Most Profitable Bot ?

[jimibt]

I definitely agree that there is no *one size fits all* bot. I THOUGHT I'd found a generic set of parameters to lay and offset twds the end of the race. It all worked fine and returned a nice profit daily for almost 7 days racing. However, it then on occasion would fail to issue the offset back bet (probably due to liquidity) and as sods law would have it, that particular runner would romp home. In my scenario, this was due to the volatility that can surround the front runners in the final stages of the race. From here, I then deduced that there really needed to either be selective use of the file and/or different versions used for race/runner types.

To be honest, I still struggle with the occasional issue where offset bets don't get matched and the strategy fails - however on balance, I still keep ahead, so have factored this into my approach and don't gt too fussed at the ODD failure, as long as the strike rate remains above 60-70%. I'm now actively reforming my strategy towards pre-race as I just feel too edgy when basing my strategy exclusively in-play. I envisage that many in-play strategies are similar to what I detail above, i.e. fantastic set of results for days on end, only to be knocked over by a situation that even the tightest rule conditions can't contain, due to volatility.

Anyway, for what it's worth, I'd suggest looking at having a set of rules so tight that it only ever on average gets triggered every 2nd or 3rd race but in return offers up a 95% chance of success. That's where I've now found myself sitting in this in-play jungle..

[Dallas]

I envisage that many in-play strategies are similar to what I detail above, i.e. fantastic set of results for days on end, only to be knocked over by a situation that even the tightest rule conditions can't contain, due to volatility.

It would nt surprise me if more than half of the users who post on here saying they cant find a profitable rule have actually had one at some point but binned it early just because of a few early losses.

Over the years there has been plenty of graphs posted up by users where you can see there losses from a short term downturn for either days or even weeks but the overall trend of the graph in the long term is a constant steady rise.

[PeterLe]

Anyway, for what it's worth, I'd suggest looking at having a set of rules so tight that it only ever on average gets triggered every 2nd or 3rd race but in return offers up a 95% chance of success. That's where I've now found myself sitting in this in-play jungle..

Jim in my experience, the best bots are the simplest.
If I were to sell my best earning BOT on Ebay for £20, no doubt the buyers would want their money back stating that it was nothing special and that it was obvious!
The more complicated you make the criteria, the less eeffective over the long term. I spent a good few hours yesterday comparing data to see how I might simplify something and what the impact would be.

Check out 'Occams Razor' on Google
Regards
Peter

Edit : Also check this out by Kahneman in context wth occams razor:
https://en.wikipedia.org/wiki/Conjunction_fallacy

[jimibt]

Anyway, for what it's worth, I'd suggest looking at having a set of rules so tight that it only ever on average gets triggered every 2nd or 3rd race but in return offers up a 95% chance of success. That's where I've now found myself sitting in this in-play jungle..

Jim in my experience, the best bots are the simplest.

peter - i have to concur 100% on simplicity being the key. I strive for a concise key element that has the focus, backed up by a few defensive measures to try to eliminate noise from the projection. That said, I may occasionally over specify such defensiveness to the point that it tips the balance between risk and reward... A few weeks back, I had a pretty simple initial strategy running that was scooping a high return/hit rate but hit the buffers a few times in succession, thus I tried to shore this up by adding a few historic odds conditions to attempt to trace a graph of the anticipated action leading up to the trade placement. This is kinda where I'm at now with the hit rate on every 2nd or 3rd race. However, maybe taking the odd hit in a more managed way would be the way fwd for such a prolific high return ruleset.

[dee12]

Thank you for all your help, could anyone tell where i have gone wrong with the top bet in the image, Baku Bay, i used the same settings as the previous bet £10 back and lay with offset and greening with 18 offset tick, batches set to 10 yet the bet made a loss even though it was matched,thanks.

[marksmeets302]

If you deploy bots and haven't got a clue of what their intentions are... that's a recipe for disaster.

How about doing some manual trading first? Try to understand the mechanics of the market, then make the jump to automation.

[Mikky]

If you deploy bots and haven't got a clue of what their intentions are... that's a recipe for disaster.

spot on. I think people just expect to load up one of Dallas' bots and then hey presto they will make money!

Dallas is giving you a ton of free information by giving you a bot and explaining what it will do. You can then go through the parameters and conditions and work out what he used to achieve that and what you could tweak...

