Original Article Title: Monthly Outlook: Debunking the Seasonality Myth
Original Article Authors: David Duong, Global Research Director at Coinbase; Colin Basco, Research Assistant at Coinbase

Translation: xiaozou, Golden Finance

Key Points:

We predict a strong start to the cryptocurrency market in the fourth quarter of 2025, driven by ample liquidity, favorable macro backdrop, and supportive regulatory developments, with Bitcoin poised to outperform.

The ongoing technical demand from Digital Asset Treasuries (DATs) is expected to continue to underpin the cryptocurrency market, even as the industry enters a competitive "player versus player" phase.

Our research indicates that historical monthly seasonality patterns, particularly the "September Effect," have not been a significant or reliable predictor of cryptocurrency market performance.

1. Overview

We believe the cryptocurrency bull market still has room to run in the early fourth quarter of 2025, propelled by a resilient liquidity environment, favorable macro backdrop, and supportive regulatory dynamics. We see Bitcoin, in particular, as likely to continue to outperform market expectations, benefiting directly from existing macro tailwinds. In other words, unless there is a sharp move in energy prices (or other factors that could negatively impact the inflationary trend), the immediate risk to the U.S. monetary policy path is actually quite low. Meanwhile, the technical demand from Digital Asset Treasuries (DATs) should continue to provide strong support for the cryptocurrency market.

However, seasonal concerns continue to haunt the cryptocurrency space – historically, Bitcoin saw September sell-offs against the U.S. Dollar for six consecutive years from 2017 to 2022. Although this trend led many investors to believe that seasonal factors significantly influence cryptocurrency market performance, this assumption was debunked in 2023 and 2024. In fact, our research shows that the small sample size and potentially wide outcome distribution limit the statistical significance of such seasonal indicators.

A more critical question for the cryptocurrency market is: Are we in the early or late stage of the DAT cycle? As of September 10, public DATs hold over 1 million BTC (110 billion USD), 4.9 million ETH (21.3 billion USD), and 8.9 million SOL (1.8 billion USD), with latecomers starting to target lower-tier altcoins on the risk curve. We believe we are currently in the "player versus player" (PvP) stage of the cycle, which will continue to drive funds towards large-cap cryptocurrency assets. However, this also likely signals the imminent entry of smaller DAT participants into the consolidation phase.

2. Continued Optimism

Earlier this year, we suggested that the crypto market would bottom out in the first half of 2025 and reach a new all-time high in the second half of 2025. This was a contrarian view at the time—market participants were worried about a potential recession, questioned whether the price increases meant an irrational market surge, and were concerned about the sustainability of any recovery. However, we found these views to be misleading, so we returned to our unique macro outlook perspective.

As we enter the fourth quarter, we maintain an optimistic outlook for the crypto market, expecting continued strong liquidity, a favorable macroeconomic environment, and supportive regulatory progress. In terms of monetary policy, we expect the Federal Reserve to cut interest rates on September 17 and October 29, as the US labor market has provided strong evidence of softness. We believe this will not only not form a local peak but rather activate sidelined capital. In fact, in August, we pointed out that the rate cuts could prompt a significant portion of the $7.4 trillion in money market fund assets to end their wait-and-see stance.

However, a significant shift in the current inflation trajectory (such as an increase in energy prices) poses a risk to this outlook. (Note: We believe that the actual risks posed by tariffs are much lower than some views have assessed.) However, OPEC+ has recently agreed to increase oil production again, and global oil demand is showing signs of slowing down. Nevertheless, the prospect of further sanctions on Russia could also push up oil prices. Currently, we do not expect oil prices to break through the threshold that would push the economic situation into a stagflationary range.

3. Maturity of the DAT Cycle

On the other hand, we believe that the technical demands of Digital Asset Treasuries (DATs) are expected to continue to support the crypto market. In fact, the DAT phenomenon has reached a critical turning point. We are neither in the early adoption stage of the past 6-9 months nor do we believe we are nearing the end of the cycle. In fact, we have entered what is known as the "Player versus Player" (PvP) phase—a competitive stage where success increasingly depends on execution, differentiated strategies, and timing, rather than simply replicating MicroStrategy's operational model.

