




































250 

 

Finance, Accounting and Business Analysis 
Volume 7 Issue 2, 2025 

http://faba.bg/       
ISSN  2603-5324 

DOI: https://doi.org/10.37075/FABA.2025.2.10 

 

Bitcoin cyclicality and investment strategy 

 

Alejandro Rabinovich*  

  
Department of Finance, Universidad del CEMA, Buenos Aires, Argentina 

 

Info Articles   Abstract 

 
 

History Article: 

Submitted 20 July 2025. 

Revised 20 November 2025 

Accepted 5 December 2025 
 

  

Purpose: To investigate Bitcoin’s cyclic price behavior around scheduled 

halving events, develop a technical-analysis-based active investment 

strategy tailored to these cycles, and rigorously assess its performance 

relative to a passive buy-and-hold benchmark. 

Design/Methodology/Approach: This research employs historical 

daily BTC /USD price series (June 2012–May 2025), applies a suite of 

technical indicators to define systematic, halving-anchored entry and 

exit rules, and then conducts rigorous statistical evaluations to test 

whether Bitcoin’s protocol-driven supply cycles yield reproducible, 

actionable investment signals. 

Findings: Over thirteen overlapping sample windows, the active strategy 

outperforms passive BTC holding in ten, with positive “alpha” 

coefficients that are statistically significant at the conventional 5% level 

in each of those windows (and, in most cases, with p-values below 2.5%). 

It captures outsized gains in post-halving bull runs (e.g. 2013, 2017, 

2021) and meaningfully limits drawdowns in bear phases (e.g. 2014, 

2018, 2022). Equity curve simulations demonstrate compounded 

account growth that markedly surpasses passive returns. 

Practical Implications: Crypto asset managers and individual investors 

can implement the halving-centric strategy using readily available 

charting tools and API-accessible price feeds to automate buy/sell 

signals, thereby enhancing return potential and mitigating drawdowns 

without requiring deep on-chain analytics expertise. This framework also 

provides a transparent risk-management overlay—leveraging predefined 

exit rules—that can be calibrated to varying risk tolerances and 

seamlessly integrated into broader multi-asset portfolios. 

Originality/Value: This study is among the first to integrate Bitcoin’s 

protocol-driven halving schedule with a multi-indicator technical 

framework and to validate its efficacy through extensive statistical tests 

over four market cycles (including the 2024 halving). It offers 

practitioners a replicable, data-driven strategy for navigating crypto’s 

unique cyclical dynamics. 

Paper Type: Research Paper  

 

 

Keywords: Statistical 

methods, Hypothesis 

Testing, International 

Financial Markets, 

Monetary Policy 
 

 

JEL: C12, G15, E52  

Address Correspondence:  Av. Córdoba 374 
E-mail: arabinovi22@ucema.edu.ar 

 

 

  

http://faba.bg/
https://doi.org/10.37075/FABA.2025.2.10
https://orcid.org/0009-0006-8143-3167


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

251 

 

INTRODUCTION 
 

Bitcoin (BTC) has become the benchmark asset of the cryptocurrency market and a focal point in 

discussions about digital money and alternative investments. Since its launch in 2009, BTC has exhibited 

pronounced price volatility and large boom–bust cycles, which many observers link to its fixed supply 

schedule and, in particular, to the protocol-defined “halving” events that reduce the rate of new coin issuance 

approximately every four years. The first halving occurred in November 2012, followed by July 2016, May 

2020, and April 2024, and each has been associated with distinct phases of appreciation and subsequent 

correction in BTC’s price (Freeman 2025). 

A growing academic literature suggests that Bitcoin’s return dynamics are not always consistent with 

weak-form market efficiency. Studies document return predictability and momentum effects in 

cryptocurrency markets, indicating that past price behavior can have explanatory power for future returns 

(Urquhart 2016; Grobys and Sapkota 2019; Jia et al. 2022). Event-study and time-series analyses further 

show that halving events tend to coincide with systematic multi-year accumulation and distribution phases, 

in which post-halving bull markets and follow-on bear markets display relatively regular timing patterns 

(Fabus et al. 2024). Parallel work on technical trading rules finds that moving-average- and breakout-based 

strategies can generate statistically and economically significant excess returns relative to simple buy-and-

hold exposure in Bitcoin and related crypto-assets (Corbet et al. 2019; Gerritsen et al. 2020). Together, these 

strands of evidence motivate the idea that BTC’s protocol-driven supply cycle and its price dynamics may 

be amenable to rule-based exploitation. 

The present study builds on this literature by examining whether a trading strategy explicitly anchored 

to Bitcoin’s halving cycle and implemented through a set of technical indicators can outperform a passive 

BTC benchmark (Kibar et al. 2023). The objective is twofold. First, the paper characterizes BTC’s historical 

price behavior around halving events and the associated cycles. Second, it proposes and evaluates a 

transparent, rule-based strategy that combines halving timing with technical signals (such as moving 

averages and cycle-top indicators) to determine entry and exit points, and then assesses whether this strategy 

generates positive and statistically significant alpha relative to buy-and-hold. 

Methodologically, the analysis employs daily BTC/USD price data over multiple cycles and applies 

standard financial econometric tools, including regression analysis and hypothesis testing, to quantify 

performance differentials between the active strategy and the passive benchmark across overlapping sample 

windows. The study also relates the empirical findings to broader macro-financial considerations—such as 

monetary conditions, investor risk appetite, and the role of BTC as a scarce digital asset—thereby situating 

the results within both the growing academic literature on cryptocurrency markets and the practical context 

of portfolio management and investment strategy design (Singal 2023). 

 

METHODS 

 
This paper employs different theoretical tools widely used in the financial world to support or reject 

the idea behind the investment strategy. 

 

Jensen’s Alpha 
Jensen’s alpha is based on systematic risk. Any given portfolio’s systematic risk can be measured by 

estimating the market model, which is done by regressing the portfolio’s daily return on the market’s daily 

return. The coefficient on the market return is an estimate of the beta risk of the portfolio. To calculate the 

risk-adjusted return of the portfolio, it is necessary to use the beta of the portfolio and the CAPM. The 

difference between the actual portfolio return and the calculated risk-adjusted return is a measure of the 

portfolio’s performance relative to the market portfolio and is called Jensen’s alpha. By definition, α of the 

market is zero. Jensen’s alpha is also the vertical distance from the Security Market Line (SML) measuring 

the excess return for the same risk as that of the market and is given by: 

𝛼𝑝 = 𝑅𝑝 − {𝑅𝑓 + 𝛽𝑝[𝐸(𝑅𝑚) − 𝑅𝑓]} (1) 

Where: 

𝑅𝑝 − realized return of the investment 

𝑅𝑓 − risk-free rate of return for the time period 

𝑅𝑚 −  realized return of the market index 

𝛽𝑝 −   beta of the portfolio of investment with respect to the chosen market index 

 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

252 

 

If the period is long, it may contain different risk-free rates, in which case 𝑅𝑓 represents the average 

risk-free rate. Furthermore, the returns in the equation are all realized, actual returns. The sign of 𝛼𝑝 indicates 

whether the portfolio has outperformed the market. If 𝛼𝑝 is positive, then the portfolio has outperformed the 

market; if 𝛼𝑝 is negative, the portfolio has underperformed the market. Jensen’s alpha is commonly used for 

evaluating most institutional managers, pension funds, and mutual funds. Values of alpha can be used to 

rank different managers and the performance of their portfolios, as well as the magnitude of 

underperformance or overperformance. 

In this work Jensen’s Alpha is used slightly differently, given that the data analyzed is from the crypto 

market where a Rf doesn’t exist. The alpha coefficient then becomes a sort of “raw alpha”, indicating 

whether the strategy outperformed the market, or not. 

 

Regression Analysis 
Regression analysis, both the simple and multiple forms, are used by financial analysts and portfolio 

managers to examine whether a variable is useful for explaining another variable. It also allows for the use 

of hypotheses testing to examine the strength of the relationship between the variables. The variable whose 

variation is being explained is referred to as the dependent variable or explained variable, typically denoted 

by Y. Whereas, the variable used to explain the variation of the dependent variable is known as the 

independent variable, denoted by X (Drake 2023b). In the current paper the dependent variable Y is the 

investment strategy, and the independent variable X is the benchmark, BTC. Given that there is only one 

independent variable, the regression analysis used is a Simple Linear Regression (SLR) and it takes the 

following form: 

𝑌 = 𝛼 + 𝛽0𝑋 + 𝜀 
 

(2) 

Where: 

𝑌 − dependent variable 

𝑋 − independent variable 

𝛼 − Intercept 

𝛽0 − Slope coefficient 

𝜀 − residual error 

 
In the context of SLR, there are some concepts that play an important role in understanding and 

interpreting the relationship between the independent and dependent variables. These concepts are the 

Mean, Variance and Standard deviation.  

The mean is used to calculate the average return of a financial asset or investment over a specific 

period. It provides a measure of the central tendency of the data. Investors and analysts use the mean return 

to assess the historical performance of an investment or portfolio. It helps in understanding the average gain 

or loss over a given time frame. 

In SLR, the mean is often used to calculate the average values of the variables involved. For instance, 

the mean of the independent variable X and the mean of the dependent variable Y are crucial in determining 

the coefficients of the regression equation. 

𝑀𝑒𝑎𝑛(�̅�) =
∑ 𝑥𝑛

𝑖=1

𝑛
 (3) 

Where: 

𝑋 − variable value 

𝑛 − number of periods 

 
Variance measures the dispersion or spread of a set of financial returns around the mean. In finance, 

variance is used to assess the volatility or risk associated with an investment. A higher variance indicates 

greater price volatility, which is often associated with riskier investments. Investors and portfolio managers 

use variance to understand the potential fluctuations in the value of an asset. In SLR, it helps assess how 

much individual data points deviate from the mean of the dependent variable. 

𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒(𝑋) =
∑ (𝑋𝑖 − �̅�)𝑛

𝑖=1

𝑛 − 1
 (4) 

 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

253 

 

Where: 

�̅� − variable mean 

𝑋𝑖 − variable value 

𝑛 − number of periods 

 
Standard deviation is closely related to variance and is another measure of the risk or volatility of a 

financial asset. It is often preferred over variance because it is expressed in the same units as the original 

data. Investors and analysts use standard deviation to quantify the degree of uncertainty or risk associated 

with an investment. A higher standard deviation implies higher risk. 

𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 (𝑋) = √𝑉𝑎𝑟(𝑋) (5) 

Where: 

𝑉𝑎𝑟(𝑋) − variance 

 

Hypothesis testing 
In regression analysis, statistical hypothesis testing is often used to assess the significance of the 

regression coefficients, including the intercept 𝛼and the slope 𝛽0. The significance of these coefficients is 

tested using the t-statistic and the associated p-value. The null hypothesis 𝐻0 typically states that the 

coefficient is equal to zero, implying no effect, while the alternative hypothesis 𝐻𝑎  suggests that the 

coefficient is different from zero. The procedure for hypothesis testing is as follows. 

  

The first step is defining the hypotheses. 

