Opposite backtests on 99.90% quality vs normal

Back to topics list To post a new topic, please log in or register
avatar
17
ionone 2016.09.09 13:59
 

Hi

 

i recently found a strategy that would lose a LOT. You know that kind of strategy that you can reverse and make profit.

 I was working on a set of non-tick data (regular backtests)

 

later i tested the algo on a tick quality backtest and the results were the exact opposite ! 

 

yes the "to inverse" algo was making money while the "inversed" algo was losing (on 99.90% quality)

 

how weird is that! so what version should i use and more importantly why it happens and how.

 

btw i used H1 in both strategies (i know MT interpolates in H1 more than in M1)

 

thanks a lot

 

JEff 

avatar
12988
WHRoeder 2016.09.09 14:12 #
 
ionone: i recently found a strategy that would lose a LOT. You know that kind of strategy that you can reverse and make profit.

False premise. If that could work, then everyone would be millionaires.

If buying looses, does selling win? No because you pay the spread in either direction, and the market moves in either direction.

avatar
17
ionone 2016.09.10 16:04 #
 
WHRoeder:

False premise. If that could work, then everyone would be millionaires.

If buying looses, does selling win? No because you pay the spread in either direction, and the market moves in either direction.

i"m aware you need twice the spread compared to what you see to make profit inverting a strat but the problem i have is different:

 

yesterday i ended up with a nice strategy that was going UP.

 

this morning i test the exact same strat, on the same machine, and the behaviour is totally different : the curve doesn't look the same! and goes DOWN

 

what the heck

 

i think i could better trust the tick 99.90% but it's on my other machine where i don't work

 

and i develop using M1 not H1 so it should be more accurate... but it"s not it's behaving strangely : what could be the explanation?.

 thanks

Jeff 

avatar
17
ionone 2016.09.17 13:25 #
 

okay i figured it out : 

when using normal backtester without tick data, MT4 performs interpolation between values and when there is value missing, MT always picks the best value : the results will always be better without tick data and you can have surprises expecially when using short trades (< 3 min)

 

that's why i had so many differences between the tick data MT4 and the regular one.

 

 

Back to topics list  

To add comments, please log in or register