@nubcake: Are you confusing profit factor with expectancy? If W = ave net win size, L = ave net loss size, and R = win%, then:
expectancy = WR - L(1-R)
profit factor = WR / L(1-R)
Thus, for an 'edgeless' system, and ignoring transaction costs, W=L and R=50%, so
expectancy = (Wx0.5) - (Lx0.5) = 0
profit factor = (Wx0.5) / (Lx0.5) = 1
Or perhaps the confusion is coming from my definition of MM, as follows.........
@nubcake, @PipMeUp:
It seems that you're confusing MM with exits, or trade management. I use the definitions given in Dr Ryan Jones' book, p4: "Money management, as defined here, is limited to how much of your account equity will be at risk on the next trade. It looks at the whole of the account, applies proper mathematical formulas, and lets you know how much of the account you should risk on the next trade.". If we use this definition, MM is about position sizing and method, i.e. fixed frac, fixed ratio, optimal-f, etc — which, if I understand correctly, is what is being discussed by the OP in this thread. Whereas trade management is about exits, i.e. initial SL and TP, subsequently moving SL to lock in profit, etc. I explained my thoughts in more detail recently here, and why IMO it's important to make the distinction between them. So if you decide to use a 20 pip SL, that's trade management; and then if you decide to risk 1% of your equity, or $125, or trade 0.15 lots, that's MM.
I've discussed Van Tharp's coin toss entries (invariably a 'hot' topic) in some detail here. Dr Tharp is talking about exits, not MM (see footnote below).
If you isolate the MM component, then unless you meet the criteria explained in my previous post (i.e. autocorrelation, or an edge from entries and/or exits), your net expectancy will be zero. That is the hypothesis that I'm putting forward. The attached XLS can be used to test different MMs. It assumes independent events, and thus attaches equal weight to all possible 64 distribution outcomes from 6 trades. Satisfy yourself that no matter how you vary your position (bet) sizes, the total net expectancy across all possible outcomes is always 0. The only way to gain an edge is to tinker with both the win rate and RR settings, an imbalance between which is ultimately created by the efficacy of entries and exits. You can extend the XLS to cover any number of events (trades); the results remain the same. Among MMs touted by FF members, I've tested martingale, anti-martingale, d'Alembert, Guetting and Oscar's Grind — all of these MMs (in isolation) have a total net expectancy of 0. I'm still trying to find one that doesn't.
David
___________________
[EDIT] Footnote — from Dr Tharp's book, p 200, "Our random entry system — consisting of random entries, a three times volatility trailing stop, and a simple MM system involving 1% risk — made money on 100 percent of the runs". Isolating the components, and using common sense, the edge can't be coming from simply sizing consistently at 1%, nor from the random entries, hence it is the trailing SL (i.e. exits, as opposed to MM). It would be interesting to run tests on different FX pairs/timeframes, to see if that still holds true, under current FX market conditions.
expectancy = WR - L(1-R)
profit factor = WR / L(1-R)
Thus, for an 'edgeless' system, and ignoring transaction costs, W=L and R=50%, so
expectancy = (Wx0.5) - (Lx0.5) = 0
profit factor = (Wx0.5) / (Lx0.5) = 1
Or perhaps the confusion is coming from my definition of MM, as follows.........
@nubcake, @PipMeUp:
It seems that you're confusing MM with exits, or trade management. I use the definitions given in Dr Ryan Jones' book, p4: "Money management, as defined here, is limited to how much of your account equity will be at risk on the next trade. It looks at the whole of the account, applies proper mathematical formulas, and lets you know how much of the account you should risk on the next trade.". If we use this definition, MM is about position sizing and method, i.e. fixed frac, fixed ratio, optimal-f, etc — which, if I understand correctly, is what is being discussed by the OP in this thread. Whereas trade management is about exits, i.e. initial SL and TP, subsequently moving SL to lock in profit, etc. I explained my thoughts in more detail recently here, and why IMO it's important to make the distinction between them. So if you decide to use a 20 pip SL, that's trade management; and then if you decide to risk 1% of your equity, or $125, or trade 0.15 lots, that's MM.
I've discussed Van Tharp's coin toss entries (invariably a 'hot' topic) in some detail here. Dr Tharp is talking about exits, not MM (see footnote below).
If you isolate the MM component, then unless you meet the criteria explained in my previous post (i.e. autocorrelation, or an edge from entries and/or exits), your net expectancy will be zero. That is the hypothesis that I'm putting forward. The attached XLS can be used to test different MMs. It assumes independent events, and thus attaches equal weight to all possible 64 distribution outcomes from 6 trades. Satisfy yourself that no matter how you vary your position (bet) sizes, the total net expectancy across all possible outcomes is always 0. The only way to gain an edge is to tinker with both the win rate and RR settings, an imbalance between which is ultimately created by the efficacy of entries and exits. You can extend the XLS to cover any number of events (trades); the results remain the same. Among MMs touted by FF members, I've tested martingale, anti-martingale, d'Alembert, Guetting and Oscar's Grind — all of these MMs (in isolation) have a total net expectancy of 0. I'm still trying to find one that doesn't.
David
___________________
[EDIT] Footnote — from Dr Tharp's book, p 200, "Our random entry system — consisting of random entries, a three times volatility trailing stop, and a simple MM system involving 1% risk — made money on 100 percent of the runs". Isolating the components, and using common sense, the edge can't be coming from simply sizing consistently at 1%, nor from the random entries, hence it is the trailing SL (i.e. exits, as opposed to MM). It would be interesting to run tests on different FX pairs/timeframes, to see if that still holds true, under current FX market conditions.
Attached File(s)
Martingale comparison.xls
125 KB
|
360 downloads