The null hypothesis is the value assumed to be true and tested for validity. In this case, the assumption 

is that the strategy returns are similar to the benchmark, 𝐻0: 𝛼 = 0. The alternative hypothesis is everything 

that is not the null; here, that the strategy returns are different from the benchmark’s, 𝐻𝑎: 𝛼 ≠ 0. 

The second step is calculating the statistics for the test and the corresponding p-value. A generic test 

of whether a sample mean differs from a hypothesized population mean can be written as: 

𝑍 =
�̅� − 𝜇
𝜎

√𝑛⁄
 (6) 

Where: 

�̅� − sample mean 

𝜇 − population mean 

𝜎 − standard deviation 

𝑛 − number of samples 

 
The p-value associated with the t-statistic (or Z-statistic in large samples) is then used to assess the 

statistical significance of the coefficient. A low p-value, below the chosen significance level, indicates that 

the coefficient is statistically significant. 

In this study, we adopt the conventional 5% significance level (𝛼 = 0.05) for two-sided tests. 

Coefficients with p-values below 5% are therefore regarded as statistically significant. The 2.5% figure 

referred to in the analysis corresponds to the per-tail critical region (𝛼/2) of a two-tailed test at the 5% level, 

and reflects the fact that many of the estimated p-values are substantially smaller than 5%, indicating 

particularly strong evidence against the null hypothesis. 

A similar logic applies when working with proportions, where a generic Z-test can be written as: 

𝑍 =
�̂� − 𝑝0

√𝑝0(1 − 𝑝0)
𝑛

 
(7) 

Where: 

�̂� − sample proportion 

𝑝0 − assumed population proportion in the null hypothesis 

 
The third step is establishing the critical values and rejection zones of 𝐻0, taking into consideration 

Type I and Type II errors. Type I error is the incorrect rejection of a true 𝐻0(false positive), and Type II error 

is the probability of incorrectly retaining 𝐻0when it does not hold for the population (false negative). In the 

current case, the significance level 𝛼 has been set at 5%, and since it is a two-tailed test, the critical region in 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

254 

 

each tail is 𝛼/2 = 2.5%. 

 

 
Source: CFA 2023, Level 1, Volume 1 Quantitative methods 

Figure 1. Null Hypothesis rejection 

 

The fourth and final step is taking a decision based on the results. 

The critical value for rejecting 𝐻0 will be all those observations of the intercept that are in excess of 

approximately two standard deviations from zero (i.e. ∣ 𝑡 ∣> 1.96 for large samples) (Drake 2023a). The p-

value is used as an additional means of confirmation of this decision, but not as the sole criterion for the 

rejection of 𝐻0.  

 

Investment strategy 
Active vs passive approach 

There are basically two ways of approaching investments in risky assets, an active pursuit in which 

the investor seeks to be compensated by his exposure to risk by maximizing his return, called alpha. Under 

this strategy the investor will have an active role in choosing the entry and exit points of his investment, 

believing it is possible to outperform the benchmark.  And there is a passive strategy in which the investor 

believes that the performance of the benchmark cannot be beaten and therefore the strategy is simple. Invest 

in the benchmark and do not try to generate an excess return by taking opportunities during the market 

cycle. 

The strategy outlined hereafter seeks to take advantage of the key events that characterize the BTC 

market cycle and outperform the returns of a passive investment strategy. 

 
The asset and platform 

The chosen asset for this strategy is BTC paired against the US Dollar (BTC/USD). BTC is a 

decentralized digital currency and a pioneer in the world of cryptocurrencies. It was created in 2009 by an 

anonymous individual or group of individuals using the pseudonym Satoshi Nakamoto. BTC operates on a 

technology called blockchain, which is a distributed ledger that records all transactions across a network of 

computers.  

BTC was created to address various shortcomings in traditional financial systems, including 

centralization, lack of transparency, and issues related to trust and security. It aimed to provide an open, 

decentralized, and secure means of transferring value and conducting transactions in a digital world. Its 

impact has extended beyond its initial goals, influencing not only the broader cryptocurrency and blockchain 

ecosystem but also financial institutions.  Currently there are 25 BTC Spot ETFs worldwide, with 11 being 

in the US (Shen 2023). 

The chosen platform was Bitstamp, for being one of the earliest and most well-established 

cryptocurrency exchanges in the world. But also, for having one of the most complete data sets for the BTC 

pair. Founded in 2011, Bitstamp has earned a reputation for reliability and security in the cryptocurrency 

industry. Overall, the platform has played a pivotal role in the development and maturation of the 

cryptocurrency market. Its commitment to security, compliance, and user experience has made it a trusted 

platform for buying, selling, and trading cryptocurrencies for both individual and institutional investors. 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

255 

 

 

Indicators and concepts 
Simple moving average (SMA) 

Is a commonly used technical indicator in financial analysis that smooth price data over specific 

periods to identify trends. Its calculation is quite straight forward and can be calculated over different time 

frames. The formula is as follows: 

𝑆𝑀𝐴 =
∑ 𝑃𝑛

𝑖=1

𝑛
 (8) 

Where:  

P − price value. 

n − number of periods. 

 
Exponential moving average (EMA) 

Is another commonly used technical indicators in financial analysis, similar to the SMA. However, 

the key difference with the EMA, is that it gives more weight to recent price data, making it more responsive 

to recent price changes. The formula is as follows: 

𝐸𝑀𝐴 = (𝑃 ∗ 𝛼) + (𝑃𝑟𝑒𝑣𝑖𝑜𝑢𝑠 𝐸𝑀𝐴 ∗ (1 − 𝛼)) (9) 

Where:  

P − current price 

𝛼 − smoothing factor = 
2

1+𝑛
 

n − number of periods 

 
Bull market support band (BMSB) 

The bull market support band is an indicator that combines a 20 week SMA and a 21 week EMA. 

These two together create a band that has historically acted as support for the price during bull markets and 

as resistance during bear market (Senado 2023). 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 2. BMSB 

 
Pi cycle top indicator  

This indicator has gained notoriety for indicating with days difference the highs of previous market 

cycles. It combines both a daily 111 SMA and a 350 SMA x2. When the 111 SMA approaches the 350 

SMAx2 from below and crosses over, this signal indicates a market top (Swift 2019).   

 

https://www.tradingview.com/


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

256 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 3. Pi cycle indicator 

 
RSI 

The Relative Strength Index (RSI) is a popular technical indicator used alike in traditional assets and 

in cryptocurrencies. It is a momentum oscillator that measures the speed and change of price movements. 

The main feature of the RSI is that it can help investors identify overbought and oversold conditions in an 

asset, together with potential trend reversals. The formula is as follows: 

𝑅𝑆𝐼 = 100 −
100

1 + 𝑅𝑆
 (10) 

 

Where:  

RS - Relative Strength =  
Avg Gain

Avg Loss
 

 

Both Average Gain and Loss are calculated over a period of 14 consecutive days. Positive and 

negative results are summed separately and divided by 14 to obtain the RS. 

The results are going to range between 0 and 100 and the way to interpret them is the following: 

 Overbought: If the RSI is above 70 it is in the overbought region. This suggests that the asset might 

be overvalued and that a correction or reversal might be close. 

 Oversold: If the RSI is below 30, it is in the oversold region. It suggests that the asset might be 

oversold and a reversal might be possible. 

 Trend reversal: Besides the oversold or overbought regions, RSI can be used in conjunction with 

price. If divergences are forming between the price and the indicator, this might signal a bullish or 

bearish reversal. 
 

https://www.tradingview.com/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

257 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 4. RSI Indicator 

 
MACD 

The Moving Average Convergence Divergence (MACD) is used in financial analysis, including stock, 

cryptocurrency, and other asset trading. The MACD is used to analyze the strength and direction of a price 

trend and to identify potential trend reversals. It consists of three main components: 

 MACD Line (Blue Line): The MACD line is calculated by subtracting the 26-period EMA from the 

12-period EMA. The result is plotted as a continuous line on a chart. 

 Signal Line (Orange Line): The Signal line, also known as the 9-period EMA of the MACD line, is 

plotted on the same chart. It helps smooth out the MACD line and provides signals for potential buy 

or sell opportunities. 

 Histogram (Bar Graph): The Histogram is the visual representation of the difference between the 

MACD line and the Signal line. It is plotted as vertical bars on a chart. The height of each bar 

represents the divergence between the two lines. 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 5. MACD Indicator 

 
The common ways to interpret the MACD are the following: 

 Crossovers: When the MACD line crosses above the Signal line, it generates a bullish signal, 

suggesting it may be a good time to buy. Conversely, when the MACD line crosses below the Signal 

line, it generates a bearish signal, suggesting it may be a good time to sell. 

https://www.tradingview.com/?utm_source=chatgpt.com
https://www.tradingview.com/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

258 

 

 Histogram: The Histogram is used to visualize the momentum of a trend. When it moves above the 

zero line, it indicates increasing bullish momentum. When it moves below the zero line, it indicates 

increasing bearish momentum. 

 Divergence: Traders also look for divergences between the MACD and the price. For example, if 

the price is making lower lows while the MACD is making higher lows, it may signal a potential 
bullish reversal, and vice versa. 

Divergences 
Divergences in the context of cryptocurrency trading refer to a situation where the price of a 

cryptocurrency and a technical indicator (RSI, MACD or other oscillators) move in opposite directions or 

show a disparity. These divergences can provide traders with important signals about potential trend 

reversals or shifts in market sentiment (CryptoJelleNL 2022). There are two main types of divergences in 

cryptocurrency trading: bullish and bearish divergences. 

 Bullish Divergence: occurs when the price of an asset is making lower lows, but the technical 

indicator is making higher lows. This can be an early indication of a potential upward price reversal. 

It suggests that while the price is still in a downtrend, the momentum or strength of the downtrend is 

weakening, and a bullish reversal may be imminent. Bullish divergences are often seen as a buying 

signal. 

 Bearish Divergence: occurs when the price of a cryptocurrency is making higher highs, but the 

technical indicator is making lower highs. This can be a warning sign of a potential downward price 

reversal. It suggests that although the price is still in an uptrend, the momentum or strength of the 

uptrend is waning, and a bearish reversal may be approaching. Bearish divergences are often seen as 
a selling signal. 

 

BTC Halving event 

This is not an indicator, as the previously described, but rather an event that is programmed into the 

BTC protocol and occurs approximately every 4 years. Every 210,000 blocks mined, the reward that miners 

receive is halved, making the asset scarcer. The first halving occurred in 2012 when the reward went from 

50 to 25 BTCs per block. The second halving occurred in 2016, reducing the reward to 12.5 BTCs. The third 

halving occurred in 2020, bringing the reward down to 6.25 BTCs. The last BTC halving is due to occur 

around the year 2140. 

This event is of importance because it is designed to mimic the scarcity of precious metals like gold. 

By reducing the rate at which new BTCs are created, the total supply is capped at 21 million, creating a 

deflationary model. 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 6. Halvings 

 

Strategy and objectives 
The aim of the strategy outline here is to outperform the benchmark in the long term by utilizing a 

mix of signals from the indicators and concepts mentioned previously; and avoid the periods of drawdown 

https://www.tradingview.com/


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

259 

 

that BTC has become famous for. 

