I've also been researching on pyramiding / scaling entry using multiple methods suggested by my trader friends. Bascially the result is very interesting. Briefly, if pyramiding is done correctly, it will decrease the winning rate but increase the expectancy / profitibility of any logical system trememously.
Pyramiding into a trend
- Thread starter ElectricSavant
- Start date
I've also been researching on pyramiding / scaling entry using multiple methods suggested by my trader friends. Bascially the result is very interesting. Briefly, if pyramiding is done correctly, it will decrease the winning rate but increase the expectancy / profitibility of any logical system trememously.
Yes...pyramiding dampens the win percentage because some of the pyramided contracts/lots will turn into losers and drag down the original trade. Yet those that remain winners will have their win size improved to a large extend.
I would harzard a guess that the longer the trend that are present in winners, the higher profitability pyramided winners have?
Thus is it proper to conclude that pyramiding are most excellent for systems with high tolerance with regards to lose % and expects to ride long trends??
I am very weary of overfitting. I generally treat backtest results as âindicationsâ of market tendencies, rather than more reliable behavioral characteristics. As such, calculating probabilities and expected returns, etc and using these to compare âsignal strengthâ does not appeal to me as being robust.
Including more conditions (pyramiding) to a strategy reduces the degrees of freedom and increases the chances of âfitting to past dataâ exponentially.
If a signal is sufficiently good to act on â i.e. it makes money (and also makes sense and passes the robustness mustard) then I put the risk on. If it is a âweakâ signal then why risk a penny? Beware of a strategy that relies on pyramiding to work (alternatively: beware of the excess results pyramiding generates for a strategy).
As to a second signal confirming the previous âweakerâ one: it sounds like you ought to wait for the confirmation in the first place (or you donât need the confirmation at all â make a choice about which version of the signal can generate better results).
Including more conditions (pyramiding) to a strategy reduces the degrees of freedom and increases the chances of âfitting to past dataâ exponentially.
If a signal is sufficiently good to act on â i.e. it makes money (and also makes sense and passes the robustness mustard) then I put the risk on. If it is a âweakâ signal then why risk a penny? Beware of a strategy that relies on pyramiding to work (alternatively: beware of the excess results pyramiding generates for a strategy).
As to a second signal confirming the previous âweakerâ one: it sounds like you ought to wait for the confirmation in the first place (or you donât need the confirmation at all â make a choice about which version of the signal can generate better results).
Definitely, trend-following system made the most in trendy period.
Depending on the pyramiding method, such system doesn't always have a big draw-down. I see pyramiding as a entry signal and since my entry limited only to small losses, my overall system doesn't have a big draw-down. It just lesser the winning rate.
Without pyramiding (example only),
Winning rate - 50%
Breakeven rate - 20%
Losing rate - 30%
With pyramding,
Winning rate - 20%
Breakeven rate - 40%
Losing rate - 40%
Well, that's what you do whenever you compare prospective trading systems before you decide which one to trade. Aren't you contradicting yourself? You need to test your system to see the estimated likelihood of it beating your required hurdle rate - and you are comparing system strength when you do this. How can you compare system strength and not signal strength?
Also, you didn't give any reason why you shouldn't take a weaker signal or system just because you also have a stronger one too. Saying it "sounds like" you should do something is not a valid argument.
To Cuttenâs remark, (I hope I understand and address your criticism)
My backtesting kicks out a variety of performance measures (sharpe, semi-variance, drawdowns, coverage, efficiency, consistency, to name a few). There rarely is a clear winner when comparing simple strategies, since the various performance measures compare/order differently. My solution has been to use judgment in deciding if a âruleâ is good enough to trade (based on the overall showing of the various measures). Once a rule passes that hurdle, then it is treated as equal to all others. My (put-my âmoneyâ¦) experience supports that methodology (though more out of sample points are needed to construct a statistically valid inference).
If I did not address your criticism, perhaps I did not understand what you meant and you could elaborate.
My backtesting kicks out a variety of performance measures (sharpe, semi-variance, drawdowns, coverage, efficiency, consistency, to name a few). There rarely is a clear winner when comparing simple strategies, since the various performance measures compare/order differently. My solution has been to use judgment in deciding if a âruleâ is good enough to trade (based on the overall showing of the various measures). Once a rule passes that hurdle, then it is treated as equal to all others. My (put-my âmoneyâ¦) experience supports that methodology (though more out of sample points are needed to construct a statistically valid inference).
If I did not address your criticism, perhaps I did not understand what you meant and you could elaborate.
Re âsounds likeâ.
I am not sure how to put this any simpler.
The two signals are distinct. The first one says âif A then go longâ. The second one says âif A AND some move in my direction then go longâ.
To me, these are two different signals. Their backtests will have different results. Subjectively, one can decide that (1) the two ârulesâ offer essentially the same promise or (2) one is clearly superior. My point is that that does not matter! What matters (is not rule superiority (hard for me to effectively evaluate amidst so much uncertainty)) but that one, both, or neither are worth risking capital on. To me thatâs keeping it simple (âbut not simpler â¦â
I am not sure how to put this any simpler.
The two signals are distinct. The first one says âif A then go longâ. The second one says âif A AND some move in my direction then go longâ.
To me, these are two different signals. Their backtests will have different results. Subjectively, one can decide that (1) the two ârulesâ offer essentially the same promise or (2) one is clearly superior. My point is that that does not matter! What matters (is not rule superiority (hard for me to effectively evaluate amidst so much uncertainty)) but that one, both, or neither are worth risking capital on. To me thatâs keeping it simple (âbut not simpler â¦â
Q
http://www.soyroy.com/murphys_law.htm
11. If you feel like doubling up a profitable position, slam your dialing finger in the drawer until the feeling goes away
UQ
http://www.soyroy.com/murphys_law.htm
11. If you feel like doubling up a profitable position, slam your dialing finger in the drawer until the feeling goes away
UQ
Ok, you are using an "either/or" evaluation for your systems as well, fair enough (although I would question how do you "judge" if a system is worthwhile - if for example you had an alternative use of your capital that made 25% a year, how would you know whether to keep trading or not?). My basic point is that trade expectation will sometimes have an identifiably significant improvement during the lifetime of the position. If that happens, then clearly any position-sizing algorithm must increase exposure to exploit the far superior returns.
I gave one example (Sep 11th) where I would consider that expectation improvement to be undeniable. However, obviously I think it occurs more frequently - examples would be the parabolic moves that sometimes occur at the ends of major trends, where the price move causes increasing panic and thus increases the linearity of the move (which makes a momentum strategy much more profitable than during normal "noisy" market action). Another example would be owning a stock and then seeing a takeover announcement - obviously it would be insane to not buy the now-stale offer. To oppose pyramiding, you must believe that the market *never* presents an identifiable increase in trade expectation - and that is a pretty strict condition which IMO has been refuted by countless market observations.