I think  it's the emotional side  Lewisham that would be very hard to model.

You must have watched a game of football or even tennis , where discipline goes out of the window and all hell breaks loose in the scoring  - that doesn't really happen in computerised chess.

Maybe coach but I was more thinking of betting before events, and presumably you could eventually build a bot that takes advantage of human emotional weakness in any case

for example if facial recognition software (which admittedly is one of the machines weak points) was good enough it could recognise negative and positive microexpressions and bet accordingly, and also identify when markets are likely to overreact in play.

yes pxb but a well-programmed bot will be able to factor that in. Human behaviour is inherently predictable, even though it often seems completely random...

I just think there are too many variables to model . But i like the idea of computers picking up on expressions and body language etc  -  would be very interesting  .

well just factor in recent form then.

Thats the crux of the matter, Just Checking. Liqudity on Horse Racing (and presumably other sports) has declined markedly since the advent of the Sportsbook. It is killing the exchange. Like a bird with two chicks that only feeds one of them. In a year or two day to day Horse Racing betting will more or less dead on here, sadly.

and whoever uses chess, which is not a complex multidimensional dynamic game but simply a slow static limited and boring calculation exercise that results in identical positions again and again after hours of play, ends in a draw most of the times and of course will be won by the biggest computer,

Oh dear. Your breathtaking ignorance was really exposed by your last comment.

What has not been mentioned here and is crucial in my experience and is the difficulty..

PJ mentioned Mr Lumpy(bot)or Asian money possibly also, on the cricket markets........

So the problem as i see it, is often the bot is dictating/controlling the market in such a way that it alone has bought all the perceived value to the point where everyone else is drowning or and panic selling at a false price which the bot has manufactured,leaving everyone else sweating on a opposite result which may becoming to appear more lightly but which the bot wont release the pressure on the price and if/and when it does, it squashes the life out of the other side rapidly to a point again of no value.This leaves joe bloggs in a place where he might choose not to operate but if he does he knows even if he is seemingly going to be shrewd and correct he may also get cut off at the knees often when he does not get it spot on as entry and exit points will be distorted by weight of money and que jumping.......

I know some will get this and others will wonder WTF...

sORRY if i have not been clear.........

Agreed.  You just have to be careful you don't commit too great a proportion of your bank when the odds are wrong.  There's no point getting value if it bankrupts you.

Very interesting thread.

One point that I think is often overlooked is that bots are available to everyone: you can get hold of software very easily (a few apps reviewed recently in another thread), you could pay a developer to build something for you or you could try and write your own. So for very little cost, anyone with excellent judgement and knowledge could add a bot to their ability to trade in the market.

Another thing that I think is important to point out is the difference between the judgement/expertise part (i.e. pricing up a market) and the execution part (i.e. managing to get matched at target price or better). On the execution part bots have a massive advantage, as speed is important, but again bots and automated tools are available to everyone at a pretty low cost of entry. The judgement part is more difficult. AI, or machine learning can go a long way to look for patterns in a massive complex data set, and has the very valuable benefit of not allowing emotions to cloud decision making.

Market manipulation of various kinds have also been mentioned by people in this thread. One thing that I like about sports markets that is not true about financial markets is that the event itself actually happens, and so if you are sure you have a good price you can hold out until the result is in and if you're correct, over time you will make a profit. However in financial markets there is the old adage that markets can remain irrational longer than you can remain solvent, so you don't just have to be right you also have to predict the market 'correction'. Bots unfortunately do create an opportunity for market manipulation, but as suggested earlier this will be to the detriment of traders while it should present an opportunity for the gamblers who want to hold on to their bet.

That's a problem with bookmakers slavishly following the slightest move on Befair.

If a £5 bet can make them cut their prices and then when the price on Betfair drifts out not to push their prices out who's fault is that? The spineless Bookies.

If Bookies had an opinion punters could get on.

Lewisham, I'm pretty sure the bookmakers have automated the price reductions the moment something trades below their prices on here - and, of course, they really hate laying arbs.

