How big a margin

[Lardonio]

Hi, I have a lot of historical data for all football betting markets, but the market I think is most fun to bet in is the correct scores market. But I'm wondering how big a margin I should calculate in order to be successful. I'm thinking - if my historical data says the odds of a 1-0 happening is 10% then of course the true odds is 10. If I can get 11 then that is a 10 percent margin, which is enough to keep most gamblers happy. But if I look at it another way it does not seem so tempting. If the true odds is 10, then again, the true chance of it happening is 10%. If the true odds is 11, then the true chance of it happening is (cirka) 9.09% There's only a (cirka) 0.91% difference in the two chances, and of course with any betting prediction data, historical data, etc. the error margin is probably bigger than 1%. This way of looking at it is more depressing the higher the odds became, for example the difference between a 40-odds and a 50-odds is, of course 10 in odds which seems a lot, but only 0.5% in chance, which is "nothing". The way the correct score markets is priced, even in betfair, I feel like I'm almost doomed using any statistical method, and I guess this applies to any betting market. I'd like your feedback on this. Is there really any point in developing betting methods based on statistics alone?

[Gingertipster]

Re: How big a margin The way I do my betting on horses is to make a 100% book, but I am not so good to be 100% accurate. I don't back anything that is (imo) only just value. It depends on how easy I think a race is to work out but normally: e.g. something I believe is a 20% chance (true 4/1) I don't back at 9/2 18.18%, but would at 5/1 16.67% (A 3.33% difference). A true 9/1 shot 10% I'd want 11/1 8.33% (just 1.67% difference). At 40/1 2.43% I'd only want 66/1 1.5% (just 0.94% difference). Sounds as if I am just missing a price out but it's a little bit more complicated. I believe the bigger the price the easier it is to evaluate. Also, bookmakers mark ups are less the bigger the price. The bookie might put a 2% or 3% (if it's unexposed) mark up on a 4/1 shot; but would not do so at 40/1, 0.5% or 1% most likely in a competitive market. If they added the same mark up for everything there'd be no 100/1 shots and bigger. So in turn, our margin for error can be less the bigger the price. What price you are willing to take should depend on how good you are at spotting value. Hope that helps Lardonio Mark

[Gingertipster]

Re: How big a margin The higher the odds the less difference in percentages, that's true. But also, the higher the odds the bigger difference in return (between prices). Every £ bet at 4/1 returns £5, 9/2 returns £5.50. For a difference of 1.82% in price, nets £0.50 difference. Yet 40/1 returns £41 and 50/1 £51. For just 0.44% difference in price nets a difference of £10. Therefore in practice the higher the odds, the more you get back. The shorter the price the more chance of winning the bet. Yet the bigger the price the more value the bet is. ie For a 2% difference in price at bigger odds you'd get back a lot more for less money. Therefore I believe the shorter the price, the more money should be staked. Yet the bigger the price the more money should be won. So I have a staking plan that does this. Simply put on the same points as the percentage price you believe the horse to have. Something I believe is true 20% 4/1 shot I stake 20 points at say 5/1. Winning 100 points. For 10% 9/1 stake 10 points at say 11/1 winning 110 points. A 40% 6/4, 40 points at 7/4 wins 70 points. A 4% 25/1, 4 points at 40/1 wins 160 points. If that's too complicated just put the same percentage as price available at the bookies or exchange. Any 3/1 shot put 25 points on, 4/1 20% 20 points etc. Of course a point can mean any amount you want it to be 20p or £200, whichever suits.

[dannkez2407]

Re: How big a margin hi GT, i do a lot round football value betting but have always struggled with horse racing, can you explain how you estimate a horse would have in your opinion a 20% chance of winning? thanks

[Gingertipster]

Re: How big a margin Dannkez, That is too long a subject to describe here, but have already done so in the At the races forum section. Look for a thread: How to produce a 100% book. Might be easiest to go through my profile to find it. Think the last time it was added to was in January, so it's some way back.

[dannkez2407]

Re: How big a margin cheers mate, will do

[Lardonio]

Re: How big a margin

The higher the odds the less difference in percentages, that's true. But also, the higher the odds the bigger difference in return (between prices). Every £ bet at 4/1 returns £5, 9/2 returns £5.50. For a difference of 1.82% in price, nets £0.50 difference. Yet 40/1 returns £41 and 50/1 £51. For just 0.44% difference in price nets a difference of £10. Therefore in practice the higher the odds, the more you get back.

