Advanced Kelly Criterion Applications: Fractional Kelly, and Adjustments for Estimation Errors
The Black Book of Day Trading Strategies
1,000 complete strategies · 31 chapters · Full trade plans
Advanced Kelly Criterion Applications: Fractional Kelly, and Adjustments for Estimation Errors
The theoretical elegance of the Kelly Criterion, with its promise of optimal capital growth, often collides with the messy reality of real-world trading. The practical implementation of the Kelly formula is fraught with challenges, most notably the aggressive nature of the full Kelly bet and the inherent uncertainty in estimating the formula's inputs. This article will explore two advanced applications of the Kelly Criterion that address these challenges: the use of fractional Kelly and the adjustment for estimation errors.
Fractional Kelly: Taming the Volatility of the Full Kelly
The full Kelly Criterion, while mathematically optimal, is notoriously aggressive. It can lead to wild swings in the equity curve and a level of drawdown that is unacceptable to most traders. The solution to this problem is the use of a "fractional Kelly," which involves risking a fraction of the bet size recommended by the full Kelly formula. For example, a trader might use a "half Kelly" (risking 50% of the full Kelly bet) or a "quarter Kelly" (risking 25% of the full Kelly bet). The use of a fractional Kelly reduces the volatility of the equity curve and the risk of ruin, while still capturing a significant portion of the optimal growth rate.
Estimation Errors: The Achilles' Heel of the Kelly Criterion
The Kelly Criterion is highly sensitive to the accuracy of its inputs: the win rate (W) and the win/loss ratio (R). These parameters are typically estimated from historical data, and as such, they are subject to estimation error. The problem is that a small error in the estimation of W and R can lead to a large error in the calculated Kelly percentage. This can result in overbetting, which can have catastrophic consequences for the trading account.
Adjustments for Estimation Errors: A Bayesian Approach
One way to address the problem of estimation error is to use a Bayesian approach to adjust the Kelly fraction. A Bayesian approach allows the trader to incorporate their uncertainty about the true values of W and R into the Kelly calculation. This can be done by using a prior distribution to represent the trader's beliefs about the parameters before observing the data, and then updating these beliefs using the observed data to obtain a posterior distribution. The Kelly fraction can then be calculated based on the posterior distribution, which will be more conservative than the fraction calculated from the point estimates of W and R.
The Formula for Adjusting the Kelly Fraction
A simplified approach to adjusting the Kelly fraction for estimation error is to use the following formula, which was proposed by Ed Thorp:
Adjusted Kelly % = Kelly % * (N / (N + k))*
Where:
Kelly %is the Kelly percentage calculated from the point estimates of W and R.Nis the number of trades in the historical data.kis a constant that represents the trader's uncertainty about the parameters. A higher value of k results in a more conservative adjustment.
The Impact of Estimation Errors on the Optimal Kelly Fraction
The following table shows the impact of estimation error on the optimal Kelly fraction for a trading strategy with a true win rate of 60% and a true win/loss ratio of 2:1:
| Number of Trades | Estimated Win Rate | Estimated Win/Loss Ratio | Full Kelly % | Adjusted Kelly % (k=10) |
|---|---|---|---|---|
| 50 | 62% | 2.1:1 | 44% | 37% |
| 100 | 61% | 2.05:1 | 42% | 38% |
| 200 | 60.5% | 2.02:1 | 41% | 39% |
A Step-by-Step Guide to Implementing a Robust Fractional Kelly Strategy
- Gather a large sample of historical trades: The more data you have, the more accurate your estimates of W and R will be.
- Calculate the point estimates of W and R: Use the historical data to calculate the win rate and the win/loss ratio of your trading strategy.
- Calculate the full Kelly percentage: Use the Kelly formula to calculate the full Kelly percentage.
- Choose a fractional Kelly multiplier: Decide on a fraction of the full Kelly that you are comfortable with. A common choice is 0.5 (half Kelly).
- Adjust for estimation error: Use a Bayesian approach or a simplified formula to adjust the Kelly fraction for estimation error.
- Calculate the final position size: Use the adjusted fractional Kelly percentage to calculate the position size for your next trade.
In conclusion, the Kelly Criterion is a effective tool, but it must be used with caution. By using a fractional Kelly and adjusting for estimation error, traders can harness the power of the Kelly Criterion while mitigating its inherent risks. A robust and conservative approach to the implementation of the Kelly Criterion is essential for long-term success in the markets.