errrr ?Kai wrote: ↑Fri Dec 17, 2021 12:09 pmwhy not just have a look at other profitable bots to see how they do itgoat68 wrote: ↑Fri Dec 17, 2021 9:11 amif strategies can easily come and go, it then makes a mockery of backtesting. The future is truely unknown, nothing except the secret sauce the 0.01% of Betfair traders who make ALL the money off us minions, can help you... !
This bot has got today to make some money then it is off!
Trading What I see !?
This may come as a surprise but I've never back tested an automated trading strategy once. Wouldn't trust the data even if I did. Placing and execution of an order automatically changes the nature of the market that the back tested data can't account for. Particularly in some low volume markets. I'd be surprised if a back tested idea would ever work as well, or at all, once live.goat68 wrote: ↑Fri Dec 17, 2021 9:11 amif strategies can easily come and go, it then makes a mockery of backtesting. The future is truely unknown, nothing except the secret sauce the 0.01% of Betfair traders who make ALL the money off us minions, can help you... !
This bot has got today to make some money then it is off!
If I've got an idea I set it up and let it go, observe, analyse and adjust. Any immediate losses will hopefully (no guarantee) pay for itself overtime. I would split your GH data into three. Morning dogs, afternoon dogs, evening televised dogs. See if one of those is out/underperforming the others. They are different beasts.
Also the two things i look at/adjust first if things aren't working as i'd hope.
time of bet execution (measured either in time relative to start time, or volume in market)
stake of bet relative to money being matched on that runner
Hope that helps
time of bet execution (measured either in time relative to start time, or volume in market)
stake of bet relative to money being matched on that runner
Hope that helps
thanks appreciated
live bot turned off... another -£10 today already
-
ShaunWhite
- Posts: 10453
- Joined: Sat Sep 03, 2016 3:42 am
Past performance isn't a guarantee of future returns. You can say what you want about backtesting but plenty of people making proper money swear by it.goat68 wrote: ↑Fri Dec 17, 2021 9:11 amif strategies can easily come and go, it then makes a mockery of backtesting. The future is truely unknown, nothing except the secret sauce the 0.01% of Betfair traders who make ALL the money off us minions, can help you... !
This bot has got today to make some money then it is off!
-
ShaunWhite
- Posts: 10453
- Joined: Sat Sep 03, 2016 3:42 am
Exactly, its not precise and it requires interpretation rather than giving concrete numbers. What it does though is sort the wheat from the chaff.
Also its not how you find strategies it's a way of confirming what you find in analysis so if you're into sports stats then skipping backtests isn't unreasonable. And tbh not many people have access to backtesting anyway so it's a help rather than a necessity.
yeah I understand that, however backtest of the same live period showed fairly good comparison, ie.both -£100 down !ShaunWhite wrote: ↑Fri Dec 17, 2021 5:33 pmExactly, its not precise and it requires interpretation rather than giving concrete numbers. What it does though is sort the wheat from the chaff.
Also its not how you find strategies it's a way of confirming what you find in analysis so if you're into sports stats then skipping backtests isn't unreasonable. And tbh not many people have access to backtesting anyway so it's a help rather than a necessity.
The reality is you've got to try these strategies out live... so this means risking -£100 each time, the question then begs how many of those before I find one that makes money ???? 5, 10, 20,....? so that maybe £2000 down, then just hope that one makes more than £2000 back again!
-
ShaunWhite
- Posts: 10453
- Joined: Sat Sep 03, 2016 3:42 am
If it was OK for 6 months and ify for just 2 weeks I'd be very reluctant to try and force fit the 2 week sample into the 6 month sample. #1 would be to find out why the 2 weeks were different and hopefully end up with 2 variations for the 2 sets of conditions. Or #2 I'd leave what appears to be promising on a watching brief while I carried on work on another idea, then make a decision in a month or two.
It's always a good idea to have a number of things at various stages of development so you don't get rushed into making changes to your one thing.
-
ShaunWhite
- Posts: 10453
- Joined: Sat Sep 03, 2016 3:42 am
Anything decent will probably make a couple of grand in no time at all. The usual pl for a beginner is a long slow decline followed by a much more rapid growth. You have to consider these investments like buying lottery tickets, most will loose but.....goat68 wrote: ↑Fri Dec 17, 2021 5:49 pmThe reality is you've got to try these strategies out live... so this means risking -£100 each time, the question then begs how many of those before I find one that makes money ???? 5, 10, 20,....? so that maybe £2000 down, then just hope that one makes more than £2000 back again!
If it takes 20 good backtests to find 1 that's good in live then you have issues somewhere. I would expect anything that stands up in a test to have as much as a 50% chance of working. That's because I know roughly how much you'll slip when you go live so I'd want to seen a good 3-5%ish roi on a test. If a test is showing 1% then it definitely isn't ready to be moved to the next stage. Possibly worth a small spin on min stakes because very occasionally live can outperform a test but certainly not with any real expectation it'll work.
I've been surprised to see how many of your things have made it to live this year.
see your point...ShaunWhite wrote: ↑Fri Dec 17, 2021 5:53 pmIf it was OK for 6 months and ify for just 2 weeks I'd be very reluctant to try and force fit the 2 week sample into the 6 month sample. #1 would be to find out why the 2 weeks were different and hopefully end up with 2 variations for the 2 sets of conditions. Or #2 I'd leave what appears to be promising on a watching brief while I carried on work on another idea, then make a decision in a month or two.
It's always a good idea to have a number of things at various stages of development so you don't get rushed into making changes to your one thing.
good thing about backtests, I can try tweak and try things quickly, i've got one tweak result coming up..
So a couple of tweaks, based on the last 2 weeks only: Going run tweak over the 6months... will take about 6hours!ShaunWhite wrote: ↑Fri Dec 17, 2021 5:53 pmIf it was OK for 6 months and ify for just 2 weeks I'd be very reluctant to try and force fit the 2 week sample into the 6 month sample. #1 would be to find out why the 2 weeks were different and hopefully end up with 2 variations for the 2 sets of conditions. Or #2 I'd leave what appears to be promising on a watching brief while I carried on work on another idea, then make a decision in a month or two.
It's always a good idea to have a number of things at various stages of development so you don't get rushed into making changes to your one thing.
You do not have the required permissions to view the files attached to this post.
So yeah, tweak looking "ok'ish" from initial 3weeks of 6month results... not exactly "rockin-it" like Original did at the start:
Original
Week 1 TOTAL : 103.80 for 590 bets
Week 2 TOTAL : 23.66 for 555 bets
Week 3 TOTAL : 100.96 for 429 bets
Tweak
Week 1 TOTAL : 21.48 for 286 bets
Week 2 TOTAL : 16.55 for 308 bets
Week 3 TOTAL : 98.77 for 239 bets
Will see how it progresses...... Original over the 28weeks backtest had 2 "negative" weeks.
Original
Week 1 TOTAL : 103.80 for 590 bets
Week 2 TOTAL : 23.66 for 555 bets
Week 3 TOTAL : 100.96 for 429 bets
Tweak
Week 1 TOTAL : 21.48 for 286 bets
Week 2 TOTAL : 16.55 for 308 bets
Week 3 TOTAL : 98.77 for 239 bets
Will see how it progresses...... Original over the 28weeks backtest had 2 "negative" weeks.