This strategy enters a long position – non-leveraged- and exits into cash according to the signals given 

by the indicators mentioned previously.  

The Pi cycle top indicator, on a daily frequency, identifies with great accuracy the current market 

cycle top price, thus acting as a sell signal.  

The BMSB, on a weekly frequency, acts as the trigger for entering or exiting a long position.  

The market cycle bottom range is identified by a combination of signals from different indicators, that 

as a standalone, don’t tell much. But used together give a strong signal, this is known as a confluence.  

 
Long buy criteria 

 Enter long position when price opens for two consecutive weekly candles on top of bull market 

support band. 

 Exception to rule, open long position close to market cycle bottom given by the confluence of the 

following indicators: 

o Halving event must have occurred (1 every 4 years). 

o Post halving, price has closed below the BMSB (with a 2 weekly frequency) 

o Price must be below BMSB (Weekly frequency) 

o There must be over 105 weekly candles since the latest halving. 

o Divergence in MACD histogram and BTC price (Weekly frequency) 

o Weekly RSI must have bottomed out in the oversold region of 30 and higher low structure 
confirmed on RSI. When all previous signals are confirmed, higher low RSI executes Long Buy. 

Sell Criteria 

 When weekly candle price opens below BMSB. 

 Pi cycle top signal has been confirmed after daily candle close. 

 Back-up signal. In case the Pi cycle top signal is not triggered, the combination of the following 

indicators executes a sell order close to market cycle top. For this, the weekly 60 SMA and 90 EMA 

are used in combination with the MACD. Once the halving has occurred, the cross over of the 90 

EMA on top of the 60 SMA, triggers a sell order when the MACD histogram has confirmed and 

closed red (or the blue line crossed beneath the orange). But this signal is only valid once per cycle. If 

the Pi cycle top indicator is activated, the sell order of the back-up signal is cancelled. And in case the 

back-up signal sell order is executed, it is retired for the remainder of the current cycle. It will only 

become active again after the next halving—and only if, following that halving, the 90 EMA crosses 

back above the 60 SMA. 

 Exception to rule: 

o Long position has been opened closed to mkt cycle bottom, do not sell till Pi cycle top signal or 
the back-up signal. 

 

RESULT AND DISCUSSION  

 

From scratch to results 
Although several platforms provide BTC/USD price histories, not all offer sufficiently long samples 

and there are discrepancies across exchanges. The first step was therefore to identify an exchange with a 

reliable and lengthy dataset—Bitstamp—and to write a Python script to fetch the BTC/USD OHLCV 

(Open, High, Low, Close, Volume) data. Days with missing OHLCV values or zero trading volume were 

dropped. No additional manual outlier filtering was applied: extreme returns were retained as part of the 

realized price history (see Annex, Figure 10, for the data-extraction script). 

Once obtained the daily OHLCV for BTC data set in a .csv file, this was imported and formatted in 

an excel file. Within this excel file, in a new separate sheet were consolidated the date range, close and open 

prices, all related to the benchmark, BTC. An additional column was added to calculate the percentual gain 

or loss compared to the same day opening price. Immediately, next to these, three columns were added and 

are related to the strategy per se. The first column defines the condition “Long” vs “sold”. Where value 1 

represents taking a long position in the asset and 0 represents selling into cash. The second column is the 

result of multiplying the daily returns by either the long or sold condition; returning the exact same daily 

return percent as the benchmark, in the case of “Long” condition and 0 for “Sold”. The last column provides 

the name of the key event triggering the Long or Sold. In brief, the two most important columns of this table 

are: the benchmark’s daily returns and the strategy returns. The starting point in time for both the passive 

and active strategies is the 11th of June 2012 when the long signal is confirmed, and the last day of the 

dataset is 10th of May 2025. 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

260 

 

Having obtained the returns for the benchmark and the strategy, the next step was performing the 

linear regression using the strategy daily returns as the dependent variable Y and the benchmark daily returns 

as the independent variable X. For this, in a separate excel sheet were added the results of the regression 

analysis. The first period examined was from 11th June 2012 up till 10th May 2025. After these results, 

subsequent regression results were added to the same excel sheet but moving the entry point to 

approximately 1 year after, taking the starting point as 1st of June. The second regressed period was 1st June 

2013 up till dataset end. Next period was 1st June 2014 and repeating this process up till the last examined 

period of 1st June 2023 till 10th May 2025. The idea behind performing several regression analyses with 

different entry points in time, was to have an additional valuation measure as to if the results are statistical 

significant or not.   

To obtain the accumulated results of the strategy in time - and to keep things structured - an additional 

sheet containing the same data as the second sheet, was created. In this sheet, the cumulative returns were 

calculated for both the passive approach and the active strategy. To obtain a better means of comparison the 

returns were calculated on a yearly basis.  

Lastly, on a separate excel sheet, utilizing the open & close prices in conjunction with the key events 

triggering the long or sell signals, equity curves were created for the different long/sell signals. The equity 

curves simulate the growth of the trading account, assuming 1000 US$ were invested in each the passive and 

active strategy, with no other additional injection of capital. Thus, the passive strategy remains with a 

constant amount of BTC, determined at the moment the long is triggered; whereas the active strategy 

experiences a compounding effect with the different long and sell signals. In order to have a point of 

comparison with the active strategy, in terms of account value expressed in US$, the value of the passive 

account was also calculated at the date of long and sell signals.      

 

Strategy Conclusions 
The aim of this work was to explore and analyze with statistical tools if the proposed active strategy 

could beat the performance of the benchmark, BTC. The short answer is that the strategy is successful. 

Furthermore, these results align with empirical findings that Bitcoin exhibited periods of weak‑form 

inefficiency in its earlier years, providing conditions under which rule‑based active strategies can generate 

statistically significant excess returns (Urquhart 2016). 

 
Regression Analysis Results 

Based on the regression results, the null hypothesis stating that the average daily returns of the strategy 

are no different from the benchmark’s is rejected in favor of the alternative hypothesis in most of the 

examined sample windows. For the full 6/2012–5/2025 sample, the regression explains about 57% of the 

variation in daily returns (adjusted R² ≈ 0.57; F-statistic ≈ 6,188, p < 0.001), and the estimated intercept α 

is economically and statistically significant at roughly 0.17% per day, with a 95% confidence interval of 

about 0.12%–0.23% (Annex Table 8). Table 1 reports 13 overlapping regressions, with start dates from June 

2012 through June 2024 and a common end date in May 2025. Under conventional OLS inference, the 

estimated intercept α is positive and statistically significant at the 5% level in 10 of the first 11 windows 

(6/2012–5/2025 through 6/2022–5/2025), with t-statistics comfortably above the 1.96 threshold and very 

small p-values (typically below 1%). In the short 6/2021–5/2025 window α remains positive but is not 

significant at conventional levels, reflecting the reduced number of observations in this subsample. The final 

two windows (6/2023–5/2025 and 6/2024–5/2025) are also included in Table 1 for transparency. In these 

most recent periods the strategy and benchmark returns are almost perfectly collinear, so the regression is 

numerically ill-conditioned: the estimated α is extremely close to zero and the resulting test statistics are not 

economically meaningful. Rather than omitting these subsamples, they are reported explicitly in Table 1 and 

Annex Tables 19–20, with standard errors and confidence intervals suppressed and interpreted with caution. 

Because the regression windows are overlapping, the estimated α coefficients across the 13 

subsamples are not statistically independent. Sequential windows share a large proportion of observations, 

which induces dependence among their test statistics. For this reason, the overlapping OLS results in Table 

1 should be interpreted as descriptive evidence of persistence rather than as a series of independent 

hypothesis tests. To complement these overlapping regressions with a non-overlapping robustness check, a 

walk-forward out-of-sample validation is also performed, where each test window uses only data not 

included in the corresponding training window. 

 

 

 

 

 

Table 1. Regression results 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

261 

 

Period Coefficient α Standard Error t Stat >2 P-value < 2.5% 

6/2012 - 5/2025 0.00175 0.000299 5.85111 0.00000001 

6/2013 - 5/2025 0.00149 0.000288 5.15492 0.00000026 

6/2014 - 5/2025 0.00128 0.000269 4.74974 0.00000211 

6/2015 - 5/2025 0.00122 0.000271 4.49184 0.00000728 

6/2016 - 5/2025 0.00131 0.000298 4.40390 0.00001097 

6/2017 - 5/2025 0.00135 0.000329 4.11187 0.00004034 

6/2018 - 5/2025 0.00097 0.000309 3.14394 0.00168621 

6/2019 - 5/2025 0.00075 0.000311 2.42628 0.01533538 

6/2020 - 5/2025 0.00102 0.000350 2.92183 0.00352306 

6/2021 - 5/2025 0.00067 0.000376 1.77202 0.07660227 

6/2022 - 5/2025 0.00062 0.000298 2.06180 0.03946733 

6/2023 - 5/2025 0.00000 - - - 

6/2024 - 5/2025 0.00000 - - - 

Source: Authors’ data 

Note: During 6/2023–5/2025 and 6/2024–5/2025 the strategy’s daily returns are almost perfectly collinear 

to the benchmark by construction, so the regression is numerically degenerate. In these windows α is 

mechanically zero and conventional standard errors, t-statistics, p-values and confidence intervals are not 

reported. 

 
Residual diagnostics and robust inference 

To assess whether the OLS assumptions underlying Table 1 are appropriate, residual diagnostics 

were conducted on the full 6/2012–5/2025 sample. The Durbin–Watson statistic for the regression residuals 

is 2.10, which is close to the theoretical value of 2 under no first-order autocorrelation. However, a Ljung–

Box-type Q-statistic at lag 20 of 50.95 (p ≈ 2.3×10⁻¹⁰) indicates that, taken jointly, the residuals exhibit 

statistically significant autocorrelation at higher lags. This suggests that, while there is no strong single lag-

1 effect, serial dependence is present in the error structure. 

Heteroskedasticity was examined using both the Breusch–Pagan and White tests. The Breusch–

Pagan auxiliary regression of squared residuals on the benchmark return produces R² = 0.000075, an LM 

statistic of 0.35 and p = 0.55, so a simple linear relationship between the conditional variance and BM% is 

not supported. In contrast, the White test yields R² = 0.938, LM = 4424.91 with 2 degrees of freedom and p 

< 0.0001, strongly rejecting homoskedasticity in favor of a more general form of heteroskedasticity. This 

outcome is consistent with the well-known volatility clustering observed in Bitcoin returns and suggests that 

OLS standard errors are likely to be understated. 