Apart from the big meetings the early racing markets usually have tumbleweeds blowing through them - I would suspect the Dundalk market is down to the lack of action today...

screaming from beneaththewaves 23 Dec 16 18:31 Joined: 30 Jan 05 | Topic/replies: 9,391 | Blogger: screaming from beneaththewaves's blog
I find it hard to believe that bookmakers are changing their prices because of a £2 bot on an unliquid market on betfair.

Believe it.

well I don't and the firmness of your answer isn't going to make me believe it anymore.

Do you really think that multi-billion pound companies are going to be manipulated by someone fiddling around with a 2 quid bet on betfair? Are you really that delusional? I fully agree that Betfair does influence bookmaker prices but not when you're talking about those amounts.

Have posted this or similar before

For me a small punter(yes small amounts)
on tennis/cricket etc when you submit a bet at ex 1.34 to back within the delay
the market has moved to 1.32 if i put in a bet to lay the market drifts to 1.36
so i don't get matched either way.Which although i find this hard to believe is
Do we all actually have the same market price at the same time?because surely i'm not the
only person wanting a bet in the market or are the bots so quick it changes that fast
But i can not always be wrong back/lay but due to market moves i am.

So bots are not subject to the 5sec or what ever delay on the markets then
which puts you then at a massive disadvantage.Or is a bot shadowing every person operating
on said market and just keeps jumping in front?.

I have made a good living botting betfair for years so I feel well placed shed a little light on this for the normal punter.

First - there is no "bots vs humans." Bots are programmed by humans so it's humans vs humans. There are a few smart bot programmers who make a lot of money and a pyramid of less effective players beneath them who scrape by or make a loss. It is far from easy to make good returns from bots, it takes a lot of work and a very particular type of mindset.

On modelling events: This appears to be poorly understood in the posts above. When my robots are analyzing an event (a horse race for example) the only information they have available to them is the betting activity on the exchange, they cannot reasonably surmise from that what the likely outcome of the event is. What they can do is identify the patterns of betting behavior that are typically present when the prices diverge from true value, make a reasoned estimate of true value and bet when it is likely that there is some margin available. To give a really simple example: Say 20 seconds into a race a horse suddenly goes from trading at 100 to 1.5 - well that might be because a load of other horses have just fallen or it might be that the horse has been over bet in a panic response to a more minor incident. If the robot decides that such a big move is likely to be an over reaction it will lay that horse somewhere between 1.5 and whatever it thinks is a more likely true price. At the same time it will be looking to see if any other horses have suddenly become value to back. The effect of this for "normal" punters is one of two things: if the horse has been over bet then it is pushed back towards true value (good for backers in this example), and if it hasn't then you can take my money - either way this is good for anyone who wants to back the horse and bad for anyone who wants to lay it. Of course this is an over simplification of how likely value is determined but hopefully it gives the idea. On some other sports a bit more information on the actual event is available, but the basic principal remains that what you analyse is primarily the betting and NOT the event.

On AI: AI isn't a box of magic that just does things for you, it's a set of complex and difficult systems. Personally I don't use AI and I suspect very few people do but I do use Machine Learning (which is often confused with AI). Machine learning is the use of learning algorithms to analyse data. My robots don't use ML directly but they do run models whose parameters are informed by research aided by machine learning techniques. As I understand it a true AI robot would be placing bets, independently analyzing their results and then adjusting its own algorithm without intervention. It is this feedback loop that makes it 'Intelligent.' This is complex and computationally expensive so I doubt many find it worth the bother and I would imagine that machine learning informed models is about as far as it goes. There is nothing to stop non automated punters from using exactly the same type of techniques to inform their own betting. If I had to bet 'by hand' that's what I'd do.

As to whether humans can beat bots: It depends on the human and it depends on the bot. In an overall sense humans can't beat my bots because they simply won't bet without a very strong indication that they will be getting value. At the level of the individual transaction however there will be thousands of occasions every day where I'm giving away money to the person who (in the example above) can see that all the other horses HAVE fallen. Maybe somewhere out there is a person who's very good at taking all my bad bets and never takes any of my good ones. If so He's very very rich and good on him. Quite apart from bots v humans there is bots v bots. Smart automated players are able to make guesses at the strategies used by less experienced botters, anticipate where they are likely to give away value and bet against them using superior models. To know what type of players come out on top overall we'd need to see everyone's overall ranking and that's not about to happen. Frankly though, the number of substantial winners on betfair is so vanishingly small that I can hardly see why it would matter.

