how much trades for validate a strategy

Quote from Dhalsim:

Okay so why don't you give us an example of your unlimited knowledge and experience.

I am trading the ES on a 1min time-frame - average trade hold time 15 minutes. It is a mean reversion strategy with 50 % win rate.
Now assume it has fixed stops and target 5 points each.
Lets say profit factor is 1.6.

How many trades per year would you require as minimum to validate this strategy?? Surely you can give us some idea into your breakdown on the ES market? What kind of stats would make you confident in making and trading a strategy in this market?
What kind of draw-down would make you not want to trade this type of strategy.
What sharpe ratio would you be aiming for?

As to whether my strategies are working - i don't know i have only been live with my automated strategies for past couple of months so we shall see how it goes. All my stuff if documented here and i made no illusions to the fact i cannot trade discretionary based because of my lack of discipline. I have traded discretionary based for many years and hence all my programmed strategies are based on my observations during that period.

I am not a statistician or a math genius hence why i rely on data from my back-tests and common sense to figure out whether to take a strategy live. I have learned alot about back-testing and the pitfalls from many on this site and other sites.

From you sir i have not learned anything but BS. I read your journal and tried to program some of the stuff you writing about but it was to inconsistent and not detailed enough. My trading style is very different from yours and to be honest i don't believe in trend-lines and many other simplistic patterns. I just tried to test if the simple things you mentioned would actualy work from a automated standpoint (they did not in my case).- your trading is more gut based. You posted no live calls and always told us how many points you made after the fact. I think you are Jack Hershey's twin brother.
More...

The op only needs a t,n, or f test with at least 30 trades since he has 50 trades on 250 bars he's in the market at least 20% of the time with plenty of significance

You are talking about 250*390=97,500 annual bars meaning at 15 minutes apiece you'd need to have trades in at least 1-2% of the data or trades in 975-1950 bars so at 15 minutes 65-130 trades would show enough significance and you should only use it if calmar (apr:dd)>1.
 
So, any answer or advice on how do you validate a strategy working with EOD bars Please?

Will my approach was misguided?

Thanks.

Quote from clowner:

It seems to me i have always have to force the number of trades. I mean, if we always think about one unique asset(one stock, one future, one currency pair, etc.) not a basket, with daily EOD bars is it difficult to get more than 60 trades in a year:

If we take between 30 - 60 trades per year to get minimum statistical significance:

60 trades per year = 252 / 4 (average holding time - 4 days) = 63 trades.
30 trades per year = 252 / 7 = 36 trades.

If we take 60 per year, i will have to force my strategy for holding no more 4 or 5 bars, and i think there is a second problem with this, if we don't hold the trades for more that 4 days, we have to reduce our asset universe because the range, during 4 bars... can the asset move enough to reach our profit target?

Please, any thoughts about that? i would like to read the experts. is there a solution? Maybe baskets?, grouping assets to get more trades.

Best.
More...
 
Quote from dbphoenix:

42 (sorry, couldn't help it)
More...

Thanks dbphoenix.

@bwolinsky please, how do you estimate the number of trades in relation to the number of bars? i don't understand well your calculus, it's very interesting.
 
Quote from bwolinsky:

The op only needs a t,n, or f test with at least 30 trades since he has 50 trades on 250 bars he's in the market at least 20% of the time with plenty of significance

You are talking about 250*390=97,500 annual bars meaning at 15 minutes apiece you'd need to have trades in at least 1-2% of the data or trades in 975-1950 bars so at 15 minutes 65-130 trades would show enough significance and you should only use it if calmar (apr:dd)>1.
More...

Thanks.

If i understand this correctly then let me apply this to one of my strategies:

It is on 1min chart. Based on backtest results the avg bars in trade is 105.
So 95,500 bars annually and 105 bar avg trade then i need to trade in atleast 1%-2% of data so 975-1950 bars, at 105 this is 9 trades or 18.5 trades per year??? This can't be right can it?

Also, annual rate of return is 14.9% and max strategy draw down is 3.45%.
 
Quote from Dhalsim:

Thanks.

If i understand this correctly then let me apply this to one of my strategies:

It is on 1min chart. Based on backtest results the avg bars in trade is 105.
So 95,500 bars annually and 105 bar avg trade then i need to trade in atleast 1%-2% of data so 975-1950 bars, at 105 this is 9 trades or 18.5 trades per year??? This can't be right can it?

Also, annual rate of return is 14.9% and max strategy draw down is 3.45%.
More...

@Dhalsim, how do you obtain 95,500?. Thanks.
 
Quote from clowner:

@Dhalsim, how do you obtain 95,500?. Thanks.
More...

Blowinsky did in his example. I said i am using 1min data and therefore in a single day there 390 bars (1min bar in a single day).

390 (1min bars) * 250 trading days per year = 95500 bars of 1min bars in a single year
 
You must log in or register to reply here.
Back
Top