Time To Build Strategies
9 replies
Trader2098
9 years ago #113471
It seems like it is taking a very long time after clicking build new strategies – is this normal? How long should it take to build new strategies?
Thank you 🙂
eastpeace
9 years ago #129346
If your setting is proper, in random mode, about half an hour (cpu:i7,mem:8G) ,you can get something. Just in my experience.Â
tnickel
9 years ago #129348
Hi Trader 2098,
its depends on your settings.
Â
I the SQ are 4 demosettings available.
With this settings strategies are build very fast.
Â
thomas
https://monitortool.jimdofree.com/
Trader2098
9 years ago #129349
Hi Thomas,
Â
Very fast meaning hours or days? I tried a couple of those demo settings and it has been running for hours and still not finished. Maybe I need to refine parameters?
Â
Thank you 🙂
Mark Fric
9 years ago #129359
if you mean that it just runs without finish this is by design. Random generation (and also genetic optimization with some settings) can run until you stop it.
Â
You can stop it manually when you have enough promising strategies generated in the databank.
Mark
StrategyQuant architect
Trader2098
9 years ago #129362
great, thanks!
Trader2098
9 years ago #129394
… is there any advantage to letting it run longer? Will it produce more/better strategies or does it start overwriting strategies once it reaches threshold specified in settings?
Â
Thank you 🙂
tnickel
9 years ago #129397
Hi Trader2098,
-there is a build in fitnessfunktion who say what strategy is good or not. You can change every parameter of this funktion.
-you can set max amount of strategies will be stored in the database (for example 1000)
Â
If the generator find more than 1000 strategies, the strategies will be sort with the fitnessfunktion and the bad strategies will be thrown out.
Â
Most Time I store 5000 Strategies in the database and run the system 1 ore two days.
It depends what you want.
With the demosettings you get strategies faster.
I think after 1-2 hours you will have 100-200 strategies ore more.
Â
You can run the system longer and you will get better strategies. No good strategy will be lost in this case.
Â
thomas
https://monitortool.jimdofree.com/
Matusiak Adrian
9 years ago #129403
I use only random generation.Â
Â
On my expirience I see that using Genetic is good, but it produce too many almost the same strategies.Â
geektrader
9 years ago #129408
I confirm this, that´s my finding as well. Hence I use random generation + just 2 generations. Population-size is set to 1200 though and Decimation coefficient to 5 to first create a really huge pool of random strategies, then try to put them into evolution in 2 generations – but most often the extra genetic evolution on the first random population does bring nothing worthwhile new indeed. But won´t hurt neither as the strategies from the random initial population still will go to your database if they fulfill your ranking criterias.
Viewing 9 replies - 1 through 9 (of 9 total)