Indeed, early movers like MicroStrategy once enjoyed a significant Net Asset Value (NAV) premium, but competitive pressures, execution risks, and regulatory constraints have led to a compression of mNAV (market cap-to-NAV ratio). We believe the scarcity premium that early adopters benefited from has now dissipated. Nevertheless, DATs focused on Bitcoin currently hold over 1 million BTC, representing approximately 5% of the circulating supply of the token. Similarly, top ETH-focused DATs collectively hold around 4.9 million ETH (worth $21.3 billion), accounting for over 4% of the total ETH circulating supply.

Figure 1. ETH Specialized Digital Asset Treasury Continues Accelerated Acquisition Trend

In August, the Financial Times reported that 154 U.S. listed companies had raised approximately $98.4 billion by 2025 for the purchase of crypto assets, a significant increase from the $33.6 billion raised by the top 10 companies earlier this year (based on Architect Partners data). Capital investments in other tokens are also growing, especially in SOL and other alternative tokens. (Forward Industries recently raised $1.65 billion to establish a SOL-based digital asset treasury, with support from Galaxy Digital, Jump Crypto, and Multicoin Capital.)

This growth has triggered tighter scrutiny. In fact, recent reports indicate that Nasdaq is strengthening its regulation of DATs, requiring specific transactions to be approved by shareholders and advocating for enhanced disclosure. However, Nasdaq clarified that it has not issued any formal press releases regarding new rules for DATs.

Currently, we believe the DAT cycle is maturing, but it is neither early-stage nor late-stage. What can be certain is that, in our view, the era of easy profits and guaranteed mNAV premiums has ended—-in this PvP stage, only the most disciplined and strategically positioned participants can thrive. We expect the crypto market to continue benefiting from unprecedented capital inflows into these vehicles, enhancing return performance.

3. Is Seasonal Risk Actually Real?

Meanwhile, seasonal volatility has been a concern for crypto market participants. Bitcoin has experienced a September drop against the dollar from 2017 to 2022 for six consecutive years, with an average negative return of 3% over the past decade. This has left many investors with the impression that seasonal factors significantly impact crypto market performance, with September typically being viewed as an unfavorable time to hold risk assets. However, if trading is based on this assumption, it will be falsified in both 2023 and 2024.

Indeed, we believe that monthly seasonal fluctuations are not an effective trading signal for Bitcoin. Through various methods such as frequency distribution charts, odds ratios, out-of-sample scores, placebo tests, and control variables, the conclusion is consistent: annual months are not statistically reliable predictors of BTC's monthly log return positive or negative values. (Note: We use log returns to measure geometric or compound growth as they better reflect long-term trends and account for Bitcoin's higher volatility.)

Figure 2. Bitcoin Monthly Log Returns Heatmap

The following test found that the "calendar month" is unreliable for predicting Bitcoin's monthly log return positive or negative values:

（1）Wilson Confidence Interval

Figure 3 shows that, after considering small sample uncertainty, no month can surpass a clear threshold for predicting seasonality. Months that appear "high" (February/October) or "low" (August/September) have error ranges overlapping with the overall average and other months, showing random variance rather than a persistent calendar effect.

Each data point shows the probability of BTC ending the month with a positive return; the vertical line/bar represents the 95% Wilson confidence interval band—appropriate for measuring uncertainty in small samples as each month has only about 12-13 data points.

The dashed line indicates the average probability of an overall positive return. Since we are simultaneously examining data for 12 months, we use the Holm multiple testing correction method to avoid any single lucky month masquerading as a systematic pattern.

Figure 3. BTC Positive Log Return with 95% Wilson Confidence Interval

（2）Logistic Regression Analysis

We employed a logistic regression model to test the impact of specific months on Bitcoin's probability of price increase (using January as the baseline). Figure 4 shows that the odds ratios for each month are mostly concentrated around 1.0, with the key being that their 95% confidence intervals all span the 1.0 line.

A value close to 1.0 indicates "equal probability of achieving a positive log return as January," above 1.0 indicates "higher probability," and below 1.0 indicates "lower probability."

For example, an odds ratio of 1.5 means "about 50% higher probability of a positive return month compared to January," while 0.7 indicates "about 30% lower probability."

Since most confidence intervals span 1.0 and no month exhibits significance after Holm multiple testing correction, we cannot consider the calendar month as an effective indicator for predicting Bitcoin's log return positivity or negativity.