To obtain more reliable inference in the presence of both autocorrelation and heteroskedasticity, 

Newey–West heteroskedasticity- and autocorrelation-consistent (HAC) standard errors were computed for 

the intercept α in each regression window (Annex Table 5). For the full sample (6/2012–5/2025), the OLS 

estimate of α is 0.00175 with a t-statistic of 5.85 (p ≈ 1×10⁻⁸). When Newey–West HAC standard errors are 

used, the standard error of α increases from 0.000299 to 0.000410 and the t-statistic decreases to 4.27 (p ≈ 

2×10⁻⁵), which still represents strong statistical evidence of a positive intercept. Across the pre-2023 

windows (6/2012–5/2025 through 6/2022–5/2025), α remains positive in all cases and statistically 

significant at the 5% level in 9 out of 11 regressions under HAC inference. The two exceptions are the very 

short 6/2021–5/2025 and 6/2022–5/2025 samples, where the reduced number of observations and elevated 

volatility make it more difficult to distinguish α from zero once serial dependence and heteroskedasticity are 

accounted for. Overall, the HAC results confirm that the main conclusion—a positive and economically 

meaningful α over the full sample and most subperiods—is robust to more stringent statistical assumptions. 

Viewed together, the residual diagnostics, HAC-based inference and walk-forward validation indicate that 

the strategy’s excess returns are statistically robust in most of the examined periods, while also making clear 

that the evidence is sample-specific and that performance varies across market regimes. Overall, these 

diagnostics, combined with the well-known volatility clustering in Bitcoin returns, are consistent with the 

view that cryptocurrency markets are non-stationary and subject to frequent external shocks (e.g. regulatory 

announcements, exchange-specific incidents, and macroeconomic policy shifts), which limits the predictive 

stability of any single rule-based strategy. This motivates a cautious interpretation: the results provide strong 

historical evidence consistent with a positive α for this halving-centric strategy, rather than a guarantee of 

persistent arbitrage or a recommendation of the strategy as investment advice. 
 

Out-of-sample walk-forward validation 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

262 

 

In addition to the in-sample regressions, a simple walk-forward analysis was implemented to 

evaluate the strategy’s performance in a genuinely out-of-sample setting. Using the June–May convention 

for “years”, rolling 3-year windows were treated as training periods and the following June–May year as a 

test period. This procedure yields ten non-overlapping test windows from 2014–2015 up to 2023–2024 (see 

Annex Table 6). The trading rules themselves are fixed ex ante; the walk-forward split is simply a way of 

asking how the strategy would have performed if each test year had not been used in calibrating the model. 

Across these ten out-of-sample test years, the strategy delivers a positive annualized test return in nine out 

of ten cases. The average annualized out-of-sample return of the strategy is approximately 108.9% per year, 

compared with about 67.3% for the buy-and-hold benchmark over the same test windows. The largest 

relative gains occur in the 2017, 2018, 2020 and 2022 test years, where the strategy both amplifies major bull 

markets and either cushions or reverses benchmark losses. In other years, such as 2015, 2016, 2019, 2023 

and 2024, the strategy closely tracks the benchmark, so it does not introduce significant negative drag when 

its signals are less distinctive. These walk-forward results therefore support the view that the observed 

outperformance is not solely an artefact of in-sample fitting to a single long window, but persists—albeit 

with variability—across a sequence of genuinely out-of-sample periods. 

 
Cumulative Return Comparison 

When examining the yearly cumulative returns of Benchmark versus the strategy several things can 

be noticed. The first one, that when BTC rallies, the strategy outperforms the returns of a passive approach, 

as seen from years 2013, 2017 and 2021. The second notorious is that the strategy is also effective in limiting 

the negative results, as observed from the results for years 2014, 2018 and 2022. The third thing that can be 

appreciated from the returns of the benchmark, is a cyclical pattern recurring every four years. This pattern 

begins with the halving event (years 2012, 2016 and 2020), continues with a bull market phase (years 2013, 

2017 and 2021) and ensues with a bear market period (years 2014, 2018 and 2022). 

This return behavior is consistent with evidence that momentum factors are strong and persistent 

drivers of cryptocurrency performance, explaining significant portions of return variation beyond market 

movements alone (Jia et al. 2022). 

 
Table 2. Cumulative returns 

Year Comp. ret BM Comp. ret Stg 

2012 142.86% 142.86% 

2013 5753.34% 8767.95% 

2014 -61.37% -30.23% 

2015 34.39% 89.95% 

2016 130.66% 130.66% 

2017 1435.09% 2012.52% 

2018 -72.84% -3.60% 

2019 88.80% 88.80% 

2020 289.55% 289.55% 

2021 59.90% 117.66% 

2022 -63.88% -23.33% 

2023 154.52% 154.52% 

2024 122.71% 122.71% 

2025 11.93% 11.93% 

Source: Authors’ data 

 

Equity Curve Analysis 

The equity-curve analysis evaluates how a hypothetical trading account would have evolved under 

the active strategy compared with a passive BTC benchmark. Table 3 reports the end-of-period value of a 

USD 1,000 account invested in BTC at the start of each sample window (triggered by a “buy” signal), the 

corresponding benchmark end capital (“BM End capital”), the benchmark return (“BM ROI”), and the 

excess return generated by the strategy over the benchmark. Across all selected periods, the strategy delivers 

a higher terminal account value than passive holding. 

These findings are consistent with prior evidence that technical, trend-following rules can be profitable 

in cryptocurrency markets. Empirical studies show that moving-average-based and breakout-style trading 

systems can generate economically meaningful abnormal returns in Bitcoin, even after accounting for 

transaction costs and employing robust statistical procedures such as bootstrap inference (Corbet et al. 2019; 

Gerritsen et al. 2020). The equity-curve results reported here align with this literature by illustrating that a 

halving-anchored, indicator-driven strategy can systematically outperform a simple buy-and-hold 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

263 

 

benchmark over multiple overlapping windows. 

 

Table 3. Equity curve comparison 

Period BM End capital (in USD) BM ROI Strategy excess return on BM 

11/6/2012 - 10/5/2025 19,125,730  1912473% 5379% 

2/9/2013 - 10/5/2025 802,212  80121% 4948% 

6/6/2014 - 10/5/2025 159,284  15828% 3144% 

9/2/2015 - 10/5/2025 468,064  46706% 1362% 

31/12/2018 - 10/5/2025 27,358  2636% 192% 

16/8/2021 - 10/5/2025  2,229  123% 129% 

4/4/2022 - 10/5/2025 2,258  126% 115% 

Source: Authors’ data 

 

Viewed together with the regression analysis, the equity-curve evidence indicates that the strategy’s 

returns differ from the benchmark in a statistically and economically significant way over most of the 

examined periods. The estimated intercepts (alphas) are generally positive and significant at conventional 

levels, while the equity curves show that the active strategy compounds capital more effectively than the 

passive benchmark. A complementary examination of yearly cumulative returns further suggests that the 

strategy tends to participate strongly in major BTC bull runs and to limit losses during bear phases, in line 

with the documented cyclical behavior of cryptocurrency markets. 

At the same time, it is important to adopt a cautious interpretation of these results. The documented 

outperformance is conditional on the specific sample period, the chosen data source, and the particular set 

of modelling choices and trading rules implemented in this study. Future market regimes—characterized by 

different volatility, liquidity, regulatory environments, or macroeconomic conditions—may not replicate the 

historical patterns observed here, and strategy performance could deteriorate accordingly. Consequently, the 

findings are best viewed as an empirical case study of one halving-centric implementation within the broader 

family of technical trading strategies, rather than as evidence of persistent arbitrage opportunities or as direct 

investment advice. 

 

Macroeconomic Correlations 

As demonstrated by the statistical analysis, applying the active strategy during the selected period 

would have outperformed BTC with statistical significance. Although not related to the strategy, there is one 

question that deserves some consideration. Why has BTC experienced such incredible returns? While not 

exclusive, some of the arguments as to why BTC could have experienced such phenomenal periods of 

growth, are the following ones. 

 
Market capitalization (mkt cap) 

In July 2010 BTC had a mkt cap of slightly under 250,000 USD, by July 2013 it grew to 1.5 B and by 

the time of its first massive bull run mkt cap ascend to 13.6 B. Four years later, by the time of the second 

massive bull run on Dec 2017 mkt cap ascended to 320 B, only to drop to 56.4 B one year later during the 

bear mkt cycle. At the height of its latest bull run on Nov 2021, mkt cap reached 1.3 T USD for BTC, whereas 

for the remaining crypto space it amounted to 1.7 T. Current BTC mkt cap, as per the time of writing, is 

around 2.08 T and remaining crypto assets mkt cap amounts to 1.15 T, 3.24T combined. Whereas just to 

give a means of comparison, the mkt cap of Microsoft, currently the highest company by mkt cap, is of 3.42 

T. Silver stands at 1.86T and Gold at 22.24 T (CompaniesMarketCap n.d.).  

 

Table 4. Market capitalization by market 

Markets by mkt cap (in Trillion USD)  as of May 2025 

Global Bond (as of 2023 EOY) 140 

Global Equity (as of 2023 EOY) 115 

S&P 500 (as of May 2025) 49.8 

Gold 22.24 

Microsoft  3.43 

Global crypto 3.23 

Silver 1.86 

BTC 2.08 

Source: Authors’ Data 

 

The market cap of the S&P 500 currently stands at about 49.8 T, the global equity market is roughly 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

264 

 

115 T (Kolchin et al. 2024) and global bond markets, as of 2023, was 140 T (Neufeld 2023). 

It is important to remember that BTC is, after all, a new asset that has been in existence for only 16 

years, compared to more mature markets mentioned previously. Given that BTC is an asset that is traded 

24/7 and with global exposure it is essential to try to understand how global liquidity could affect the BTC 

price dynamics. Particularly during earlier phases when mkt cap was relatively small, and could be easily 

pumped by the inflow of institutional investors or private companies that decided to buy and hold BTC as 

part of the assets within their balance sheet (Benzinga 2023). This scenario occurred during the last cycle 

that saw BTC price propelled to 67,000 USD per BTC. Microstrategy (MSTR) is the top reference, currently 

holding over 580,250 BTC. Also equally important, is the fact that some governments hold BTC in their 

balance sheet and some countries, such as El Salvador and Central African Republic, have adopted BTC as 

legal tender currency. Recently the US established a strategic reserve of crypto assets, becoming the 

government with the largest quantity of BTC holdings (207,189), China follows suit with approximate 

holdings of 194,000 BTC and other governments are showing interest in doing the same. All of the 

aforementioned have a direct impact in the demand for BTC and thus have an impact in the price dynamics 

(Del Castillo 2023).  

 
Inflation and interest rates 

BTC is a consequence of the financial crisis of 2008, one of its purposes is to be a store of value and 

a hedge against inflation. Maybe it is a mere coincidence, but it is interesting to observe that the three bull 

rallies occur while inflation numbers (in the US) were on the rise and that the respective bear cycles take 

place while inflation numbers decrease. Another possible factor that could have contributed to BTC’s growth 

and the bull periods, are the US Fed’s interest rates and its economic impact during these periods. The 

growth cycles occur while interest rates are at their lowest value in decades, sub 0.5 points. Rates started 

going up since Nov 2016 till they plateaued at 2.5 on Dec 2018, also marking the bottom for the 2018 bear 

cycle. Then on Feb 2020, the Covid-19 black swan event triggered the return of low interest rates of 0.25 

points. Which could have benefited institutional investors by having access to cheap money for investment 

purposes.  

