In sample/Out of sample of live strategies inside!
- This topic has 23 replies, 4 voices, and was last updated 6 years ago by
.
-
I cherry-pick a fantastic piece of data (I choose a wonderful perfect trade setup in the chart) and i create my strategy based on that single trade.
Aha I think I’m right in saying this is the first time you have said that?
Interesting method!
This might be the first time im saying this!
When i go into “creation mode” i usually think about 1 way to get money out of the markets, be it a trend following thought, or a mean reverting plan. (for me its trend following 9/10 times, “sadly” hehe would love more diversification for my systems..)
So i think about “how can i enter the start of a good trend as early as possible, and how can i exit as close to the “top of the trend” as possible? without taking too many false signals/bad trades.
So i dive into the chart at any random point of time, and i search for a trade i would consider a great trade, a very fine looking “swing up” or whatever. Then i just create something out of my theory to match that 1 spesific trade, and as i said, if it dosnt look like shit just before/after that trade is done, i check 100% and if its not a big fat turd, i go into part 2 and optimize!
And yes ofc i do walk forward leaving all/most of my variables up for change, and im looking for good walk forward % (between 60-100%) and i look at the variables chosen for each period, obviously wanting them to be as close to my optimized values as possible.
Edit: “But maybe you have curve fitted for a trend and then price level reaches / consolidates in a plateau or base and so your System has no place in that structure?” If i have optimized a system for 10 years (as i had in that example) and it looks like shit the next 10 years (out of sample) then i have curvefitted the strategy too hard on the first 10 years. This is obvious. MAYBE i could go back to other values, MAYBE i could fit the model for ALL the data, MAYBE im drowning a baby hehe… BUT if it looks good on 100% then im gonna be confident that its more robust. Confidence is key. If i tkae a curvefitted strategy and start to well.. curvefit it some more, im not gonna be as confident and im probably gonna turn that strategy off during the drawdowns…
Maybe this is just me hehe
1 user thanked author for this post.
@GraHal I think that if you optimize a strategy and it turns out to be curvefitted then your strategy might be shit but if u optimize and the OOS is good, then that means ur strategy might be something robust.
I like to stress-test my algos. If its robust it should work (profitable, not beautiful) on other timeframes and other similar markets. Ive included the 1h DAX strategy “Poirot” on other markets and timeframes. A couple of hard tests are Italy and Spain index. They have seen some SHIT lol. If its profitable and OK-looking in other markets, on other timeframes, i feel like the foundation of the system is not curvefitted. (i wouldnt trade it on the other markets, unless i fitted it for that data and it would as good as the original.. DAX index is worth like 12K and Wall st is worth like 25K (the index i mean), so a stop loss 60 pips might not be a good stop loss in every market.. BUT if its ish-similar and profitable, then my foundation would not be shitty (i think and hope)
As i said, all the EQ curves in the photo is the EXACT SAME algo optimized for 1h dax.
1 user thanked author for this post.
Nice job Jebus!
Wasn’t too long ago you joined our beautiful forum, and all help i’ve given you, seems to have given some sort of reward!
Nice to see son, keep it up!
Nice joke lol.
Find exclusive trading pro-tools on 
