Difference between tick backtesting and m1 backtesting
Difference between tick backtesting and m1 backtesting
15 Oct 2014, 00:40
So I wrote an algorithm that works beautifully while backtesting with m1 open bars, but fails with tick-data. The algorithm itself has no logic that executes every tick, so I was thinking that it shouldn't matter, but apparantly it does. Why is this, and can it be adjusted in some way?
Replies
Spotware
15 Oct 2014, 09:29
RE:
Pannekoek said:
So I wrote an algorithm that works beautifully while backtesting with m1 open bars, but fails with tick-data. The algorithm itself has no logic that executes every tick, so I was thinking that it shouldn't matter, but apparantly it does. Why is this, and can it be adjusted in some way?
Which version of cAlgo do you use? You can find it in menu Help -> About
@Spotware
Pannekoek
15 Oct 2014, 15:01
It's Pepperstone cAlgo v1.27.133. A two-year backtest with m1 data gives me a net of +507% on gold, with a spread set to random between 25-40 and a win-ratio of about 50% trading 100 Oz lots max. I have a few different settings derived from optimization but don't have access to all of them right now, but I had a ton of net winning passes.
A back-test on tick-data stops at -100% after 6 months, with a win-ratio of about 25%.
So yeah, being realistically the algo's logic is probably flawed in that it does not have a definitive edge, but it got me wondering on this discrepancy in back-testing results.
@Pannekoek
Spotware
16 Oct 2014, 10:14
There is a chance that backtesting cache became corrupted. Try to remove folder C:\Users\%UserName%\AppData\Roaming\%Broker Name% cAlgo\BacktestingCache and try your experiment again.
Tick data backtesting is more accurate than m1 backtesting. If you use m1 data for optimization, please backtest your cBot with tick data after that.
@Spotware
Pannekoek
15 Oct 2014, 00:48
Ok so I found the following topic:
/forum/calgo-support/3431
Is it possible to make an algo behave as if it were in m1 mode?
@Pannekoek