Hi All,
I seem to be getting myself in a pickle and not make any progress with things into the right direction and I know I need to resolve these issues and its as simple as looking into why, but when you tweek something, it does the opposite thing then, never ending battle to get profitable at some ideas, however its really annoying.
What are some points of advice you would give?
Just After Some Advice
-
ShaunWhite
- Posts: 10559
- Joined: Sat Sep 03, 2016 3:42 am
Are you giving your modifications long enough to give you a statistically significant indication of their impact? And also changing things too often is a sign you might just be backfitting too.
I don't want to take human experience out of the loop entirely, but gpt is good on general principals around this topic so can expand on that.
I don't want to take human experience out of the loop entirely, but gpt is good on general principals around this topic so can expand on that.
I can sort of Instantly know when its not working as Its not doing what I imagine it doing and it requires more ground work. I just need to keep sucking it up. But just anything thats helped you, logical approach, cognative, what helped etc. Chat GPT is really good tbf, its good to bounce ideas from, I have used it in the past with things
-
ShaunWhite
- Posts: 10559
- Joined: Sat Sep 03, 2016 3:42 am
It's difficult in part because version A and version B are run on different markets. Ie before the change, then after. It might be worth thinking about running the two versions side by side and comparing their relative performance rather than the actual performance of each. Simply, is B better than A on the same markets.
And sample size is a PITA.
But how are you measuring the performance? If you're betting multiple times per market then do you know what side /price/stake/time etc is profitable and which bets aren't. Market totals are subtotals, and you might need to dig down into the bets that gave rise to them.
Hopefully you find something simple like your backs being profitable but the lays aren't.....and then it's an easy fix, ditch the lays and re-stake for the new expected liability. Or better still it's that prices above or below a point where the pl changes, so you limit the bot to just those prices.
And sample size is a PITA.
But how are you measuring the performance? If you're betting multiple times per market then do you know what side /price/stake/time etc is profitable and which bets aren't. Market totals are subtotals, and you might need to dig down into the bets that gave rise to them.
Hopefully you find something simple like your backs being profitable but the lays aren't.....and then it's an easy fix, ditch the lays and re-stake for the new expected liability. Or better still it's that prices above or below a point where the pl changes, so you limit the bot to just those prices.