Position Sizing Optimization

Questions about MultiCharts and user contributed studies.
Posts: 261
Joined: 28 Feb 2008
Has thanked: 2 times
Been thanked: 1 time

Position Sizing Optimization

Postby flipflopper » 29 Mar 2010

One thing that I have always wondered about is position sizing during optimization. I think that you are NOT supposed to vary position size during optimization as a rule but below is the point logic I am weighing.


* You are willing to risk a fixed amount on each trade.
* You are optimizing the stop and target values that are based on ATR.


* The optimizer quickly realizes that the tighter your stop is the larger the position size can be.
* The optimizer thinks it is most profitable to trade HUGE size and super tight stops and eventually you will hit big.
* Reality is with slippage and savvy market makers these strategies are not realistic at all.

My issue:

* How do you balance the TRUTH that when you have a tighter stop you CAN trade larger size and risk the same dollar amount with the REALITY that when you trade larger size slippage is bigger and win rate becomes very low and paper results become a fantasy.
* In other words how do you tell your optimizer not to get too carried away with the concept of tight stop larger size hit a homerun and make the most money?
* I want to find the sweet spot between position size and tight stop.

Posts: 79
Joined: 18 Feb 2010
Location: Dallas, TX

Postby ctrlbrk » 30 Mar 2010

The way I do this is setup different entry triggers within the strategy, and then assign a weight to each signal that you can optimize against. This will basically tell the strategy to optimize the significance/profitability of each trigger within my strategy.

You could then easily multiply that weight or whatever to get your position size, and also wrap a MaxList/MinList around it to prevent it from going crazy.


User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times

Postby Bruce DeVault » 30 Mar 2010

This is a complex subject, but to over-simplify quickly in the interest of the "big picture" discussion, it's generally helpful in quantification and strategy improvement work (which is what most people are doing with optimizers, although there are some exceptions) to optimize your stop and target considerations separately from position sizing.

While in "real life" you might scale your position based on % of equity at risk etc., if you do this during optimization you will introduce a starting point dependency such that your optimization would get different results for a particular current point in time, just because for instance you included an extra day or month in sample, and those kinds of starting point dependencies are generally "bad" in that they distract and confuse from what you're actually trying to measure which is how one set of parameters compare with another or how performance has changed as adjusted for the addition of more recent data with the same parameters.

You also generally don't want your optimization in this case to take into account "compounding" which is the simple idea that as your equity increases you trade bigger size - the reason is much the same as the above - what you want in general to assess which input parameters give you the "best" performance on typical measures is to trade a "fixed" size e.g. always 10 contracts of a commodity, or always X dollars of an equity - in this way you're not starting point dependent, and the optimizer can focus on getting the stops and targets right without getting wrapped around the axle with the question of how many $ you have in current equity right now, which really could be anything depending on where you started.

In summary, the way you set up a strategy when you're trying to reoptimize or determine how to make it better is often different than the way you would deploy a strategy in the real-life context of trading it in your account together with other strategies and instrument/time frame combinations.

The reason it's different is so you can get a more statistically meaningful, cleaner measurement of what it is you're varying, and so you can see through the skew introduced by the position size varying, which for instance for a profitable strategy would result in great increases in the volatility of the equity curve later in the test due to position size being increased, when in fact, the strategy's performance statistics aren't inherently then more volatile.

Remember, if you start trading a strategy right after your optimization, that's when the compounded optimized strategy would have its largest size, yet that isn't the size you would be trading because you don't "really" have the profits from the optimization trades in sample to work with, so introducing this tends to make the results less useful rather than more for helping you understand its characteristics in the near future - and that's important so you can know if it's working or if you should re-evaluate.

glen demarco
Posts: 35
Joined: 16 Nov 2009

Postby glen demarco » 30 Mar 2010


Very well said and informative.

Money management is a fascinating topic and some claim more important then entry and exit criteria. There is something called "Kelly bet size" for those interested in the topic. Some of the best articles I've seen on the topic are on www.tssupport.com in the knowledge database search on money management.


Return to “MultiCharts”