On Queue Jumper bots: Let's be clear - I don't do this, but I suspect the reason some people do is simply that they want to make up the market close to 100% and they don't have any information with which to do it except the prices that others are willing to put up. If you ask me it's a pretty lame way of operating, but on the other hand I don't really see why people get so upset about it. It might be a bit annoying but were you ever going to get matched anyway? Ultimately the market has to build to close to 100% before any significant volume will be matched and it's all part of the process.

I hope that clarifies a few things, what bots do and (broadly) how they work isn't a secret so if anyone has questions I'll answer.

Regarding AI - I mean "computationally expensive" - as in it takes a lot of computing power. Most of the time you're trying to write the fastest, simplest, cleanest code you can so that it executes quickly. You wouldn't want to go gumming it up with a lot of unnecessary AI stuff unless you really saw huge advantages from it.

On the wider question - I get that it's frustrating for some people to find that they're facing such opposition but what can I say - that's the game as Betfair laid it out. Betfair actively supports bots and does everything to make it easy for us - there is specialist support, they publish helpful documentation on how to interact with their systems and even host a dedicated forum. They're making me welcome - so why shouldn't I play?

Recently I haven't posted much on these forums but in the past I've made no secret of my activities. As to whether Betfair should advertise differently - that's a question for them.

Maybe you think people wouldn't play if they knew what they were taking on - I'm not so sure. Everybody knows that you can't beat the high street bookies but in they go.... I think there are people out there who rather like the idea of taking money off the likes of me and I wish them good luck with it. On the other hand there are numerous recreational punters who just want to get a bet on - I'm glad to oblige. Of course there will be people who think the reason they're losing their money is bots but that's just denial. Gamblers have always had their excuses.

Truth be told - most of the time you'd have no idea what bots are operating or whether you're betting against them. Sure there are a few small time simpleton efforts trying to be first in the queue at 9am but the real players are predominantly getting stuck in to the in-running chaos where ordered thinking and cool heads rule.

Finally - I wouldn't want you to think that people like me are somehow just exploiting technical loopholes or doing trivial nerdy tricks - we're not. I take on gambles - big, meaty gambles. If my robots think a horse is good value they back the nuts off it. They're doing it rationally and based on a lot of research but push come to shove they get real money - MY money - and wager it like there's no tomorrow. I use some good tools to do it but in the final analysis I'm a gambler same as you.

I'll tell you exactly what I do and you can decide on the description:
Every time I get new data (the market refreshes) I estimate the value of every horse (if it's a horse race). If something's value to back I back it, if it's value to lay I lay it, if nothing's showing value I do nothing. If I think one horse is value to lay all the way through the race then I keep on laying it and rack up big liabilities. If bets don't get taken I may or may not cancel them depending on how the event plays out. I make hundreds of bets in almost every race. Sometimes all those bets mean that I'm green on everything but usually not and it's certainly not a goal to be green all over. The event finishes and I win or lose same as anyone else, sometimes I win or lose a lot, sometimes not much. Over all it averages out in my favor.

You can label that whatever you want but personally I don't really believe in 'traders' - they're gamblers in denial.

Do you know anything about the horses you are betting on?

No. Not one end from the other.

I used to find it scary but I'm accustomed to it now. Yes I do have liability limits though, not because I'm scared of betting liabilities so much as because I'm worried about some sort of programming error pouring all my money down the drain. Ultimately I have a float in my account as the absolute backstop.

I understand what you mean about the more you're able to get on the more likely it is you're wrong - but I would think about it like this:
Say I lay at 2 and the price moves to 1.5 - that means my bet at 2 is unlikely to pay off, but it doesn't have any baring on either on whether a lay 1.5 would be good value or whether 2 was good value at the time the bet was placed. If I'm being rational I should forget about my bet at 2 and calculate the likely value of a bet at 1.5 independently. I know that overall my reckonings of value are sound and the fact that I may have got the bad end of some variance a moment ago is neither here nor there. If 1.5 looks like value to lay then I lay again. On the other hand if it's value to back I switch sides and back it.