TraderZones - why would you recommend to stay away from Optimal f. I am currently reading Ralph Vince (Handbook of Portfolio Mathematics) and he seems to make a point - although he's the most biased person on this topic! ;-) I'd be interested to know your point of view on that and why you would not recommend it?
Optimal f sizing makes a potentially bad assumption about underlying distribution based on historical data. One or two fat tails can destroy you based on that assumption. There's the rub.
The other option I would have suggested is a volatility-adjusted variable fractional. This basically rejoins some of the earlier posts: determine and ATR-based stop-loss. Calculate the position so that the loss amount is a given percent of your equity and apply an anti-martingale: increase the percentage in winning periods and decrease it in losing ones. You probably want boundaries on min/max pct (which I would suggest are based on Optimal f values: ie half of optimal f)
It seems that the common theme which is repeated over and over here is to keep everything proportional in terms of sizing, bumps up or down in sizing, and stop-loss levels. Good stuff. Many traders, including myself, can get a little freaked-out if I do something rash like just double my size from one day to the next (one lots excluded). I like to keep my bumps up in size somewhat innocuous so as to minimize any subconscious or conscious trepidation taking entry or exit signals. All depends on the individual.
+ I have seen enough position sizing articles concluding that optimal f is a dangerous way to go based on analysis. There are better methods, as has been said above. But I think 2 important things to remember: 1) No mathematical tweaking overcomes the fact that you are still trying to increase leverage. That brings increased risk 2) If you don't have a significant edge, position sizing will make that very apparent.
There are several aspects of trading that come into play for position sizing. The market's capacity varies over the day. What I use to check this is the number of partial fills that comprise an execution. I feel most comfortable with under 8. The time lapse of the filling, I like to see not exceed a second. The next ranking considerations are the trend and bar volatility and bar overlap. All my comments are in terms of maintenance margin. I trade on the observable trend traverses only and I do not jump fractals in either direction. Bar volatility also involves bar overlap as seen on the lookup tables for filtering trades according to market pace. This effort in my routine focuses primarily the difference between using maintenance margin and using the position buying power handles. I use a mental equation which is the difference of the two and I think of this difference as a number of dollars that is a multiple of a contract's maintenance margin. As I look at three concurrent volatilities (bar, overlap, and trend observable container heighth), I make sure that the difference cited handles these values in terms of the magnitude of the given trade position.. Ordinarily it is a four digit or low five digit value in cash terms. As a trade ensues the whole amount of the first cited difference accumulates rapidly as unearned profits that show on the position pane. This just means I have reached a certain time in an ongoing trade and that part of the inital trade risk has come to a close; it has zeroed out, so to speak. Next, I focus on the relationship of dealing with the other part of a profit segment; the end effects. Here the focus is on price progress to the extreme value of the traverse of the profit segment. Here money velocity is how position size is determined. I do not take off contracts ever, but there is a question of how to use the excess unearned cash (unearned by virtue of not having itl ocked in by an execution). I measure it in terms of what it represents as maintenance margin or multiples of, say 1500 dollar increments.) I do not do division since my excess buying power states this value in terms of contracts directly. It is a comparison of the order already set on the execution platform (which coincides with my contemporay positon at any time in the day) compared to the actual present buying power which is stepping as a postage stamp curve would. Lastly, there exists a time window whose width is the first exit point in time and the last exit point in time. This is precisely an overlap of trending period that is well defined; it has a beginning and an end. (Google SKO). During this period, I look mostly at risk going away and then beginning to return. It is a matter of the dominant or non dominant residence time(calculated as a mental ratio) of the price BBid/BAsk residence time. There is also translation to the limit and away from the limit. (The market makes this limiting value available well in advance.) The optimum position management comes at least risk when the acceleration goes through zero. Since I have to trade twice as many contracts as my position; the first half is to realize profits and the second half is to establish a proper new management position. I do it in three steps: out, in and add ons to go through the above cycle from top to bottom. This approach is not for edge based traders. This approach is a construct that is related to always having in hand what the market is doing and trading at a multiple of the market's capacity. I do a weekly adjustment of capital (with regard to market capacity) to sweep proifits into other applications.
+1 Great post. I need to look further into this, it makes complete sense but its just never crossed my mind before. Thanks for sharing.
Tharp's book had a pretty decent section on position sizing (I found it informative, anyway). I think the Turtle method also had a very simple way of reducing position size as well...something along the lines of a 5% drawdown means you assume a 10% loss in available equity, etc. Basically you get some "friction" while you slide in the losing direction and some "lube" when you're heading in the right direction.