Well, that's not true, because you will only get back more if you are able to find more value in those bets than you would find in bets with smaller odds. And even if that is the case, playing 40 and 50-oddsers has so many extra challenges for the gambler and it does require quite a powerful mental stamina to sit around for 40-60 bets waiting for a winner while unsure whether you have a 0.10% edge or a 0.10% (or worse) disadvantage. In practice that's impossible to know until you've bet thousands of bets. But my original point was more the question: Is it possible to make money on statistical football betting at all?

[dannkez2407]

Re: How big a margin

Is it possible to make money on statistical football betting at all?

definitely, there is more than enough systems out there all over the net to quantify that

[Gingertipster]

Re: How big a margin

Well, that's not true, because you will only get back more if you are able to find more value in those bets than you would find in bets with smaller odds. And even if that is the case, playing 40 and 50-oddsers has so many extra challenges for the gambler and it does require quite a powerful mental stamina to sit around for 40-60 bets waiting for a winner while unsure whether you have a 0.10% edge or a 0.10% (or worse) disadvantage. In practice that's impossible to know until you've bet thousands of bets. But my original point was more the question: Is it possible to make money on statistical football betting at all?

Was not saying you should back a horse at 50 if you think it is a true 40 chance. You need a margin for error, at least 67. A bookie who's odds compiler works a horse out as a 1.5% 67 chance will often offer 41. So if he is capable of working out races that tightly, so should we. In my original post was just illustrating the difference in money won between the two prices, compared to shorter prices. Finding big priced value bets is just as easy / difficult as finding short priced value bets. Had I only backed 10/1+ shots this year my record would be considerably better. Much shorter strike rate , but you don't need a large strike rate at those prices. Of course you need "powerful mental stamina" Lardonio, that's needed whatever way you gamble. If you're not willing to put the study in, then you are right, it is impossible.

[Lardonio]

Re: How big a margin I think we've strayed a bit from my original purpose of this post, possibly because I'm not a native English speaker. I'm sorry about that. My problem is that I think it's very difficicult to find big priced value bets in football when allowing a margin for error. I have 30,000 matches worth of data, and statistically it seems that the bookies have got it spot on, with a comfortable margin for themselves. On Betfair occasionally value can be found, but with such a tiny margin for error which makes me suspect that the so-called value is not actually there. It would be very interesting to hear about a purely statistical system on correct scores that works. I have put a lot of study in, but just because you put a lot of time in does not mean value can be found..

[Gingertipster]

Re: How big a margin You seem to write perfectly good english to me Lardonio. I have no experience on football betting. Suspect knowledgeable punters in any code can identify value bets at big prices. Doubt if it can be done with statistics though. Agree, just because a punter puts in a lot of time does not mean he will find value. But if a punter is not willing to put in the work then he has no chance. Anyway, am sure you put in enough work Lardonio and wish you luck with your quest.

[KThom]

Re: How big a margin Interesting thread. I’ve been working on this very aspect when it comes to football betting.

I’ve managed to get a working formula together which will automatically calculate the total book, and give odds for the home team – away team and the draw.

The formula I’ve used is pretty complexed. It incorporates the last 6 games played by that team at home, away and also the last 5 head to head meetings between the sides. I already have this data and it automatically updates as the season progresses. I have weighted the form 85% to the recent head to heads 15% and at the end of it out spews a home team % win chance.

From here I am able to look at the total draw frequency for the past 5 seasons and weight the draw odds accordingly. (The championship in England has high draws – so the projected odds are lower compared to the premiership where the draws are less frequent historically and therefore the projected odds are higher). I am then able to weight the total percentages and calculate the odds from here.

Below is my table for the last weekends Championship games.

