Hi kut2k2, with one single asset, and EOD time frame no way. If you need 400-600 trades per period. 252 / 400 = 0,63 bars. Or If you have 30 trades per year EOD, you need more than 14 years historical EOD data using your numbers. But i don't want to validate 14 years of trading, i want to validate 1 year. This is the reason because i did the question. You are limited by the time frame where the strategy is working. Regards.
Well, 97,500, With a 15 minute chart having 26 possible bars, a 105 bar system requires a 105/15 1-2% ratio increase. so 9 times the data would be the 250*390=97,500 9% to 18% of the data into 105. 97500*0.09/105=83 to 166 trades. The 1-2% is from the highest exposure of systems being the most successful with the least amount. 1-2% of a 1 minute 15 minute trade average needs 9 times as many trades to be valid because the 1-2% is where the highest frequency trades have been with the most successful having low exposure meaning they aren't in the market very much of the time. At 1-2% r:r had very high returns and 15 minutes was a reasonable rate of time for 1 minute charts to use in duration. At 105 minutes just from experience this is more intraday trading than hft. If the only data that is available is 1 year you can't expect hundreds of trades to validate EOD data since more exposure is a sign of a less robust system.
@bwolinsky sir, what variables would we need to set a limit of exposure (average bars, holding time per trade) where the system starts losing robustness? (or please give me a reference where i can read more about that). I have some difficulties to understand. Where is the 9 come from? 105/15 = 7 Regards.
My mistake. I thought you were making a sane inquiry. If your question is how many trades you can stuff into 252 bars, my response is I don't know and couldn't care less.
Thanks anyway, i think it's no posible to validate one EOD/system in a one year period, is it not posible to stuff 30 trades into 252 daily bars. Sorry, maybe i do not get explain it well Best.
Again, Clowner, it's not the number of bars but proportion of it to the time in the market. You only need one more year to validate or determine if it's robust by using a normal distribution, checking for positive skew and excess kurtosis with a t-test and count what percent of bars have trades in them. I guess I take for granted the valuable analytical tools Multicharts has, but just the same you have to be a master statistician to tell with certainty. Look up Leptokurtic then think about why having a positive skew as such means it's robust with excess kurtosis.
You're right. Brain hiccup, ahem. Guess we're still at the 50 trade zone. It's really the 1-2% exposure you're looking for that gives a calmar apr:dd≥1 then lever up.
Kind of makes me question the data now, though, with "when the system was working." or whenever it seems to be "working."
There is no magic number of trades or trades per bars. The backtest period must include different market conditions, otherwise you make the assumption that only the price behavior of last year is relevant. Even if the strategy is on 1-minute scale and has 10 000 trades in the past 6 months, it's not validation due to not including any bearish or volatile markets.