There isn't really a one-size-fits-all type answer. Just find something that does ok and then try and work out why it does well in some areas and not others.

[dee12]

Get what your saying guys, i thought i would followed Peters new youtube video and copied the settings : https://www.youtube.com/watch?v=6fkRZHWSmbc
Just cannot work out why the top selection in the image did not make money, for some reason it overlayed the selection.

[Tanden]

Hi dee12

There was a non-runner in the 15:15 which could explain the issue.

You say you used an offset of 18 ticks but the prices shown are only 9 ticks apart.

If you add another 9 ticks to the back price shown of 4.70 you get 5.60 so perhaps that was your original back price.

If it was then the correct additional offset lay stake would be (5.60 - 3.90) * £10.00 / 3.90 = £4.36

Add this to the original stake of £10.00 and you get £14.36 which is the lay stake shown.

If the non-runner was withdrawn after your back bet was matched at 5.60 then a reduction factor would have been applied which would have reduced the price.

We can see that if this was the case then the reduction was 5.60 - 4.70 = 0.9 which is a reduction of roughly 16%

Sporting Life show the non-runner priced at 5-1 = 6.00 in decimal which represents 16% of the market (1/6 = 0.16) so the figures all seem to add up

What also supoprts this is the timestamp on the bets. The back bet which would have been placed first shows as being placed just over 5 minutes after the lay.

I would assume that the timestamp on the back bet is actually the time that Betfair applied the reduction factor.

[dee12]

Thank you Tanden for your reply, the ticks must have changed by themselves as i never changed any settings once i set it up, i will take note in future and keep a check on them but it does look like what you say has happened, i would never had worked that out myself, so thanks.

[jimibt]

Get what your saying guys, i thought i would followed Peters new youtube video and copied the settings : https://www.youtube.com/watch?v=6fkRZHWSmbc
Just cannot work out why the top selection in the image did not make money, for some reason it overlayed the selection.

I think in this particular race, a selective approach (in automation) of looking at the fave BSP of 6.5 or above, with a 6-10 tick separation from the 2nd fave could bring similar results on each occasion. However, when you find the fave popping down to odds of 2-3 BSP, then you have a tougher time gaining the 40 tick offset bet to green up. This bears out the advice from many automation discussions regards picking an appropriate scenario for a given approach., In this case, I'd imagine you are EASILY 50-60% likely to green up, given that the top fave is trading above 6.5 - one for the toolkit but only to be used where you have empirical evidence that the stats are on your side. However, it is a pretty low risk strategy to double your stake and if used on the appropriate races, might well work out well in conjunction with other defensive conditions.

[ricardodeano]

If I were to sell my best earning BOT on Ebay for £20, no doubt the buyers would want their money back stating that it was nothing special and that it was obvious!

Hi Peter

I have £20, an Ebay account and won't pass judgement - let me know when it is listed

How obvious are we talking here? A one liner?!

[GunDog]

If I were to sell my best earning BOT on Ebay for £20, no doubt the buyers would want their money back stating that it was nothing special and that it was obvious!

Hi Peter

I have £20, an Ebay account and won't pass judgement - let me know when it is listed

How obvious are we talking here? A one liner?!

As soon as it is live ricardodeano, I will outbid you!!!

[jimibt]

If I were to sell my best earning BOT on Ebay for £20, no doubt the buyers would want their money back stating that it was nothing special and that it was obvious!

Hi Peter

I have £20, an Ebay account and won't pass judgement - let me know when it is listed

How obvious are we talking here? A one liner?!

As soon as it is live ricardodeano, I will outbid you!!!

i just saw it on ebay... stupidly, he put it up as a Buy it Now -sorry guys!!

[nigelk]

If you look at Peters back to lay video, he chooses Indian Stream as a likely trade.

Go the Racing Post website and the in running comments for that horse

http://www.racingpost.com/horses/horse_ ... horse_form

most of them are tracked leader, with leader, took keen hold etc, etc and also the finishing positions 2nd, 1st, 3rd etc. So here we have a horse that runs well and likes to be up with the leaders.

As it happens, I've been working with something along these lines for the past week or so (see attached), and have highlighted 3 runners with similar characteristics, good placings and likes to be upfront with the leader(s).

Don't know what the prices will be. Don't even know if I'll trade them, I'm just trying to show one way you could use spot likely candidates for a back to lay trade.