Reverse Martingale position sizing

[tech/a]

Say I have a win rate of 80+% on a very basic system. Small profitstop of 1% and wide stoploss of 5%. Max consecutive losers is 2 on any bluechip tested. Max consecutive winners of around 10. The equity curve is basically flat for most securities, but can be optimized to slope upwards.

I wonder how this would go with a Martingale.

Recovering 5% each loss is going to make for a huge position size after 2 losers. I guess one could start small.

Any ideas?

Personally I'd play around with the win rate
I'd want to increase the profit stop while decreasing the stop loss
The objective is to decrease the gap when you have consecutive losers.

If you ever got 3 or 4 consecutive losers you'd need very deep pockets and may never get back.
Just my take.

 

[howardbandy]

HI GB --

I discuss the importance of using the entire distribution of trade results to answer questions such as you are asking in my book "Modeling Trading System Performance."

In short summary:
1. Final equity or terminal wealth, TW, of a set of trades can be computed as the geometric return per trade, G, raised to the power of the number of trades, N. TW = G ^ N. You must have a positive G -- a positive expectancy. You want a large N.
2. While the final equity of a set of trades is always the same without regard to the sequence in which they occur, the drawdown is very sequence dependent.
3. The reason most traders stop trading a system is that the drawdown exceeds their personal risk tolerance.
4. Drawdown is determined primarily by the number and size of losing trades.
5. Position size is independent of the logic of the trading system, including being independent of the exit methods, including whatever profit target and maximum loss exits are included in the system logic. Position size depends on the relationship between the health of the system (the synchronization between the logic and the data over time) and the trader's personal risk tolerance.

Given a list of trades (which can be any combination of real trades, paper, out-of-sample, in-sample, or hypothetical), you can determine the risk of drawdown, then maximum safe position size, then profit potential, using the techniques I describe.

You need the list of trades, or a probability density function of the distribution that represents those trades, in order to do the analysis. Knowing only the expectancy, or the mean and standard deviation, or even the first four moments (mean, standard deviation, skewness, and kurtosis) is insufficient.

Best regards,
Howard

 

[Gringotts Bank]

Thanks tech and howard.

Howard, I followed along until those last 2 paragraphs. If possible, could you explain in more basic language? Thanks.

 

[howardbandy]

Hi GB --

It took a large part of the book to explain the background, the concepts, and the techniques. See if some of the articles on my blog site help:
http://www.blueowlpress.com/WordPress/

There are a couple of chapters of the book that give an introduction on the book's website:
http://www.modelingtradingsystemperformance.com/book.html
but the meat is in the chapters that are not posted.

Regards,
Howard

 

[Gringotts Bank]

Hi GB --

It took a large part of the book to explain the background, the concepts, and the techniques. See if some of the articles on my blog site help:
http://www.blueowlpress.com/WordPress/

There are a couple of chapters of the book that give an introduction on the book's website:
http://www.modelingtradingsystemperformance.com/book.html
but the meat is in the chapters that are not posted.

Regards,
Howard

Thanks, will do.