In this blog, I’ll be looking at the way I would calculate my position size using a dollar criteria.
The inputs are:
- The average dollar win per trade.
- The average dollar win on a one-contract basis.
- The maximum dollar loss
- The volatility of the market as measured by the stop location. My stops are based on swing extremes. For this reason, the volatility are an automatic part of the stop calculation.
The process takes place over 4 steps:
Work out the units to determine the maximum loss. I do this by:
- Avg$Win per trade/Avg$Win on a single contract basis. For example my Avg$Win per trade is $12K, and Avg$Win on a single contract basis is $1k. My units are 12k/1k = 12.
- Multiply the units by 3 (as a buffer against a Black Swan). In this case, 12 x 3 = 36.
- The Avg$Win per trade is the total profits/number of winning trades. The Avg$Win per trade includes the total number of contracts of every trade. The Avg$Win on a single contract is the average $ profit of all winning trades on the basis that each trade had only 1 contract.
2. Work out the maximum loss: calculate the mean and standard deviation of losses. Our loss should not exceed this figure. This is standard statistics. So, let’s assume that our mean is $9k, and the standard deviation is $3k. The maximum loss will be $9 + 3×3 = $27k. I normally round up to the next ‘5’ or ‘0’. In this case, the maximum loss would be $30k.
3. I then divide step (2) by step (1). In this case: $(30,000/36) = $833. Rounded up, this gives a maximum dollar loss of $850.00.
4. The $850 represents the Normal Position Risk and controls the normal position size. Let’s say our entry and stop show a risk of $400. This means we can take 2 contracts. If the risk is $1000, we would have to skip the trade.
The advantage of this approach is it utilizes our trading statistics. But it does have one drawback, it fails to normalize. By that I mean that it assumes, for example, that an $800 movement in the ES is the same as an $800 in oats. This is clearly untrue. I get around the problem by including in my stats normalization of results using the initiating trade price.
When I first did this, I struck two problems but I eventually solved them. Since the material forms part of the seminar material in August 2008, it would be unfair for me to disclose the problems and solution here. However, the process above is still better, in my view, than many of the position sizing algorithms currently available.