Hi, You can hear many rumors, ideas and not-so-true stories about pyramiding, I decided to do some research of my own. I have to stress, that my research is pure system based. So I don't have any discretionary elements in my tests. I'm trying to avoid arguing, and try to calculate some figures which show me what is true and what is not so true. What I basically do is: - write some strategies - test them with and without pyramiding - compare the results, based on some key figures It's as easy to describe as it's hard to do Let's start with a definition. I see pyramiding as "adding to an already open position". I'm well aware that this is not the classical definition ("using unrealized profit to add to an already open position"), but for me it was more convenient to stick to this more "relaxed" definition. For the first step I focused on trend-following strategies. So I copied some from various sources, and executed them on several instruments of several market in different time frames. Next I added pyramiding - adding positions of constant size, increasing and decreasing size - and compared the results. Although I'm not yet done with my research, I now would like to share results and hopefully initiate some discussion. As a first step I attach a summary Excel-Sheet showing some examples (1 strategy, 3 markets) of improved performance by applying pyramiding. Please note columns: - MRE: showing the "average max risk exposure" of a trade (smaller figures are better, since risk is reduced) - AP/DD: annual profit / max drawdown ratio - PF: Profit factor Many details are still unmentioned (how are key figures calculated ? what market/timeframes/instruments? how can different markets be compared ? ...). If this issue is welcome, I would be more than happy to share details. Dierk
i view it as an approach that acknowledges what we all know - even if we only admit it in the wee dark hours of the night - which is that collectively we all suck at calling turning points. look forward to seeing more.
Attached you find details about the test setup. A lot for a first step, I know. But since I want let my numbers speak, I felt that you should know what I'm doing. I would appretiate any comment in the test setup. If there's some interest in this issue, I'll continue the thread and publish results and (my proposed) conclusions. Dierk PS: Here in the south of germany it's 37 degree celsius. I better should go to the pool ...
@droth Since you tested commodities, did you take limit-days into account? These days might yield some profits in your testing while in fact you could not trade at all, or, even worse, would have made a substantial loss, because you could not close a position. I do not know how large the impact on the overall results are, maybe it is negligible. If you like this kind of testing, the book "The encyclopedia of trading strategies" by Katz / McCormick might be of interest to you. regards Bernd Kuerbs
seems that this would come to an academic thread. I'm not smart enough for that, but looking forward to an interesting discussion. I've done a bit research with position size but without all this numbers. I'm only interested in how the equity curve of the portfolio looks like and the average yearly profit to the max DD. I found 2 things. If you trade trendfollowing, above water pyramiding can make sense. If you trade counter trend, under water pyramiding can make sense. But it always depends on which markets you are trading where you step in a trend, from the bottom at a range break out or when the ADX is well above 25 and so on. btw I don't think you can impress the american guys with 100° Fahrenheit
Interesting, but ... the results could not be applicable for real trading systems. The drawdown for some securitiies is > 50%. Such systems not for real trading. I would see the following steps for testing: 1) Develop system that can be traded, Max DD < 20% 2) Explore if pyramiding can increase the performance
@BKuerbs (we already hat PM contact) No, I don't take into account limit days. This is not for real trading. The turtletrader data is much dirty to base some "real" strategy on it. @NYSEtradinglog I'm not yet smart enough for real trading, so I do some academic stuff @a5519 Well, I think if there is something in pyramiding, it should be obvious on applying to some "simple" strategies, even when these strategies are non-profitable. So I start with some simple demos. I course they can't be traded for real. Here we go, 1st strategy is as simple breakout strategy: - go long when current high is higher than highest high of last 25 periods - go short when current low is lower than lowest low of last 25 periods This strategy is always on. I would like to point out again, that this strategy is not for being profitable. I only for demonstration purposes concerning the effects of pyramiding. You'll find the results in http://mitglied.lycos.de/ddroth/Pyramiding file "Breakout.zip". "Pyramiding.txt" holds a slightly updated version of the "regulations". These spreadsheets contains several pages please focus page "summary" and "results". "summary" holds a summary of all instruments - mean values are calculated as described earlier, "results" holds the details results for each tested instrument. Column "M" or "Magic" is the "adjusted return" as described earlier in the details paper. - EUREX futures (1Breakout0-EU*.xls) No improve can be seen. Instead some of the major key figures got worse (e.g. PPM). Looking at the result page we'll find that pyramiding mainly had an effect for FDAX: although optimization improved "M" (significant reduce of the MRE), the overall picture is worse. It also can be seen, that the optimizer found the best solution on placing the pyramiding orders directly at the initial entry price (column "PyramidATR" values are 0 for FESX, FGBL,FGBM). This is a clear indication that pyramiding here makes no sense. - Stocks (1Breakout0-TKD*.xls) We can the the pyramiding had positive effects. * SR improved * DD decreased * PPM increased ... Overall a nice picture. Please note, that the optimal pyramid type seems to be type "2" (increasing position sizes, starting with 1, next 2,3,4,5). - Commodities (1Breakout0-TT*.xls) Again an improve which can be seen by better key figures. Again pyramid type "2" seems to be optimal. Please not, that the distribution of CumProfit (on the "summary" page) tends to be smoother by applying pyramiding (StdDev of CumProfit decreased). Ok, that was the fist step. I'll be back ... Dierk