I have noticed one thing when developing strats is that intraday price action strategies (ones which look for recurring themes in the data e.g. 1min chart), these strats can perform excellently with very little draw-down (sometime less than 1.5%) but yet they always seem to breakdown. I only have emini data upto 2007 and hence most of my testing is done for last 6 years on futures markets. I have found some strats do amazingly well on this data and test well across NQ,ES and YM. However, when i then test this on the SPY or QQQ i find that the results perform poorly prior to 2007. I know the results from cash market to futures will differ as anyone with half a brain can look at 1min chart of ES and 1min chart on SPY and see there are small differences. Some of these price action strats have well over 1000 trades over 6 years and very few optimizing parameters. Also, the entry edge was tested and developed separately from stops and targets. Yet i find no matter what, any price action type strategy will break down. Moreover, i only optimize any parameter right at the end of development, this is after i have ensured at-least 1.5 profit factor usually. Please note, what i am referring to price action is not support or resistance or any type of level, i am referring to recognizing patterns with short time frames of data. Is this normal for price action strats to breakdown on these short timeframes or is it just bad design and development on my part. All my other strats which are not pattern based or indicator based do not breakdown and usually perform well across all markets, however they are usually of slightly longer trade duration. Any advice?
Are you saying "since-2007" is In-Sample, and "pre-2007" is Out-Of-Sample, and asking why the strats don't seem to perform Out-Of-Sample? Thanks.
You just answered your own question. You're not trading price action; you're trading patterns and indicators, both of which will fail you over time. This doesn't mean that you're required to abandon them if you really like them for some reason. But you won't be trading price action. That entails support and resistance, demand and supply, and trend, none of which lend themselves to computerized backtesting or mechanical trading strategies and tactics.
No, i usually use 3 random years from 2007 to 2013 as in sample and then the other 3 years as out of sample for futures market strategy development (as my esignal data only feeds me 6 and half years worth of emini 1min data). Then i usually test any strat developed across its equivalent cash market and therefore you could say even further out of sample is from 1997 to 2007 on cash markets. I find nearly all my price action based strategies breakdown at some point in the 1997 to 2007 testing no matter what. e.g. i am firm believer that each market behaves differently, however i have a price action strat which looks at 1min data on futures markets from 2007-2013 - this strategy clocks up 1000 trades on average on one instrument. I have tested this across 16 different futures markets from 2007 - 2013 (ES,NQ, YM, Oil, Gold, Silver, Wheat etc.). This strat performs well on all futures data from 2007-2013, it also does reasonably well on 2007-2013 SPY,QQQ. Hence, you would think this strategy can't be curve fitted as it was developed on one market and then subsequently seemed to work on 16 other markets. It the precedes to work on some cash markets. But when i then tested this price action strategy from 1997 to 2007 on SPY, QQQ (as i have 1min data for these all the way upto 1997) i find this strategy breaks down and hits about break even during this period. So what i am thinking is that no pattern or intraday theme will work throughout the ages - i have a feeling they all breakdown no matter what. A short term strat which trades frequently and can be sized up with very little drawdown is somewhat of holy grail, i think its near impossible to get an edge in these types of strategies which will not come to ruin when there is a regime shift in the markets. Have others found this to be the case with price action strategies or do i need to work on this further?
No indicators were used. Yes somewhat i am trading patterns in these strats - nothing like head and shoulder, wedges and so forth like any books states. The price action i am talking about is recurring themes to the fractal nature of the markets, or patterns which take advantage of the noise intraday. I agree with you point however i think anything can be computerized, anything a human can see so can a computer.
Of course they do. Swing up, swing down... trend change. Incumbent on you to get a "read" on the market and risk appropriately.
Nothing works all the time. Basically, you need different strategies for different market conditions and/or the knowledge of when to stay out of the market if you don't have a strategy that do well in those conditions. So, assuming you have a wide variety of strategies, the challenge is then to know what market condition we're currently in and what strategy to apply, when does it end and what is the new market condition we're transitioning to? Of course, easier said than done, but that's the gist of it. As always, my humble opinion.
Thanks for replies guys, yes i have many strats developed which will will trades both long and short and i try to even these out with RTM and trend. I don't play a theme to the market and have starts where some might be breakeven for the month and other will be doing well and so forth. This is just a question i was asking out of curiosity with intraday price action/pattern rec strats. Just wanted to know if you guys have strats like this that never fail or just guys just know when to pull them and start them up again.
Have you broken down the performance by time of day? Are you using time of day as a variable for the patterns?
I have 3 types of these price action/pattern rec strats developed for use on ES,YM and NQ and they all have same problem of breaking down in the equity cash markets SPY.QQQ. The only time of day i use is that they usually trades can only take place after the first 15mins of the day. Apart from that i don't want to use time of day that much because i feel it is over optimization. I have tested each hour of the day trades were made and it tends to be not that much difference, obviously i could tweak it here and there and increase PF but i don't.