Expanding on the aforementioned, empirical studies have shown that Bitcoin exhibits time‑varying 

inflation‑hedging properties, acting as a partial hedge during periods of monetary expansion and elevated 

inflation uncertainty (Bouri et al. 2017). 

As of time of writing, May 2025, the FED is still applying quantitative tightening but there is 

increasing pressure to switch towards quantitative easing. When this happens, the crypto markets might 

benefit from the tail wind and this might translate into appreciation of price (Trading Economics n.d.). 

The figure below illustrates in a graphical sense what was happening and when. Yellow indicates the 

halving event, green the top of cycle and red, the bottom of it. 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 7. BTC’s timeline along FED’s interest rate & US inflation rate 

Inverse relation with the DXY U.S. Dollar Index (USDX) 

https://www.tradingview.com/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

265 

 

Curiously enough, throughout BTC’s history there has been an inverse relationship with the DXY. 

Whilst also occurring in other periods, this inverse relationship is not as significant as in the months 

following the halving event. This could well be a mere coincidence, or it could be an indicative of global 

economic factors. Under this premise, it is of interest studying the relationship between BTC and the DXY, 

given that the later factors in 7 major currencies: the US dollar, Euro, Japanese Yen, British Pound Sterling, 

Canadian Dollar, Swedish Krona and Swiss Franc.  

At its core the index reflects the appreciation or depreciation of the US dollar versus the other major 

currencies in the basket, therefore its use as an economic indicator. Under this light, an inverse relation 

between BTC and the DXY could be explained by some of the objectives of BTC: 

 Acting as a safe haven: BTC was conceived as a store of value and a form of digital gold. Inspired 

by the 2008 financial crisis, it was designed to act as a safe heaven during times of uncertainty when 

there is a lack of confidence in traditional financial markets.  

 Inflation Hedge: Under an inflationary context, when there are doubts about currency devaluation 

and rising inflation, investors might take shelter in an asset like BTC, that has a capped supply and is 

not subject to central bank policies. 

In line with the discussed, previous research also documents that Bitcoin tends to display a negative 

correlation with the U.S. Dollar Index, especially in risk‑on environments, reinforcing the relevance of 

dollar‑driven macro cycles for BTC valuation (Dyhrberg 2016). 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 8. BTC-DXY inverse relation 

 

Figure 8 illustrates the inverse relationship that has been observed between BTC and the DXY after 

the halving event or shortly before. The figure plots in yellow the halving event, color green and red are 

used to show the percentual increases or decreases for the same time window in BTC and DXY.  

 
Halvings effect 

The halving might mistakenly be overlooked as a simple event that reoccurs every four years and that 

just halves BTC’s miners reward in half. In fact, the BTC halving is one of, if not, the most important 

characteristic of the asset, with profound implications for its economic model. The price fluctuations of the 

asset in the weeks prior and after the halving can be explained by the following:   

Supply Reduction: the reward that miners receive for validating and adding new blocks to the 

blockchain is reduced by half. Initially, when BTC was launched in 2009, miners received 50 BTCs per 

block. The first halving occurred in 2012, reducing the reward to 25 BTCs. The second halving occurred in 

2016, reducing it further to 12.5 BTCs. The third halving occurred in 2020, reducing the reward to 6.25 

BTCs. The fourth halving occurred in 2024, reducing the reward to 3.12 BTCs This reduction in the rate of 

new BTC creation is designed to control its overall supply. 

Scarcity and Deflationary Nature: BTC's total supply is capped at 21 million coins. By halving the 

reward every four years, the rate at which new BTCs are created slows down over time. This controlled 

https://www.tradingview.com/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

266 

 

issuance creates a sense of scarcity, similar to precious metals like gold. The idea is that as the supply 

becomes more limited, and if demand remains constant or increases, the value of each BTC could rise. 

Market Perception and Speculation: Traders, investors, and the broader market pay close attention 

to the halving events. The anticipation of reduced new supply often leads to increased speculation about 

potential price increases. This heightened interest can lead to increased demand in the period leading up to 

and following a halving, affecting the market dynamics. 

Previous empirical halving analyses also document systematic post‑halving appreciation cycles, 

strengthening the argument that Bitcoin’s issuance schedule plays a central role in shaping long‑term price 

dynamics (Fabus et al. 2024). 

In the figure below, the yellow lines mark when the halving occurred and the consecutive growth in 

BTC’s price after the event. 

 

 
Source: Authors’ data (https://www.tradingview.com/ platform) 

Figure 9. BTC’s halvings 

 

CONCLUSION  
 

This study has examined whether a halving-anchored, technically driven trading strategy can 

outperform a passive buy-and-hold exposure to Bitcoin. Using multiple overlapping sample windows, the 

empirical results indicate that the proposed strategy generates positive and statistically significant alpha 

relative to a BTC benchmark in most periods, and that the associated equity curves display superior 

compounded growth. These findings suggest that Bitcoin’s protocol-driven supply schedule and its cyclical 

price behavior can be translated into systematic trading rules with economically meaningful performance 

over the historical sample considered. 

Beyond the strategy itself, the analysis highlights several potential drivers of BTC’s long-term growth 

that warrant further investigation. First, questions remain about market structure and price formation, 

especially in Bitcoin’s early years when market capitalization was low and formal regulation was limited or 

absent. Under such conditions, the possibility of price manipulation, including pump-and-dump dynamics, 

cannot be ruled out and deserves dedicated study, particularly in light of the growing influence of large 

institutional players and the global reach of the asset. Second, macroeconomic forces and investor sentiment 

appear to interact with Bitcoin’s cycles: episodes of elevated inflation, shifting interest-rate regimes, and 

changing risk appetite may amplify or dampen BTC’s performance, suggesting that macro-financial 

conditions are an important part of the broader narrative. Third, the halving mechanism itself may shape 

investors’ value perceptions and expectations, potentially giving rise to recurring accumulation and 

distribution phases that extend beyond the immediate supply shock. 

From an academic standpoint, the main contribution of this study is to document, over several 

overlapping windows, that a halving-anchored set of technical rules can generate positive and statistically 

significant alpha relative to a passive BTC benchmark, and to outline macro-financial channels—such as 

liquidity conditions, monetary policy regimes, and protocol-driven supply shocks—through which such 

patterns may arise. Importantly, the analysis is intended as a contribution to the empirical literature on 

Bitcoin cyclicality and investment strategies, and not as personalized investment advice. 

https://www.tradingview.com/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

267 

 

From an institutional portfolio-management perspective, these findings are best interpreted as a case 

study in designing rule-based, risk-managed crypto exposures that can be slotted into diversified multi-asset 

portfolios, rather than as a stand-alone trading mandate. The emphasis on transparent rules, explicit 

drawdown control and long-horizon evaluation is also consistent with the broader sustainable-fintech 

agenda, in which digital-asset strategies are engineered to balance innovation with risk management, 

governance and investor protection. 

 

ACKNOWLEDGEMENTS 

  

I would like to thank Prof. Dr. José P. Dapena for guidance on the original work, Dr. Maximiliano 

Ivickas Magallan for recognizing its potential as an academic paper, and Lic. Bahía Solla Rouquaud for 

assistance during the publication process. 

 

REFERENCES 

Benzinga. 2023. 10 Public Companies With Largest Bitcoin Holdings in 2023. Investing.com, March 6, 2023. 

Accessed May 2025. https://uk.investing.com/news/stock-market-news/10-public-companies-with-

largest-bitcoin-holdings-in-2023-2939608. 

Bouri, E., P. Molnár, G. Azzi, D. Roubaud, and L. I. Hagfors. 2017. On the Hedge and Safe Haven 
Properties of Bitcoin: Is It Really More Than a Diversifier? Finance Research Letters, 20: 192–198. 

https://doi.org/10.1016/j.frl.2016.09.025. 
CompaniesMarketCap. n.d. Top Assets by Market Cap. CompaniesMarketCap.com. Accessed April 2025. 

https://companiesmarketcap.com/assets-by-market-cap/. 

Corbet, S., V. Eraslan, B. M. Lucey, and A. Sensoy. 2019. The Effectiveness of Technical Trading Rules in 
Cryptocurrency Markets. Finance Research Letters, 31: 32–37. 

https://doi.org/10.1016/j.frl.2019.04.027. 

CryptoJelleNL. 2022. How to Use the Market Structure in Trading. CoinMarketCap Academy. Accessed April 

2025. https://coinmarketcap.com/academy/article/how-to-use-the-market-structure-in-trading. 

Del Castillo, M. 2023. U.S. Government Owns Way More Bitcoin Than Any Other Country—So Why 
Aren’t They Selling It? Forbes, June 16, 2023. Accessed April 2025. 

https://www.forbes.com/sites/michaeldelcastillo/2023/06/16/us-government-owns-way-more-

bitcoin-than-any-other-countryso-why-arent-they-selling-it/.  
Drake, P. P. 2023a. Reading 6: Hypothesis Testing. In CFA Program Curriculum 2023 Level I, Volume 1: 

Quantitative Methods. CFA Institute. 

Drake, P. P. 2023b. Reading 7: Introduction to Linear Regression. In CFA Program Curriculum 2023 Level I, 

Volume 1: Quantitative Methods. CFA Institute. 

Dyhrberg, A. H. 2016. Bitcoin, Gold and the Dollar–A GARCH Volatility Analysis. Finance Research Letters, 

16: 85–92. https://doi.org/10.1016/j.frl.2015.10.008. 

Fabus, J., I. Kremenova, N. Stalmasekova, and T. Kvasnicova-Galovicova. 2024. An Empirical 

Examination of Bitcoin’s Halving Effects: Assessing Cryptocurrency Sustainability within the 
Landscape of Financial Technologies. Journal of Risk and Financial Management, 17(6): 229. 

https://doi.org/10.3390/jrfm17060229. 
Freeman, J. 2025. The Four-Year Bitcoin Halving Cycle. eToro Academy. Accessed April 2025. 

https://www.etoro.com/crypto/bitcoin-four-year-cycle/. 

Gerritsen, D. F., E. Bouri, E. Ramezanifar, and D. Roubaud. 2020. The Profitability of Technical Trading 

Rules in the Bitcoin Market. Finance Research Letters, 34: 101263. 

https://doi.org/10.1016/j.frl.2019.08.011. 

Grobys, K., and N. Sapkota. 2019. Cryptocurrencies and Momentum. Economics Letters, 180: 6–10. 

https://doi.org/10.1016/j.econlet.2019.03.028. 

Jia, B., J. W. Goodell, and D. Shen. 2022. Momentum or Reversal: Which Is the Appropriate Third Factor 
for Cryptocurrencies? Finance Research Letters, 45: 102139. https://doi.org/10.1016/j.frl.2021.102139. 

