new of way optimisation / robustness test. (ORT)
- This topic has 7 replies, 2 voices, and was last updated 2 years ago by
.
-
An experiment of a new of way optimisation / robustness test. (ORT) It’s touch explain it but here it goes.
I’ve tried a few times similar approach, but here I’ve fixed an important issue.
- Implementing stop/limit orders instead of market orders with reentry.
- if using cumulative orders in step1 there can be multiple positions. Each position has it’s own fixed profittarget & stoploss.
- So EACH position = Signal has i.e. 100 stoploss & 100 profittarget tested
- Important is that there is no noise of a trailingstop or other exit criteria. It should be as clean as possible.
With step 1 sometimes the results of the optimisation rapport does not correspond to the equitycurve, but the equitycurve is leading in the choice of the parameters.
With step 2 with reentry it acts on every signal in the same direction. But instead of adding to the current position (compared to step1) now it closes the position and enters again at the next bar after the signal on stop/limit. This like above creates extra spread costs but the moment the position is reopened again it has the same profit/loss target compared to step 1. Because it uses stop/limit orders step2 can not be used for live trading because of the minimum distance requirement.
Target here is that the equity-curve is similar to step 1 and the win% in higher then 50.
In a perfect scenario the parameters would be the same for step1 or step2, but here has a choice to me made. Focus on step 1 or step2 or a compromise in both.
In any case in step 3 the trailingstop is activated & optimised and if a position is openend it ignores the other signals while in position. Entry parameters should not be optimised again I think. Here are no cumulative orders and no closing of a position when a new signal appears in the same direction, so a normal trade scenario.
The end result I hope should be that because of some kind of robustness test is applied on every signal in step 1 and/or 2 while optimizing that trading live with step3 is simply more reliable and less curve-fitted.
Attached is an example strategy. The only thing to look at is the difference in detail between step 1, 2 and 3.
Any idea’s to improve on the concept or strategy are welcome.
again
2 users thanked author for this post.
codes of an older project.
2 versions, lite & full, with test strategy (optimise parameters on fast timeframe < 30k bars on dow)
the lite version has no reentry and does not use cumulative orders
the full version with step 4 has the same results as the lite version
1 user thanked author for this post.
the round function (display lines) didn’t work properly on forex, that’s removed & small fix. dow, 30k 1 minute
3 users thanked author for this post.
Find exclusive trading pro-tools on 
