Friday, December 4, 2009

My Take On Optimal Position Sizing

A subscriber recently asked me about my take on position sizing, and more specifically using Optimal f. For those unfamiliar, Optimal f is a calculation derived by Ralph Vince which computes the optimal position size that would lead to a portfolio’s fastest growth rate. Mr. Vince has written several books on the subject and I do recommend his work.

Here's my take on using Optimal f (or any other position sizing formula for that matter).

Optimal f may be the mathematically optimal size for a position IF the trades are managed in an optimal manner. For many people, positions of "optimal" size feel aggressive. While our perception is often that we have a large edge on a particular trade or strategy, the edge often isn't THAT great. In order to maintain your edge you need to be able to trade the strategy in an "optimal" manner. Whether the strategy is automated or discretionary doesn't matter. You need to be able to take entries when they trigger. You need to be able to take exits when they trigger. But just as important and sometimes most difficult is you also need to be able to SIT when there is neither an entry nor an exit triggering.

The sitting can often be difficult when traders have what they perceive as a large position that they are managing. Many traders have a tendency to want to manage the trade more carefully with large positions. This may mean tightening stops sooner or taking partial profits a little quicker. Doing this "feels" good and it may get the trader to a breakeven level faster, but in the long run it will almost always either reduce or destroy the edge of the strategy. Tighter stops and quicker profit taking effectively means more losers and smaller winners. Few strategies can withstand more losers and smaller winners and still look good.

Optimal f may be the mathematical optimal size, but in my view the optimal size is the closest you can get to Optimal f and still execute your strategies as they were designed - while keeping your sanity.


Toptick said...

Good points on the practicalities, Rob. I would add that you also would want to look at the assumptions behind optimal-f: that you have a good handle on the probabilities, they are stable, and that you know the maximum loss.

It's perfect for a roulette wheel, but not so easy to apply to investing/trading.

John said...

There is no more important topic than position size. Almost anybody can suggest a name to trade and even an entry price.

Few people know their exit price and even fewer know the most important question of all: how much to trade?

Michael said...

Your comments on tweaking a large position are very timely. I am in analysis paralysis on a new automated true reversal trading strategy around AUD/USD currency using 5m minute bars. Unfortunately, there are times when it is holding a position that continues to grow the unfavorable excursion because it is trending counter to my position. Analysis revealed that largest profits are in first 30 minutes and after holding more than 80 minutes there was not one trade that became profitable probably because the automated strategy reversed the trade eventually. Given that, I have yet to find a way to tweak the strategy to improve it, and just missed out on Thursday and Friday's big moves (more than $2k in less than 2 hours on 3 globex contracts) because of the time being spent to reduce risk. So far all risk reduction methods I have tried have resulted in at least 10% more trades, larger individual loss (seems contradictory), lower overall return, yet the max drawdown is smaller. Anyways, trending markets and normal mean reversion help to make risk avoidance difficult. So, these practical limitations as being verified in strategy analysis as well.

nzbryant said...

Exactly Rob!