Kibar, A., B. Sine, and R. Strong. 2023. Learning Module 4: Technical Analysis. In CFA Program Curriculum 

2023 Level I, Volume 6: Portfolio Management and Ethical and Professional Standards. CFA Institute. 

Kolchin, K., J. Romulus, and M. Paluzzi. 2024. 2024 Capital Markets Fact Book. Securities Industry and 

Financial Markets Association (SIFMA), July 2024. Accessed April 2025. 

https://www.sifma.org/wp-content/uploads/2023/07/2024-SIFMA-Capital-Markets-

Factbook.pdf. 
Neufeld, D. 2023. Ranked: The Largest Bond Markets in the World. World Economic Forum / Visual 

Capitalist, April 17, 2023. Accessed April, 2025. 

https://www.weforum.org/stories/2023/04/ranked-the-largest-bond-markets-in-the-world/. 

https://uk.investing.com/news/stock-market-news/10-public-companies-with-largest-bitcoin-holdings-in-2023-2939608?utm_source=chatgpt.com
https://uk.investing.com/news/stock-market-news/10-public-companies-with-largest-bitcoin-holdings-in-2023-2939608?utm_source=chatgpt.com
https://doi.org/10.1016/j.frl.2016.09.025
https://companiesmarketcap.com/assets-by-market-cap/?utm_source=chatgpt.com
https://doi.org/10.1016/j.frl.2019.04.027
https://coinmarketcap.com/academy/article/how-to-use-the-market-structure-in-trading?utm_source=chatgpt.com
https://www.forbes.com/sites/michaeldelcastillo/2023/06/16/us-government-owns-way-more-bitcoin-than-any-other-countryso-why-arent-they-selling-it/
https://www.forbes.com/sites/michaeldelcastillo/2023/06/16/us-government-owns-way-more-bitcoin-than-any-other-countryso-why-arent-they-selling-it/
https://doi.org/10.1016/j.frl.2015.10.008
https://doi.org/10.3390/jrfm17060229
https://www.etoro.com/crypto/bitcoin-four-year-cycle/?utm_source=chatgpt.com
https://doi.org/10.1016/j.frl.2019.08.011
https://doi.org/10.1016/j.econlet.2019.03.028
https://doi.org/10.1016/j.frl.2021.102139
https://www.sifma.org/wp-content/uploads/2023/07/2024-SIFMA-Capital-Markets-Factbook.pdf?utm_source=chatgpt.com
https://www.sifma.org/wp-content/uploads/2023/07/2024-SIFMA-Capital-Markets-Factbook.pdf?utm_source=chatgpt.com
https://www.weforum.org/stories/2023/04/ranked-the-largest-bond-markets-in-the-world/?utm_source=chatgpt.com


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

268 

 

Senado, X. 2023. Bull Market Support Band — Market Update: July 07, 2023. Medium (Coinmonks), July 7, 

2023. Accessed April 2025. https://medium.com/coinmonks/bull-market-support-band-market-

update-july-07-2023-fc06463a63ec. 
Shen, C. 2023. What Europe’s New Spot Bitcoin ETF Means for Global Markets. Forkast, September 18, 

2023. Accessed May 2025. https://forkast.news/what-europes-spot-bitcoin-etf-means-for-markets/. 
Singal, V. 2023. Learning Module 3: Portfolio Risk and Return, Part II. In CFA Program Curriculum 2023 

Level I, Volume 5: Fixed Income, Derivatives, Alternative Investments, and Portfolio Management. CFA 

Institute. 
Swift, P. 2019. The Golden Ratio Multiplier. Medium, March 18, 2019. Accessed May 2025. 

https://positivecrypto.medium.com/the-golden-ratio-multiplier-c2567401e12a. 
Trading Economics. n.d. United States Fed Funds Interest Rate. TradingEconomics.com. Accessed May 2025. 

https://tradingeconomics.com/united-states/interest-rate. 

Urquhart, A. 2016. The Inefficiency of Bitcoin. Economics Letters, 148: 80–82. 

https://doi.org/10.1016/j.econlet.2016.09.019. 

 

 

 

 

 
 
 
 
 
 
 
 
 
 
 

 

 

 

  

https://medium.com/coinmonks/bull-market-support-band-market-update-july-07-2023-fc06463a63ec?utm_source=chatgpt.com
https://medium.com/coinmonks/bull-market-support-band-market-update-july-07-2023-fc06463a63ec?utm_source=chatgpt.com
https://forkast.news/what-europes-spot-bitcoin-etf-means-for-markets/?utm_source=chatgpt.com
https://positivecrypto.medium.com/the-golden-ratio-multiplier-c2567401e12a?utm_source=chatgpt.com
https://tradingeconomics.com/united-states/interest-rate?utm_source=chatgpt.com
https://doi.org/10.1016/j.econlet.2016.09.019


Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

269 

 

ANNEXES 

1. # Script to fetch data from Bitstamp 

2. import json, requests, datetime 

3. import pandas as pd 

4.  

5. #Parameters 

6. currency_pair = "btcusd" 

7. url = f"https://www.bitstamp.net/api/v2/ohlc/{currency_pair}/" 

8.  

9. # Convert dates to Unix timestamps 

10. start_date = pd.Timestamp("2012-03-01").timestamp() 

11. end_date = pd.Timestamp("2025-05-10").timestamp() 

12.  

13. step = 86400  # 1 day 

14. limit = 1000  # max candles per API call 

15.  

16. master_data = [] 

17. current_start = int(start_date) 

18. while current_start < end_date: 

19.     current_end = current_start + (step * limit) 

20.     if current_end > end_date: 

21.         current_end = int(end_date) 

22.  

23.     print(f"Fetching from {datetime.datetime.utcfromtimestamp(current_start)} " 

24.           f"to {datetime.datetime.utcfromtimestamp(current_end)}") 

25.  

26.     params = {"step": step, 

27.         "limit": limit, 

28.         "start": current_start, 

29.         "end": current_end,} 

30.  

31.     try: 

32.         response = requests.get(url, params=params) 

33.         if response.status_code != 200: 

34.             print(f"Failed at {current_start}: Status {response.status_code}") 

35.             break 

36.         json_data = response.json() 

37.         ohlc = json_data.get("data", {}).get("ohlc", []) 

38.         master_data += ohlc 

39.     except Exception as e: 

40.         print(f"Exception at {current_start}: {e}") 

41.         break 

42.  

43.     current_start = current_end 

44.     time.sleep(1) 

45.  

46. # Create DataFrame 

47. df = pd.DataFrame(master_data) 

48. df = df.drop_duplicates() 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

270 

 

49. # Format date 

50. df["timestamp"] = df["timestamp"].astype(int) 

51. df = df.sort_values(by="timestamp") 

52. df["date"] = pd.to_datetime(df["timestamp"], unit='s').dt.strftime("%d/%m/%Y") 

53. # Save to Drive 

54. output_path = '/content/drive/MyDrive/Data/BTCUSDohlcv.csv' 

55. df.to_csv(output_path, index=False) 

56.  

57. print('End of data fetching. Saved to:', output_path) 

Authors’ data (google colabs) 

Figure 10. Python script to fetch data 

 

58. # Script for Newey-West HAC 

59. # Drive mount 

60. from google.colab import drive 

61. drive.mount('/content/drive') 

62.  

63. import pandas as pd 

64. import datetime as dt 

65. import statsmodels.api as sm 

66.  

67. # Adjust path to file directory in Drive 

68. excel_path = "/content/drive/MyDrive/Strategy results 2025.xlsx" 

69. # Load and clean data 

70. raw = pd.read_excel(excel_path, sheet_name="Strat + Acc ret", header=2) 

71. header_row = raw.iloc[0] 

72. col_map = { 

73.     "Benchmark (BTC)": header_row["Benchmark (BTC)"],  # Date 

74.     "Unnamed: 2": header_row["Unnamed: 2"],            # close 

75.     "Unnamed: 3": header_row["Unnamed: 3"],            # open 

76.     "Unnamed: 4": header_row["Unnamed: 4"],            # BM % 

77.     "Strategy": "Long/Sold", 

78.     "Unnamed: 6": header_row["Unnamed: 6"],            # Strategy % 

79.     "Unnamed: 7": header_row["Unnamed: 7"],            # Key event 

80. } 

81. df = raw.iloc[1:].reset_index(drop=True) 

82. df = df.rename(columns=col_map) 

83. df = df[["Date", "BM %", "Strategy %"]].copy() 

84. df["Date"] = pd.to_datetime(df["Date"]) 

85. df["BM %"] = pd.to_numeric(df["BM %"], errors="coerce") 

86. df["Strategy %"] = pd.to_numeric(df["Strategy %"], errors="coerce") 

87. df = df.dropna() 

88.  

89. df["Date_only"] = df["Date"].dt.date 

90. data_start = df["Date_only"].min() 

91. data_end = df["Date_only"].max() 

92. print("Data from", data_start, "to", data_end, "rows:", len(df)) 

93. # Define 13 sampling periods (including 2023 & 2024) 

94. period_specs = [ 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

271 

 

95.     ("6/2012 - 5/2025", dt.date(2012, 6, 1)), 

96.     ("6/2013 - 5/2025", dt.date(2013, 6, 1)), 

97.     ("6/2014 - 5/2025", dt.date(2014, 6, 1)), 

98.     ("6/2015 - 5/2025", dt.date(2015, 6, 1)), 

99.     ("6/2016 - 5/2025", dt.date(2016, 6, 1)), 

100.     ("6/2017 - 5/2025", dt.date(2017, 6, 1)), 

101.     ("6/2018 - 5/2025", dt.date(2018, 6, 1)), 

102.     ("6/2019 - 5/2025", dt.date(2019, 6, 1)), 

103.     ("6/2020 - 5/2025", dt.date(2020, 6, 1)), 

104.     ("6/2021 - 5/2025", dt.date(2021, 6, 1)), 

105.     ("6/2022 - 5/2025", dt.date(2022, 6, 1)), 

106.     ("6/2023 - 5/2025", dt.date(2023, 6, 1)), 

107.     ("6/2024 - 5/2025", dt.date(2024, 6, 1)), 

108. ] 

109. desired_end = dt.date(2025, 5, 31) 

110. period_end = min(data_end, desired_end) 

111. print("Using period end:", period_end) 

112. results = [] 

113. # Loop over all 13 periods 

114. for label, start_cal in period_specs: 

115.     mask_after_start = df["Date_only"] >= start_cal 

116.     if not mask_after_start.any(): 

117.         print(f"No data for {label}, skipping.") 

118.         continue 

119.  

120.     actual_start = df.loc[mask_after_start, "Date_only"].min() 

121.     mask = (df["Date_only"] >= actual_start) & (df["Date_only"] <= period_end) 

122.     sub = df.loc[mask].copy() 

123.     n = len(sub) 

124.     if n < 50: 

125.         print(f"Period {label} has only {n} rows, skipping regression.") 

126.         results.append({ 

127.             "Period": label, 

128.             "N": n, 

129.             "alpha_OLS": None, 

130.             "alpha_se_OLS": None, 

131.             "alpha_t_OLS": None, 

132.             "alpha_p_OLS": None, 

133.             "alpha_HAC": None, 

134.             "alpha_se_HAC": None, 

135.             "alpha_t_HAC": None, 

136.             "alpha_p_HAC": None, 

137.         }) 

138.         continue 

139.  

140.     y = sub["Strategy %"] 

141.     X = sm.add_constant(sub["BM %"]) 

142.     ols = sm.OLS(y, X).fit() 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

272 

 

143.     maxlags = int(n ** 0.5) 

144.     hac = ols.get_robustcov_results(cov_type="HAC", maxlags=maxlags) 

145.     results.append({ 

146.         "Period": label, 

147.         "N": n, 

148.         "alpha_OLS": float(ols.params[0]), 

149.         "alpha_se_OLS": float(ols.bse[0]), 

150.         "alpha_t_OLS": float(ols.tvalues[0]), 

151.         "alpha_p_OLS": float(ols.pvalues[0]), 

152.         "alpha_HAC": float(hac.params[0]), 

153.         "alpha_se_HAC": float(hac.bse[0]), 

154.         "alpha_t_HAC": float(hac.tvalues[0]), 

155.         "alpha_p_HAC": float(hac.pvalues[0]), 

156.     }) 

157. # Collect & save 

158. res_df = pd.DataFrame(results) 

159. res_df = res_df[ 

160.     [ 

161.         "Period", 

162.         "N", 

163.         "alpha_OLS", 

164.         "alpha_se_OLS", 

165.         "alpha_t_OLS", 

166.         "alpha_p_OLS", 

167.         "alpha_HAC", 

168.         "alpha_se_HAC", 

169.         "alpha_t_HAC", 

170.         "alpha_p_HAC", 

171.     ] 

172. ] 

173. res_rounded = res_df.round( 

174.     { 

175.         "alpha_OLS": 8, 

176.         "alpha_se_OLS": 8, 

177.         "alpha_t_OLS": 5, 

178.         "alpha_p_OLS": 8, 

179.         "alpha_HAC": 8, 

180.         "alpha_se_HAC": 8, 

181.         "alpha_t_HAC": 5, 

182.         "alpha_p_HAC": 8, 

183.     } 

184. ) 

185. print(res_rounded.to_string(index=False)) 

186. out_path = "/content/drive/MyDrive/hac_alpha_results_13_periods.csv" 

187. res_rounded.to_csv(out_path, index=False) 

188. print("Saved results to:", out_path) 

Source: Authors’ data (google colabs) 

Figure 11. Python script for HAC test 

 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

273 

 

Table 5. Newey–West HAC regression results by sample window 

Period N α (OLS) 
SE 

(OLS) 

t 

(OLS) 
p (OLS) 

α 

(HAC) 

SE 

(HAC) 

t 

(HAC) 
p (HAC) 

6/2012–5/2025 4717 0.00175 0.00030 5.851 1.00×10⁻⁸ 0.00175 4.1×10⁻⁴ 4.267 2.00×10⁻⁵ 

6/2013–5/2025 4362 0.00149 0.00029 5.155 2.60×10⁻⁷ 0.00149 3.5×10⁻⁴ 4.228 2.40×10⁻⁵ 

6/2014–5/2025 3997 0.00128 0.00027 4.750 2.11×10⁻⁶ 0.00128 3.1×10⁻⁴ 4.117 3.90×10⁻⁵ 

6/2015–5/2025 3632 0.00122 0.00027 4.492 7.28×10⁻⁶ 0.00122 3.3×10⁻⁴ 3.595 3.29×10⁻⁴ 

6/2016–5/2025 3266 0.00131 0.00030 4.404 1.10×10⁻⁵ 0.00131 3.6×10⁻⁴ 3.641 2.76×10⁻⁴ 

6/2017–5/2025 2901 0.00135 0.00033 4.112 4.03×10⁻⁵ 0.00135 3.7×10⁻⁴ 3.585 3.42×10⁻⁴ 

6/2018–5/2025 2536 0.00097 0.00031 3.144 1.69×10⁻³ 0.00097 3.4×10⁻⁴ 2.838 4.57×10⁻³ 

6/2019–5/2025 2171 0.00075 0.00031 2.426 1.53×10⁻² 0.00075 3.5×10⁻⁴ 2.132 3.31×10⁻² 

6/2020–5/2025 1805 0.00102 0.00035 2.922 3.52×10⁻³ 0.00102 3.9×10⁻⁴ 2.602 9.35×10⁻³ 

6/2021–5/2025 1440 0.00067 0.00038 1.772 7.66×10⁻² 0.00067 3.7×10⁻⁴ 1.768 7.73×10⁻² 

6/2022–5/2025 1075 0.00062 0.00030 2.062 3.95×10⁻² 0.00062 4.3×10⁻⁴ 1.402 1.61×10⁻¹ 

Source: Authors’ data 

Note: Rows for 6/2023–5/2025 and 6/2024–5/2025 are omitted from this HAC table because, in those 

short subsamples, the strategy and benchmark returns are nearly collinear and the regression becomes 

numerically ill-conditioned. The corresponding OLS coefficients and test statistics are nevertheless reported 

in Table 1 for completeness. 

 

Table 6. Walk-forward out-of-sample performance (3-year train / 1-year test, June–May years) 

Train start 
Train 

end 

Test 

year 
N_test 

Test cumulative 

return (strategy) 

Test 

cumulative 

return (BM) 

Test 

annualized 

return 

(strategy) 

Test 

annualized 

return (BM) 

2012 2014 2015 366 1.311 1.311 0.780 0.780 

2013 2015 2016 365 3.598 3.598 1.867 1.867 

2014 2016 2017 365 7.614 2.367 3.423 1.312 

2015 2017 2018 365 1.232 0.169 0.741 0.114 

2016 2018 2019 366 0.032 0.032 0.022 0.022 

2017 2019 2020 365 5.323 2.942 2.573 1.578 

2018 2020 2021 365 -0.036 -0.134 -0.025 -0.094 

2019 2021 2022 365 0.381 -0.147 0.250 -0.104 

2020 2022 2023 366 1.498 1.498 0.878 0.878 

2021 2023 2024 344 0.550 0.550 0.379 0.379 

Mean test annualized return (strategy) = 1.088 

Mean test annualized return (benchmark) = 0.673 

Share of positive test years (strategy) = 9/10 

Share of positive test years (benchmark) = 8/10 

Source: Authors’ data 

 
Table 7 below shows all the buy and sell signals, defined by the strategy. Note that on the sell signals, 

the Long/Sold is also 1, that is due to the signal confirming on the trading day end, executing the sell signal 

on the following day open. 
  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

274 

 

Table 7. Buy and sell signals 

Benchmark + strategy returns 

Benchmark (BTC) Strategy 

Date  close   open  BM % Long/Sold Strategy % Key event 

11/6/2012             5.5              5.5  0.73% 1 0.73% Bull mkt crossover 

5/4/2013         141.8          134.7  5.29% 1 5.29% Pi Cycle Top 

2/9/2013         130.2          130.7  -0.38% 1 -0.38% Bull mkt crossover 

5/12/2013     1,023.9      1,135.0  -9.79% 1 -9.79% Pi Cycle Top 

9/6/2014         648.8          658.0  -1.39% 1 -1.39% Bull mkt crossover 

17/8/2014         496.9          523.5  -5.08% 1 -5.08% Bull mkt crossunder 

9/2/2015         220.9          223.9  -1.35% 1 -1.35% RSI bottom 

16/12/2017   19,187.8    17,478.0  9.78% 1 9.78% Pi Cycle Top 

31/12/2018     3,693.3      3,831.0  -3.60% 1 -3.60% RSI Bottom 

12/4/2021   59,831.7    59,972.3  -0.23% 1 -0.23% Pi Cycle Top 

16/8/2021   45,930.5    47,025.0  -2.33% 1 -2.33% Bull mkt crossover 

5/12/2021   49,463.2    49,240.8  0.45% 1 0.45% Bull mkt crossunder 

4/4/2022   46,598.2    46,414.9  0.39% 1 0.39% Bull mkt crossover 

10/4/2022   42,133.9    42,774.9  -1.50% 1 -1.50% Bull mkt crossunder 

29/8/2022   20,302.0    19,571.0  3.74% 1 3.74% RSI Bottom 

10/5/2025 104,809.0 102,992.0 1.76% 1 1.76% dataset end 

Source: Authors’ data 

 

Table 8. Regression results for the period June 2012 – May 2025 

Regression Statistics               

Multiple R 0.7534               

R Square 0.5676               

Adjusted R 

Square 0.5675               

Standard Error 0.0205               

Observations 4717               

                  

ANOVA 6/2012 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 2.590723 2.590723 6188.022 0       

Residual 4715 1.974017 0.000419           

Total 4716 4.564739             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00175 0.00030 5.85111 5.2133E-09 0.00116 0.00233 0.00116 0.00233 

X Variable 1 0.56662 0.00720 78.66 0 0.55250 0.58074 0.55250 0.58074 

Source: Authors’ data 
  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

275 

 

Table 9. Regression results for the period June 2013 – May 2025 

Regression Statistics               

Multiple R 0.7822               

R Square 0.6118               

Adjusted R 

Square 0.6117               

Standard Error 0.0190               

Observations 4362               

                  

ANOVA 1/6/2013 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 2.479121 2.479121 6870.217 0       

Residual 4360 1.573308 0.000361           

Total 4361 4.052428             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00149 0.00029 5.15492 2.64976E-07 0.00092 0.00205 0.00092 0.00205 

X Variable 1 0.61064 0.00737 82.89 0 0.59620 0.62509 0.59620 0.62509 

Source: Authors’ data 

 

Table 10. Regression results for the period June 2014 – May 2025 
Regression Statistics               

Multiple R 0.8199               

R Square 0.6723               

Adjusted R 

Square 0.6722               

Standard Error 0.0170               

Observations 3997               

                  

ANOVA 1/6/2014 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 2.360475 2.360475 8195.721 0       

Residual 3995 1.150612 0.000288           

Total 3996 3.511087             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00128 0.00027 4.74974 2.10827E-06 0.00075 0.00180 0.00075 0.00180 

X Variable 1 0.67103 0.00741 90.53 0 0.65650 0.68556 0.65650 0.68556 

Source: Authors’ data 

  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

276 

 

Table 11. Regression results for the period June 2015 – May 2025 
Regression Statistics               

Multiple R 0.8447               

R Square 0.7135               

Adjusted R 

Square 0.7134               

Standard Error 0.0163               

Observations 3632               

                  

ANOVA 1/6/2015 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 2.408330 2.408330 9038.519 0       

Residual 3630 0.967220 0.000266           

Total 3631 3.375550             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00122 0.00027 4.49184 7.28097E-06 0.00069 0.00175 0.00069 0.00175 

X Variable 1 0.71226 0.00749 95.07 0 0.69757 0.72695 0.69757 0.72695 

Source: Authors’ data 

 
Table 12. Regression results for the period June 2016 – May 2025 

Regression Statistics               

Multiple R 0.8339               

R Square 0.6953               

Adjusted R 

Square 0.6952               

Standard Error 0.0170               

Observations 3266               

                  

ANOVA 1/6/2016 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 2.150193 2.150193 7448.660 0       

Residual 3264 0.942214 0.000289           

Total 3265 3.092406             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00131 0.00030 4.40390 1.09735E-05 0.00073 0.00190 0.00073 0.00190 

