At what point can you call a bot you've set up a suucess?

[ferrypaul]

Would like to get peoples opinions on this. Over the last couple of weeks iv making subtle changes to a automation iv been working on, went live with it 5 days ago and so far its 100% 88 succesful trades. I put it to run on the aussie racing overnight last night and that threw in another 8 succesful trades. Im keeping it on low stakes at the moment but should i think about scaling it up?

[jimibt]

paul - unless you need to use the profits for living off of, i'd be tempted to continue with the smaller stakes in the HOPE that you hit a minor roadbump so that you can assess the impact if such a thing were to occur on higher stakes. I personally would think that 88 straight successes sounds good but depending on the strategy (lay based etc), I wouldn't place a lot of significance on things until you've cycled thro the markets for maybe 400-500 trades.

just my thinking as I've been on many a winning streak, only to find that seasonal bias suddenly takes effect and without an awareness of the changes that can cause, you could erode the edge that you seem to have found.

[ferrypaul]

ok cheers for the reply. Iv already ammended it with little differences for jumps and flat races which has made a huge difference

[marksmeets302]

There's some discussion about this here: https://www.betangel.com/forum/viewtopi ... =2&t=12099

[spreadbetting]

Would like to get peoples opinions on this. Over the last couple of weeks iv making subtle changes to a automation iv been working on, went live with it 5 days ago and so far its 100% 88 succesful trades. I put it to run on the aussie racing overnight last night and that threw in another 8 succesful trades. Im keeping it on low stakes at the moment but should i think about scaling it up?

That's the thing with bots even subtle changes can make a big difference but you shouldn't underestimate the fact higher stakes can be a major change not a subtle one.

So much will depend on what 'edge' your bot is trying to take advantage of and only you will have an idea if higher stakes will tip the balance. Sadly some bots won't ever return high profits simply because the edge you're exploiting is very small. Sometimes small edges get overlooked by other bots as their sights are set higher and you increasing stakes may end up triggering their strategies.

The only way I've ever found out if they work at larger stakes is by scaling up slowly and doing it for real. It's all well and good doing some statistical analysis but once the money increases it's a major change in dynamics and can almost be an entirely new bot depending on the liquidity in the market and what you're doing.

[jimibt]

So much will depend on what 'edge' your bot is trying to take advantage of and only you will have an idea if higher stakes will tip the balance. Sadly some bots won't ever return high profits simply because the edge you're exploiting is very small.

this is sadly the case in many strategies. in particular the magical balance between book% and odds gap works reasonably well (if not spectacularly often) on small stakes but seldom on higher stakes, so developing such a strategy is useful only to bag the odd £x here and there but far more likely to not pick off the offset on higher stakes.

[dannycutts]

I have a bot that is doing pretty well and I have set it to use % liability of the bank... so as the bank grows so does the stakes...

How long does it take to test whether my bot is a success... not sure if you ever know for sure But I am always trying to protect my bank

[EyePeaSea]

How long does it take to test whether my bot is a success...

Perhaps another way of looking at it is - how long to run a system and be confident that the profits aren't just the result of a 'lucky streak'. I guess you could say that I'm a glass is half-empty kinda guy.

I've been running my current automated system for 6 weeks. For me - 3 months is the psychological barrier that I want to get past.

Ian

*Edited to correct the typo - I'm never a half-full kind of person.

[Euler]

I tend to look for the worst possible run and work that way around.

I have something that works 99.9% of the time, but if the loss is 100% when it occurs there isn't much value in the process.

[Dublin_Flyer]

Agree with Euler, I prefer to be glass half empty and think of the worst case scenario, double it, square the result, then add an internet outage and see what your hypothetical position might be.

Plus 88 trades over 5 days at this time of year must be triggering in about 50% or more of the races (if it's 1 trade per race) so a bad string of results due to fog/heavy ground/bad luck could leave you in an awful position depending on your liability.

If you're in no rush to scale up, and it triggers frequently, you could keep at low stakes for a good few weeks and see how it goes over a few hundred races, over a thousand. If you were to start with your finished product now and run it at minimum stakes/liability for 8 weeks, (without modifying it whatsoever, no cheating yourself!) you would cover this weekends racing, the lull before Chelt, the Festival itself, the lull after Chelt, then the Grand National meeting. Extend it by a bit more and you could cover the start of the flat season up to the Guineas meetings.

You would have covered in the time quite a bit of the peaks and troughs of the annual calendar, and see is it still doing well, is it more profitable for big/small/NH/AW races etc, or has it just been a lucky week this time!

[ShaunWhite]

How long does it take to test whether my bot is a success...

From a non-statisticians perspective...
When the nice coloured columns on the graph fuzz into a solid black blob, and the running STDEV graph looks like a straight line when I burr my eyes a bit...I figure that's probably enough data.

Right now I feel I'm just prepping for next year as my winter testing has just started producing results in time for the daffodils to sprout.

[workpeter]

Can't answer without knowing your loss tolerance rate.I'm guessing it's very low