rubbish. an optimal positioning strategy is central to any real successful strategy. to use fixed-ratio you're more likely to break-even. a poor strategy can be made profitable by just optimising its positioning of trades. potentials of a good strategy is difficult to discover without any probable positioning strategy. JMO.
Ok so I went to show an example of what I was worried about and picked a sample strategy that had made me think of this question and because of this discussion it led me to do more testing. The initial 2 parameters I used were 5k fixed size vs a 1% fixed ratio. You can see why from the picture I was so concerned about using the fixed ratio. Now I'm thinking you just need to have a big enough bet out there to allow the strategy to suffer through drawdowns before giving up on it. Now the problem I'm seeing is that even with fixed size bets that survive the drawdowns, the Sharpe varies. Ideally, one would want something that's consistent regardless of position size (% or fixed). Does that mean that Monte Carlo is really the best way to go? Edit: For those who don't feel like downloading the picture, the 1k fixed and 1% ratio tests both ruined. The 5k to 10k fixed and 2.5% and 5% ratio tests had Sharpe's varying from 1.8 to 2.6. p.s. I tried getting the image to show up but couldn't - does anyone know how please?
Finally, somebody who gets it. Fixed-size trades aren't anywhere near optimal. They can't be rigged to even approximate optimality. Fixed-ratio, or fixed-fractional (if there's a technical difference between the two labels, save it -- it doesn't matter in the end), trades might be optimal, but since that can only be determined after the fact, they don't offer optimality for a going-forward goal. The only use of a fixed-size or a fixed-ratio trade scheme is in the beginning, where you don't know what the trade distribution is and you have to establish a baseline. But after the first 30 trades or so, you should switch from fixed position sizing to an antiMartingale sizing strategy, like Kelly sizing. Look at it another way: the easiest way to beat buy-and-hold is by using leverage. And the only intelligent time to use leverage is when the odds are in your favor. And no way does fixed position sizing allow you to ever use leverage in an intelligent manner. You have to use some adaptive position sizing strategy like Kelly to do that.
This is what I'm asking. I'm not asking about the optimal trade size nor am I asking about money management at this time. I was just asking about determining a strategies potential during backtesting. The quote above sounds like you might think it would make sense to use a fixed strategy for backetesting - am I interpreting that correctly? Thanks.
Read the next sentence I wrote after the one you quoted. If all you have are 30 trades in your backtest, it's a pretty weak backtest. In other words, you should switch to adaptive position sizing during your backtest at some point -- 30 is just a suggestion, not an ironclad rule. To give an example, my position sizing formula starts as follows: @MIN(leverage, @MAX(kmin, @IF(tradecount < kstart, kzero, { formula for position sizing } ))) where leverage is determined by the type of instrument (e.g., 2 for US stocks), kmin is the minimum allowable value of the Kelly ratio (>0, to keep you in the test), kstart is the start-up period (e.g, 30) and kzero is the position sizing fraction during the start-up (>= kmin). HTH
I see, so you suggest varying the position size even during backtesting - possibly starting with a fixed size but when/if the equity increases enough then you use the actual money management technique you use during trading. So that goes with what's seeming to become a consensus - i.e. to test the way you trade...
How come the number of trades in the backtest changes as the position size changes? I thought you were comparing the same trading strategy (entries and exits) under various position sizing strategies. I would have expected the number of trades under each column to be the same.