RevertX Strategy
Risk Category: Medium
Introduction
Mean reversion is a well-established trading strategy that seeks to capitalize on temporary price deviations from an expected value. While the concept is widely known, Teahouse’s new strategy — RevertX — applies a specialized pairs trading approach to $BTC and $ETH, leveraging statistical models to identify profitable trading opportunities.
TL;DR:
RevertX capitalizes on mean reversion between $ETH and $BTC by tracking price deviations and trading when they exceed a threshold.
The strategy succeeds by allowing small losses and capturing larger profitable opportunities.
53% annualized return with manageable drawdowns.
33.6% win rate, but profitable due to a favorable risk-reward ratio (wins 4.5x larger than losses).
Small losses and frequent trade resolutions (avg. 5 days).
Understanding the Mechanics Behind RevertX
The core process of this strategy can be visualized in three key steps:
This strategy is particularly well-suited for Tea-REX, as it allows the vault to strategically capture mean-reversion opportunities with leveraged exposure. Before diving into the mathematical framework, it’s important to understand the risks associated with mean-reversion pairs trading:
Divergence Risk: The assumption that ‘price relationships will revert’ may fail due to fundamental market shifts.
Execution Risk: Slippage, liquidity constraints, and trading costs can impact profitability.
Timing Risk: The reversion process may take longer than expected, increasing carrying costs and the opportunity cost of capital being tied up in positions.
With these considerations in mind, let’s explore the mechanics behind RevertX, starting with how we construct the trading signal.
1. Constructing the Trading Signal — The Key to the RevertX Strategy
A key aspect of this strategy is to identify moments when $ETH and $BTC temporarily deviate from their typical price relationship. When this happens, the strategy aims to take advantage of the expectation that their prices will eventually revert to their usual pattern. To build our trading signal, we analyzed the price relationship between $ETH and $BTC using hourly data from Binance, covering the period from January 2017 to November 2024.
In this strategy:
We analyze the historical price relationship between $ETH and $BTC and confirm that they are cointegrated.
We track how much $ETH’s price deviates from its expected value relative to $BTC to identify trading opportunities and capitalize on these deviations.
This expected relationship is calculated using a simple regression model that updates daily, based on data from the previous 20 days.
The formula we use is:
Signal = ETH Price — (β₁ × BTC Price + β₀)
Where:
$ETH price is how much Ethereum costs at a given moment.
$BTC price is how much Bitcoin costs at the same moment.
β1 is a multiplier that allows us to compare $ETH and $BTC, given that they trade at different price levels.
β0 is an adjustment factor that accounts for the long-term relationship between the two cryptocurrencies.
We fit β₀ and β₁ to create a linear combination that ensures the signal remains stationary.
What the data shows:
On average, β1 has been about 0.077, meaning a $1 change in $BTC typically corresponds to a $0.077 change in $ETH.
This multiplier has remained relatively stable, with a standard deviation of 0.04.
β0 has averaged around -271 but varies considerably more, with a standard deviation of 1251.
2. Ensuring Signal Stability
To ensure our trading signal is reliable, we need to verify that it behaves in a stable and predictable manner over time. In statistical terms, the signal must be stationary — fluctuating around a consistent average rather than drifting unpredictably. This is crucial because our strategy depends on the assumption that when prices move too far in one direction, they will eventually revert to their typical relationship (baseline value).
To test for stationarity, we applied the Augmented Dickey–Fuller (ADF) test to both:
In-sample residuals (historical data used to build the model)
Out-sample trading signal (new data not used in model training)
Both tests provided strong evidence of stationarity (p-value < 0.000001), reinforcing our confidence that it is likely to revert when the signal deviates significantly from zero.
3. Placing Trade Positions:
For each trade, we allocate a fixed amount of capital. For our study, this was set at $1,000 per trade. Maintaining consistent position sizes helps manage risk effectively.
Position Sizing Formulas
The strategy determines the trade size for each cryptocurrency using the following formulas:
For Ethereum:
For Bitcoin:
Where:
K is the fixed capital allocation per trade
β1 is the regression coefficient linking $ETH and $BTC prices
The Threshold Value (δ = 18.06)
The strategy only enters a trade when the signal crosses a threshold value (δ) of 18.06, which is 1.5 times the standard deviation of the signal. This value was carefully chosen because:
It is set high enough to filter out normal market noise, avoiding unnecessary trades.
It is sensitive enough to catch meaningful deviations in the $ETH-$BTC relationship.
It assumes the price deviations follow a random walk (Brownian motion) pattern.
By waiting for the signal to cross the threshold before entering a trade, the strategy focuses on genuine opportunities rather than reacting to minor market fluctuations.
Visualizing the Mean-Reversion Signal
The chart below illustrates how the trading signal fluctuates around the fair value line. Trades are triggered when the signal crosses predefined thresholds, identifying opportunities for mean reversion while avoiding minor price fluctuations.
Backtesting and Findings
To evaluate the strategy’s effectiveness, we conducted a backtest using actual historical data and a fixed trade size of $1,000 per trade.
Trade Statistics
The strategy executed 372 complete trading cycles (excursions) over the test period from late 2017 through early 2024, with the following characteristics:
Average Position Duration: 118 hours
Win Rate: 125 trades (33.6%)
Loss Rate: 247 trades (66.4%)
Strategy Performance
Total PnL: +$3,766 over the testing period
APR: 53%
Maximum Drawdown: $112 (just 11.2% of the trade size), indicating that losses were limited and manageable (low downside risk).
Performance Visualization
Green line: The model’s signal or prediction for trading. When it deviates from zero, it indicates when to open a long or short position.
Orange line: $BTC Price
Gray line: $ETH Price
Red line: Cumulated PnL
The Risk-Reward Profile
Despite a low win rate (33.6%), the strategy remains profitable due to its favorable risk-reward dynamic:
Winning trades yielded an average profit of $55 (std = 94)
Losing trades had a much smaller average loss of $12.25 (std = 14.5)
The average winning trade was ~4.5x larger than the average losing trade
This “small losses, big wins” approach ensures that even though the strategy loses more often than it wins, the overall gains far outweigh the losses.
Conclusion
The RevertX strategy demonstrates a strong approach to mean-reversion pairs trading by focusing on the relationship between $ETH and $BTC. Its backtest results show a 53% annualized return with a maximum drawdown of $112, reflecting effective risk management. Despite a low win rate of 33.6%, the strategy remains profitable due to its favorable risk-reward ratio, with winning trades averaging 4.5 times larger than losing trades.
Key takeaways include:
The Strategy Works: The performance results suggest the approach is fundamentally sound.
Risk Management is Effective: Even with a two-thirds loss rate, the strategy limits losses while allowing profits to run.
Psychological Challenge: While profitable overall, this strategy requires being comfortable with frequent small losses, understanding that larger wins will eventually compensate for them.
Capital Efficiency: With trades typically resolving within 5 days, capital can be redeployed frequently, potentially increasing the overall return on investment.
RevertX shows why mean-reversion strategies are favored among quantitative traders: steady profitability with a low win rate, as long as the wins significantly outweigh the losses.
Last updated
Was this helpful?