X Variable 1 0.69403 0.00804 86.31 0 0.67826 0.70979 0.67826 0.70979 

Source: Authors’ data 
  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

277 

 

Table 13. Regression results for the period June 2017 – May 2025 
Regression Statistics               

Multiple R 0.8175               

R Square 0.6683               

Adjusted R 

Square 0.6682               

Standard Error 0.0177               

Observations 2901               

                  

ANOVA 1/6/2017 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 1.824587 1.824587 5842.049 0       

Residual 2899 0.905415 0.000312           

Total 2900 2.730002             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00135 0.00033 4.11187 4.0337E-05 0.00071 0.00200 0.00071 0.00200 

X Variable 1 0.66705 0.00873 76.43 0 0.64994 0.68416 0.64994 0.68416 

Source: Authors’ data 

 
Table 14. Regression results for the period June 2018 – May 2025 

Regression Statistics               

Multiple R 0.8442               

R Square 0.7127               

Adjusted R 

Square 0.7126               

Standard Error 0.0155               

Observations 2536               

                  

ANOVA 1/6/2018 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 1.518074 1.518074 6287.112 0       

Residual 2534 0.611855 0.000241           

Total 2535 2.129929             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00097 0.00031 3.14394 0.001686206 0.00037 0.00158 0.00037 0.00158 

X Variable 1 0.71183 0.00898 79.29 0 0.69422 0.72943 0.69422 0.72943 

Source: Authors’ data 
  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

278 

 

Table 15. Regression results for the period June 2019 – May 2025 

Regression Statistics               

Multiple R 0.8763               

R Square 0.7678               

Adjusted R 

Square 0.7677               

Standard Error 0.0145               

Observations 2171               

                  

ANOVA 1/6/2019 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 1.498918 1.498918 7173.241 0       

Residual 2169 0.453234 0.000209           

Total 2170 1.952152             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00075 0.00031 2.42628 0.015335382 0.00014 0.00136 0.00014 0.00136 

X Variable 1 0.76721 0.00906 84.69 0 0.74944 0.78497 0.74944 0.78497 

Source: Authors’ data 

 

Table 16. Regression results for the period June 2020 – May 2025 
Regression Statistics               

Multiple R 0.8218               

R Square 0.6754               

Adjusted R 

Square 0.6752               

Standard Error 0.0149               

Observations 1805               

                  

ANOVA 1/6/2020 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 0.828433 0.828433 3751.438 0       

Residual 1803 0.398158 0.000221           

Total 1804 1.226591             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00102 0.00035 2.92183 0.003523064 0.00034 0.00171 0.00034 0.00171 

X Variable 1 0.67444 0.01101 61.25 0 0.65285 0.69604 0.65285 0.69604 

Source: Authors’ data 

  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

279 

 

Table 17. Regression results for the period June 2021 – May 2025 
Regression Statistics               

Multiple R 0.8045               

R Square 0.6473               

Adjusted R 

Square 0.6470               

Standard Error 0.0142               

Observations 1440               

                  

ANOVA 1/6/2021 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 0.535521 0.535521 2638.808 0       

Residual 1438 0.291829 0.000203           

Total 1439 0.827350             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00067 0.00038 1.77202 0.076602273 -0.00007 0.00140 -0.00007 0.00140 

X Variable 1 0.64687 0.01259 51.37 0 0.62217 0.67157 0.62217 0.67157 

Source: Authors’ data 

 

Table 18. Regression results for the period June 2022 – May 2025 
Regression Statistics               

Multiple R 0.9216               

R Square 0.8494               

Adjusted R 

Square 0.8492               

Standard Error 0.0098               

Observations 1075               

                  

ANOVA 1/6/2022 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 0.577390 0.577390 6049.540 0       

Residual 1073 0.102411 0.000095           

Total 1074 0.679801             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0.00062 0.00030 2.06180 0.039467333 0.00003 0.00120 0.00003 0.00120 

X Variable 1 0.84837 0.01091 77.78 0 0.82697 0.86977 0.82697 0.86977 

Source: Authors’ data 

  



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

280 

 

Table 19. Regression results for the period June 2023 – May 2025 
Regression 

Statistics                 

Multiple R 1               

R Square 1               

Adjusted R 

Square 1               

Standard Error        0               

Observations 710               

                  

ANOVA 1/6/2023 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 0.474147869 0.474147869 - -       

Residual 708 2.36968E-32 3.34701E-35           

Total 709 0.474147869             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% 

Upper 

95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0 - - - - - - - 

X Variable 1 1 - - - - - - - 

Source: Authors’ data 

 
Table 20. Regression results for the period June 2024 – May 2025 

Regression Statistics               

Multiple R 1               

R Square 1               

Adjusted R 

Square 1               

Standard Error 0               

Observations 344               

                  

ANOVA 1/6/2024 - 10/5/2025       

  df SS MS F 

Significance 

F       

Regression 1 0.240305728 0.240305728 - -       

Residual 342 0 0           

Total 343 0.240305728             

                  

  Coefficients 

Standard 

Error t Stat P-value Lower 95% Upper 95% 

Lower 

95.0% 

Upper 

95.0% 

Intercept 0 - - - - - - - 

X Variable 1 1 - - - - - - - 

Source: Authors’ data 

Note: In the periods June 2023–May 2025 and June 2024–May 2025 (Tables 19 and 20), the strategy and 

benchmark returns are almost perfectly collinear, so the regression is degenerate. α = 0 and β = 1 are implied 

mechanically; residual variance is effectively zero, and standard errors, test statistics and confidence intervals 

are therefore not reported. 

 

 

 



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

281 

 

 
Source: Authors’ data 

Figure 12. Equity curves 
 

 

 

  

 $ 1 000

 $ 200 001 000

 $ 400 001 000

 $ 600 001 000

 $ 800 001 000

 $ 1 000 001 000

 $ 1 200 001 000

1
.6

.2
0

1
2

1
.3

.2
0

1
3

1
.1

2
.2

0
1

3

1
.9

.2
0

1
4

1
.6

.2
0

1
5

1
.3

.2
0

1
6

1
.1

2
.2

0
1

6

1
.9

.2
0

1
7

1
.6

.2
0

1
8

1
.3

.2
0

1
9

1
.1

2
.2

0
1

9

1
.9

.2
0

2
0

1
.6

.2
0

2
1

1
.3

.2
0

2
2

1
.1

2
.2

0
2

2

1
.9

.2
0

2
3

1
.6

.2
0

2
4

1
.3

.2
0

2
5

Equity curve 6/2012 -5/2025

Benchmark

 Strategy

 $ 1 000

 $ 10 001 000

 $ 20 001 000

 $ 30 001 000

 $ 40 001 000

 $ 50 001 000

1
.9

.2
0

1
3

1
.5

.2
0

1
4

1
.1

.2
0

1
5

1
.9

.2
0

1
5

1
.5

.2
0

1
6

1
.1

.2
0

1
7

1
.9

.2
0

1
7

1
.5

.2
0

1
8

1
.1

.2
0

1
9

1
.9

.2
0

1
9

1
.5

.2
0

2
0

1
.1

.2
0

2
1

1
.9

.2
0

2
1

1
.5

.2
0

2
2

1
.1

.2
0

2
3

1
.9

.2
0

2
3

1
.5

.2
0

2
4

1
.1

.2
0

2
5

Equity curve 9/2013 - 5/2025

 Benchmark

 Strategy

 $ 1 000

 $ 1 001 000

 $ 2 001 000

 $ 3 001 000

 $ 4 001 000

 $ 5 001 000

 $ 6 001 000

1
.6

.2
0

1
4

1
.2

.2
0

1
5

1
.1

0
.2

0
1

5

1
.6

.2
0

1
6

1
.2

.2
0

1
7

1
.1

0
.2

0
1

7

1
.6

.2
0

1
8

1
.2

.2
0

1
9

1
.1

0
.2

0
1

9

1
.6

.2
0

2
0

1
.2

.2
0

2
1

1
.1

0
.2

0
2

1

1
.6

.2
0

2
2

1
.2

.2
0

2
3

1
.1

0
.2

0
2

3

1
.6

.2
0

2
4

1
.2

.2
0

2
5

Equity curve 6/2014 -5/2025

 Benchmark

 Strategy



Alejandro Rabinovich / Finance, Accounting and Business Analysis, Volume 7, Issue 2, 2025 

282 

 

 
Source: Authors’ data 

Figure 13. Equity curves pt 2 

 
 

 $ 1 000

 $ 2 001 000

 $ 4 001 000

 $ 6 001 000

 $ 8 001 000

1
.2

.2
0

1
5

1
.9

.2
0

1
5

1
.4

.2
0

1
6

1
.1

1
.2

0
1

6

1
.6

.2
0

1
7

1
.1

.2
0

1
8

1
.8

.2
0

1
8

1
.3

.2
0

1
9

1
.1

0
.2

0
1

9

1
.5

.2
0

2
0

1
.1

2
.2

0
2

0

1
.7

.2
0

2
1

1
.2

.2
0

2
2

1
.9

.2
0

2
2

1
.4

.2
0

2
3

1
.1

1
.2

0
2

3

1
.6

.2
0

2
4

1
.1

.2
0

2
5

Equity curve 2/2015 -5/2025

 Benchmark

 Strategy

 $ 1 000

 $ 21 000

 $ 41 000

 $ 61 000

 $ 81 000

 $ 101 000

1
.1

2
.2

0
1

8

1
.5

.2
0

1
9

1
.1

0
.2

0
1

9

1
.3

.2
0

2
0

1
.8

.2
0

2
0

1
.1

.2
0

2
1

1
.6

.2
0

2
1

1
.1

1
.2

0
2

1

1
.4

.2
0

2
2

1
.9

.2
0

2
2

1
.2

.2
0

2
3

1
.7

.2
0

2
3

1
.1

2
.2

0
2

3

1
.5

.2
0

2
4

1
.1

0
.2

0
2

4

1
.3

.2
0

2
5

Equity curve 12/2018 - 5/2025

 Benchmark

 Strategy

 $ -
 $ 1 000
 $ 2 000
 $ 3 000
 $ 4 000
 $ 5 000
 $ 6 000

1
.8

.2
0

2
1

1
.1

1
.2

0
2

1

1
.2

.2
0

2
2

1
.5

.2
0

2
2

1
.8

.2
0

2
2

1
.1

1
.2

0
2

2

1
.2

.2
0

2
3

1
.5

.2
0

2
3

1
.8

.2
0

2
3

1
.1

1
.2

0
2

3

1
.2

.2
0

2
4

1
.5

.2
0

2
4

1
.8

.2
0

2
4

1
.1

1
.2

0
2

4

1
.2

.2
0

2
5

1
.5

.2
0

2
5

Equity curve 8/2021 - 5/2025

 Benchmark

 Strategy

 $ -

 $ 1 000

 $ 2 000

 $ 3 000

 $ 4 000

 $ 5 000

 $ 6 000

4.4.2022 4.4.2023 4.4.2024 4.4.2025

Equity curve 4/2022 - 5/2025

 Benchmark

 Strategy


