Hi guys,
Just wondering if you get to a stage where you think you may have a bot that's starting to turn a profit, what's the best way to test it's scalability? As in is there a better method than just plowing cash in?
Thanks
W
Scaling a bot
-
spreadbetting
- Posts: 3140
- Joined: Sun Jan 31, 2010 8:06 pm
Without knowing how your bot plays it's hard to say. The more money you stick into the markets the more effect it'll have on the prices and moves within the market. Even relatively small changes may have the 'butterfly effect' and switch a winning system into a losing one if bets start going unmatched.
If you're straight backing/laying and taking prices it may not have as much effect as offering odds and trading. The only real way to find out is to scale up slowly over a period of time and don't overeact to short term trends that may crop up.
Sometimes you just have to accept there's a market 'cap' on all bots just because there isn't endless liquidity in the market. Plenty of my bots have capped out for various reasons, you just have to leave them to it and search for other ideas once they're max'd out.
If you're straight backing/laying and taking prices it may not have as much effect as offering odds and trading. The only real way to find out is to scale up slowly over a period of time and don't overeact to short term trends that may crop up.
Sometimes you just have to accept there's a market 'cap' on all bots just because there isn't endless liquidity in the market. Plenty of my bots have capped out for various reasons, you just have to leave them to it and search for other ideas once they're max'd out.
I agree with what spreadbetting said.
Scale up slowly. Keep a detailed log of what stakes you used on what days, and what the results for each race you traded were. Over a reasonable period of time (rolling window), use excel to calculate statistics like average P&L, s.d. of P&L, max win, max loss, edge as a measure of % profit against turnover, edge as a measure of ratio of win/loss, etc.., etc.. and just see how these statistics are being affected as you increase the stake.
You need to do this over a reasonable rolling window of time though. As Euler has suggested in the past, your statistics will eventually reach a fairly steady state. At that point you can change your stakes and see how it affects the stats over the next window.
Scale up slowly. Keep a detailed log of what stakes you used on what days, and what the results for each race you traded were. Over a reasonable period of time (rolling window), use excel to calculate statistics like average P&L, s.d. of P&L, max win, max loss, edge as a measure of % profit against turnover, edge as a measure of ratio of win/loss, etc.., etc.. and just see how these statistics are being affected as you increase the stake.
You need to do this over a reasonable rolling window of time though. As Euler has suggested in the past, your statistics will eventually reach a fairly steady state. At that point you can change your stakes and see how it affects the stats over the next window.
There is a maximum profit per race per strategy.
For myself I can easily choose the wrong bot that presses too hard and produce a loss. So part of what I am always doing is matching my bots to the market - or what I guess the characteristics of the market will be. Some of that is also automated within the bot based on market characteristics. But mostly, for me, it's about not making up too much of the market.
For myself I can easily choose the wrong bot that presses too hard and produce a loss. So part of what I am always doing is matching my bots to the market - or what I guess the characteristics of the market will be. Some of that is also automated within the bot based on market characteristics. But mostly, for me, it's about not making up too much of the market.
Absolutely great advice from spreadbetting, gauzy and titian - given me lots to consider and I can see the limitations in me asking the question obviously not explaining the full system out but none-the-less thanks for all the help and ideas going forward.
W
W
-
Crazyskier
- Posts: 1301
- Joined: Sat Feb 06, 2016 6:36 pm
I'm finding this too. I 'bot' the early markets to make a market when there are often large gaps, and in the early days I kept tripping over 1000 'transactions' an hour and had over £200 removed by Betfair at one stage. I've since reduced the triggers and extended the kill-or-fill times, however this can leave me exposed sometimes, with the stop losses I used to use costing me dearly.spreadbetting wrote:Without knowing how your bot plays it's hard to say. The more money you stick into the markets the more effect it'll have on the prices and moves within the market. Even relatively small changes may have the 'butterfly effect' and switch a winning system into a losing one if bets start going unmatched.
If you're straight backing/laying and taking prices it may not have as much effect as offering odds and trading. The only real way to find out is to scale up slowly over a period of time and don't overeact to short term trends that may crop up.
Sometimes you just have to accept there's a market 'cap' on all bots just because there isn't endless liquidity in the market. Plenty of my bots have capped out for various reasons, you just have to leave them to it and search for other ideas once they're max'd out.
Since I've kept refining, I have very few substantial losses versus the average gain, and still get out before the last race finishes and the prices go bananas. I'm sure I could reduce the size of my infrequent but sometimes large losses further if I had more time to look at the patterns of repeated movement in a price that has shortened in the hours pre-off, then just after my bots greens / hedges up, then goes back out again! It's such a dark art this botting business...