Figure 4. Logistic Regression—Monthly BTC Log Return Probability Ratios relative to January (baseline)

（3）Out-of-Sample Prediction

At each step, we only re-estimated two models using data available up to that month (initially trained on half the dataset):

The baseline model is a logistic model with only an intercept term, which predicts a constant probability (equal to the baseline rate, which is the proportion of positive return months in history).

The Month of Year (MoY) model is a logistic regression model that includes month dummy variables; it predicts the probability of a positive return month based on past performance during that month.

Our results are presented in Figure 5, where the X-axis represents the predicted probability of a positive log return month, and the Y-axis represents the actual proportion of months with positive returns. When plotting the predicted results, data points from a perfectly calibrated model should lie along the 45-degree line—i.e., predicting a 50% increase probability should result in an actual 50% proportion of positive return months.

The Month of Year (MoY) model exhibits significant bias. For example:

When predicting an increase probability of around 27%, the actual realized frequency is around 50% (overly pessimistic);

It only roughly approximates the target within the 45-60% prediction range;

It is overconfident in high probability ranges—e.g., around 75% prediction corresponds to around 70% realization, while extreme predictions of around 85% result in around 0% realization.

In contrast, the baseline model, which consistently predicts the historical baseline rate (around 55-57% increase probability), closely follows the 45-degree line. Given the relatively stable probability of positive return months in Bitcoin's history, this line remains almost stationary. In summary, these results suggest that calendar months have little predictive power in out-of-sample forecasting.

Figure 5: Out-of-Sample Forecast Accuracy of the Month of Year (MoY) Logistic Regression Model

(4) Placebo Randomization Test

To verify whether the "month labels" help predict positive or negative log returns, we used a simple logistic model with month dummy variables and conducted an omnibus test to determine if these variables improve the model fit compared to a baseline model without month labels (standard likelihood ratio joint test). The observed p-value was 0.15, indicating that even if the month factor is irrelevant, the likelihood of observing a pattern as significant as this by chance alone is approximately 15%. We then randomly shuffled the month labels thousands of times and repeated the same joint test each time.

The results show that approximately 19% of random shuffles produced results with p-values less than or equal to the observed p-value (Figure 6).

In summary, this result is highly common under purely random conditions, reinforcing the conclusion that "there is no month signal." If month labels were statistically significant, a true data joint test should yield a p-value less than 0.05, and the proportion of shuffled outcomes producing such a small p-value should be less than 5%.

Figure 6. Placebo p-Value Distribution Generated by Randomly Shuffling the "Month" Label in the Logic Model

(5) Control Variable Test

Adding a real-world calendar flag did not unlock tradable advantage—and typically degraded the accuracy of directional predictability. We re-estimated the "positive return month probability" using the same month dummy variable, then added two major event practice control variables: 1) potentially influencing Bitcoin log return; 2) annual month occurrence variability—Lunar New Year and Bitcoin halving window (±2 months). We only used control variables corresponding to different calendar months annually to avoid redundant dummy variables leading to unstable model estimates.

This test aims to verify two common concerns: (i) phenomena that seem like a "month effect" may be mere disguises of cyclical events—such as Lunar New Year (LNY) liquidity or the Bitcoin halving effect; (ii) even if the original month pattern is weak, considering these driving factors may still yield utility. In the initial stage, we used half of the data set for training and the other half for testing. Brier scores were used to evaluate monthly probability predictions, with the score reflecting the average square error between predicted probabilities and actual up/down outcomes (i.e., the deviation of predicted values from reality).

In Figure 7, the bar chart shows the Brier improvement value of each model compared to the simple baseline (using only the training window's historical uptrend rate as a single value). All bars are below zero, indicating that the performance of each control variable variant is worse than the constant probability baseline. In short, introducing additional calendar flags based on the month label only added noise.

Figure 7: Brier Improvement Scores of Logistic Regression Models with Control Variables in Out-of-Sample Predictions

4. Conclusion

The market's seasonal concept imposes a harmful restraint on investors' psychology and may form a self-fulfilling prophecy. However, our model shows that the performance of simply assuming the monthly up/down probability is roughly in line with the long-term historical average is superior to all calendar-based trading strategies. This strongly suggests that calendar patterns do not contain effective information for predicting Bitcoin's monthly direction. Since calendar months cannot reliably predict the positive or negative direction of log returns, the likelihood of predicting return magnitudes is even more minuscule. The simultaneous September decline from previous years or the surge in Bitcoin's "October effect" legend may have statistical interest, but they lack statistical significance.

Original Article Link

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia