The System Bot Dream...

[sclarence]

I couldn't help but wonder today whether 'the Math' of any system/bot is doomed to fail when put into the context of a betting exchange? It seems to me we're just kidding ourselves to think that next 'Eureka bot' is waiting around the corner? 'Just one tweak to the in-play settings', 'maybe if I exclude sprint races', 'maybe if I loose a couple of ticks or if I wasn't as greedy'. Sure short term it could work for a day or so or even a week but sooner or later 'the math' always wins!

Maybe I'm talking rubbish, I've probably wasted some 15years (on and off) of my live trying to find an edge of some kind but no matter what happens ultimately they have all failed one by one... For instance, the other week I had some good success backing horses as they shorten up in-play, only for all my profit and more to be wiped out a day later. 'OK' I thought, 'well maybe if I try it as a lay bot instead?' But no, that didn't work either.....

I'm no mathematician but you would think if something is loosing as a backing system, surely it must therefore win as a laying system? But no... It's difficult to explain and I'm not just talking about this specific instance. Its as though the odds are always against you no matter which way around you try??

So why do we keep trying?

Steve

[ShaunWhite]

I'm no mathematician but you would think if something is loosing as a backing system, surely it must therefore win as a laying system? But no... It's difficult to explain

Sadly you can't just flip a losing system because of the spread between the back and lay prices. 'The maths' and market efficiency make breaking even pretty easy, the tricky part is beating the commission or the rake. I'm pretty new to bots so this is just my opinion, not advice.

As we're told, it's rarely if ever the case that one size fits all. Therefore it must be beneficial to reduce the number of variables to tweak, otherwise the permutations are literally endless. There's lots of variables in the event itself (course, distance, type, runners, going, starting odds relationships, day of the week, season etc) and we can't do anything to reduce them. Then there's heaps more in the bot (offsets, stop losses, this coming in, that going out etc, greening etc), but these are ones we can choose to hinder ourselves with, or not.

The frustrating part for me is running long enough tests to be meaningful; if you're too selective it takes months to gather the intel. By then things may have changed and it's stopped working (markets change or maybe someone else bot comes along and nicks your action). I'm dubious for the same reason about back testing on historical data but people do it, and as I've said, I'm no expert.

[cybernet69]

Well, I don't know about sports trading... But there is one FX trading system which makes decisions based on fundamentals, technicals and news data inputs from around the world. It's not commercially available but is used by several hedge funds. I read a news article about whereby it stated 89% accuracy on trades.

Up until 2011 everyone and their dog knew that Japan manipulated there currency at opening time if it was oversold/overbought and likewise just before closing time. I used to be part of a forum which traded the Yen pairs making rather allot of dosh riding the up/down turn. There was no risk as you knew the Japan central bank would just send the currency one way or the other for around 30 minutes.

However, their central bank doesn't seem to get involved that much anymore.

[Euler]

If you look at the market from a top level perspective it's pretty efficient. But when you dig below the surface it becomes apparent it is not, or it is only efficient because people are correcting those inefficiencies.

In racing the market is efficient, but if you cut it by a number of metrics you can find a number of oddities and the markets in general seem biased.

I don't think it's enough for outright backing or laying often, but they do exist and trading an inefficiency is much better than betting it.

[spreadbetting]

Euler's comments pretty much sum it up, whilst the market may appear 'efficient' that's usually because the statistics are taken from fixed points i.e. start of the race and compared against the end of the race.

If the markets were truely efficient we'd have straight line graphs between those points but we all know the graphs bounce all over because of 'inefficiences' within the market. To bot successfully you need to think why these 'inefficiences' occur and then how to exploit them. You can't simply lay every runner at 1.21 because the betfair's hstorical data show it returns a 5% profit and expect to retire.

[Euler]

The earth is flat or round depending on where you look from, but it's actually neither.

[Naffman]

The earth is flat or round depending on where you look from, but it's actually neither.

Don't start confusing us Peter

[foxwood]

The edge I just discovered would have yielded 11% nett profit after 5% commission over the last 4 years at BF SP.

An easy £9k+ pa at £100 level stakes and lots of similar opportunities to be found.

How ?

Simply back any runner whose name includes the string "win"

[ShaunWhite]

If the markets were truely efficient we'd have straight line graphs between those points

Surely an efficient IP market is one where the prices at any given moment reflect the actual odds of an outcome occuring. (or have I totally misunderstood the word 'efficient', happy to be corrected). It could never be a straight line unless every horse ran at one pace like a machine.

There are far more than 2 points and each are joined by a step, theres' one for each moment observed and assessed by the hive mind, and not just the one at post time and one at the end of the race.

I see it as the reason why, over the entire duration of the average event, if you look for how many contenders get below price n.nn it's almost always exactly n.nn.

I don't think I'm disagreeing with anyone, just seeing the market in the moment(s), rather than as something in transition from a known state (the 'off') to an unknown state (the finish). I get the feeling though that I'm completely missing a vital point somewhere, unless it's to do with deviation from a line.... but it has an unknown end point ??

if you cut it by a number of metrics you can find a number of oddities

I will say Mr Euler, that you rarely write a sentance that doesn't require a whole coffee pot's worth of pondering. It's like talking to Confucius

I 'get' why your answers are like that (no fixed way to do things/trade secrets/to give us a net instead of a fish etc), that's all fine and I do like a puzzle but; I am left wondering how to measure the market in new ways in order to reveal the bias and how it occurs. I'm missing a clue that will get me off and running, but I'm not expecting one.

That's all far to much to read, but the main question I have for my homework is 'how can I spot an inefficiency if I don't know where the market is going?'...it's time to put the coffee on.

[EyePeaSea]

Hi Steve,

That's a really interesting post - thank you. It's something that I've been thinking for the last 6 years or so (since I started dabbling).

I've always found the same; start a system and it seems to work for a while then just when, or soon after, I start betting real money the P&L either crashes in a downward spiral (think WW I bi-plane with the Red Baron behind it) or floats surely but consistently downwards until I give up.

Peter posted a great link to this sort of behaviour; one moral that I took from that - don't panic about fluctuations in the short term; observe over a longer term.

I think you're right; the maths will always win. But that doesn't mean that a system that's based on (simple?) logic (rather than just human judgement/skill/intuition) can't work.

I couldn't help but wonder today whether 'the Math' of any s'Just one tweak to the in-play settings', 'maybe if I exclude sprint races', 'maybe if I loose a couple of ticks or if I wasn't as greedy'.

With previous systems, I've tinkered with this sort of thing (and its always been a mistake in my experience). My latest system - which is still in the green after 34 days (not that I'm counting) has thrown up some interesting 'facts':
I've been in the green on races at Lingfield 11 times compared to one loss. But Ludlow, I've got a 100% loss rate (2 races, lost both). Should I tweak things so I never bet at Ludlow? No - the sample amount is just too small. Likewise, should I increase the bank I use at Lingfield, I think no.
I can keep trying to improve the system, but looking at trends can be misleading if you don't give it long enough.

Regards

Ian

[sclarence]

There's no question BetAngel is a fantastic tool but as with any tools in the wrong hands......

I run my own bathroom design & fitting company, I have been fitting bathrooms and generally been in the building industry on and off since I was 10 years old (thanks Dad)

We all have our own skills and abilities that come naturally to us, I mean not to be modest, but I would consider myself the bathroom fitting equivalent of Peter. Now I run my own company I thought well I'll get a few experienced fitters to work for me so I don't have to anymore, now I'm pushing 37 .. I've gone through 3 fitters in 2 years and still can't seem to get them to install a bathroom like I can, no matter how much teaching I try and no matter how willing they are to learn. In life I think you either 'got it' or you 'don't got it'. And trying to 'learn it' would be like me thinking I could paint like Rembrandt just because I went to Art school.

So what hope is there? In trading I mean...?? We can all try to learn and understand, experiment and practice, but try as we might, we will not become Peter
So I say well sod it, I love art, I love painting, I know I'm not going to become Rembrandt but I'll just paint anyway!



Oh my brain hurts... this is too much this early in the morning

[HRacing]

Rembrandt had his edge, Neil Buchanan had his edge, we all dont need to strive towards the same target.

[spreadbetting]

Surely an efficient IP market is one where the prices at any given moment reflect the actual odds of an outcome occuring. (or have I totally misunderstood the word 'efficient', happy to be corrected). It could never be a straight line unless every horse ran at one pace like a machine.

There are far more than 2 points and each are joined by a step, theres' one for each moment observed and assessed by the hive mind, and not just the one at post time and one at the end of the race.

I see it as the reason why, over the entire duration of the average event, if you look for how many contenders get below price n.nn it's almost always exactly n.nn.

I don't think I'm disagreeing with anyone, just seeing the market in the moment(s), rather than as something in transition from a known state (the 'off') to an unknown state (the finish). I get the feeling though that I'm completely missing a vital point somewhere, unless it's to do with deviation from a line.... but it has an unknown end point ??

I don't think you misunderstood I just didn't phrase things too well to try and get my point across. There are obviously more than two price points for each runner but if we pick any of those price points we'll find an 'efficient' market for each of those points where the price will reflect the probabilty if we look simply at the bare 'efficient' stats.

It's no surprise the Betfair stats appear to show as efficient as we've had large numbers of markets all bet on by 'sharp minds' so you'd expect the averages to be close to true probabilites. Plus any long term inefficiences get picked up and exploited or closed.

If we consider foxwood's system, has he found the holy grail or will the system eventually reflect the starting odds once more data is collected.

The edge I just discovered would have yielded 11% nett profit after 5% commission over the last 4 years at BF SP.

An easy £9k+ pa at £100 level stakes and lots of similar opportunities to be found.

How ?

Simply back any runner whose name includes the string "win"

[Euler]

Surely an efficient IP market is one where the prices at any given moment reflect the actual odds of an outcome occuring. (or have I totally misunderstood the word 'efficient', happy to be corrected). It could never be a straight line unless every horse ran at one pace like a machine.

At any point in a sports event the market probably is efficient (apart from where it isn't, but that's a whole different discussion). But as a trader you are effectively selling volatility, what happens between point A & B. Nothing to do with whether the price is correct at that exact moment or not.

Because the market has a spread, at the tiniest level that's where the profit comes from. On a broader lever it's how that spread moves and the high and low point traded.

[Euler]

Probably the hardest thing to do is something that just generally works well in generally most markets. You have to trade like you would play a round of golf. Look at the hole, decide what you are going to do and they pick a club that best suits that purpose. If you make a mistake, hopefully, you have rescue club in the bag.