Date	Home	Away	H Odds	D Odds	A Odds	My H Odds	My D Odds	My A Odds
24-Oct-09	Barnsley	Bristol City	2.50	3.40	3.00	1.77	3.59	6.34
24-Oct-09	Coventry	West Brom	3.50	3.40	2.20	4.80	3.24	2.07
24-Oct-09	Crystal Palace	Nott'm Forest	2.50	3.40	3.00	3.81	3.14	2.39
24-Oct-09	Newcastle	Doncaster	1.57	4.00	7.50	1.89	3.46	5.45
24-Oct-09	Peterboro	Scunthorpe	2.38	3.40	3.25	2.39	3.24	3.68
24-Oct-09	Plymouth	Ipswich	2.50	3.40	3.00	3.35	3.14	2.61
24-Oct-09	Preston	Middlesbrough	2.50	3.30	2.88	4.23	3.24	2.20
24-Oct-09	Reading	Leicester	2.60	3.40	2.80	4.30	3.24	2.18
24-Oct-09	Sheffield United	Cardiff	2.50	3.40	3.00	3.86	3.14	2.37
24-Oct-09	Swansea	Blackpool	2.30	3.40	3.40	1.81	3.59	5.98
24-Oct-09	Watford	Sheffield Weds	2.38	3.40	3.25	3.54	3.14	2.51
24-Oct-09	Derby	QPR	2.75	3.40	2.80	2.13	3.35	4.31

The problem now is – how do I transform and “value” into profit. I trust enough the odds calculation, the stumbling block now becomes the staking, or selection process.

If we take last nights game between Reading and Leicester for example.

Actual best odds for Reading was 2.60. My real odds were 4.30 so there is no value there.

The Best odds for a draw were 3.40, and my projected odds for the draw were 3.24. So there is some value here.

The best odds for Leicester were 2.80 but my odds calculate they should be at 2.18.

I have done some simple testing, but only at a basic level (staking £20 on any single value selection per game)

Which would include Barnsley, Draw (Peterborough v Scunthorpe) and Swansea.

Also looked at lay selections where the two other markets are below actual market price. So laying Coventry, Crystal Palace, Newcastle, Plymouth, Preston, Reading, Sheffield United and Watford.

I’ve also briefly looked at adjusting the backing stake dependant on how difference the projected odds are too the actual odds.

I’ve only really looked at historical data for the last 3 prem and championship seasons. The Premiership seems to be more profitable than the championship which would end in negative income.

I am open for suggestions on staking or selecting. I don’t want too may – nor too few.

If anyone has suggestions I can test this against a range of leagues. I’d like to find a selection and stake process which can work on any league. But sometimes creating the one size fits all system is impossible.

[Smokin Joe]

Re: How big a margin KThom, I would have thought that you should be able to use Kelly staking for this. Using your Leicester example, if the "true" odds were 2.18 but they were available at 2.80, then on a bank of £1,000 you would bet £158. I find that backtesting any system by using Kelly staking gives a pretty good idea of its worth.

[Lardonio]

Re: How big a margin KThom, I think you should do your back-testing based on various intervals of value, so maybe you can find out if there's any lower limit of value you need to be profitable. For example in your Barnsley - Bristol City match your calculated odds is 1.77 while you were offered a staggering 2.50! 1.77 equals a 56,5% chance with your system, while the bookmakers have estimated the chance to be 40%. That gives a percentage advantage of 16,5% which is HUGE. You can calculate the statistical value of all your bets using calculations like this, and then start tracking your bets, sorting the bets into various categories, f.eks. 0-2%, 2.1-4% etc. and see how the betting performs. The formula in Excel would be =(100/A1-100/B1) where A1 is the odds offered by bookies and B1 is the true odds. I would however be a little suspicious to finds of 16,5% value as such huge perceived value suggest, in my experience, flaws in the system.

[KThom]

Re: How big a margin Thanks guys, will give them a whirl when i have the chance

[Vilamoura04]

Re: How big a margin Kthom I imagine you already know this but what you are doing sounds similar to what this guy here does http://goonersguide.com/numbers_spreadsheet.htm The "DV Sheet PK line" is his value bets (currently there are no selections), where the odds offered are flagged as being 'value' in comparison to the probability. TheGooner says this system has only backtested as being profitable if you level stake. The system also doesnt pick up odds which offer 'too much' value...ie if the odds are say more than 120% of the probability...this means something is 'up' and that bet should be avoided.

[KThom]

Re: How big a margin

Kthom I imagine you already know this but what you are doing sounds similar to what this guy here does http://goonersguide.com/numbers_spreadsheet.htm The "DV Sheet PK line" is his value bets (currently there are no selections), where the odds offered are flagged as being 'value' in comparison to the probability. TheGooner says this system has only backtested as being profitable if you level stake.

I hadn't seen that before Vilamoura. After some stringent testing i agree with "TheGooner" regarding level staking being the profitable way to go. I've backtested the last 5 seasons in the Premiership and Championship. I've also added an additional filter (ie if the best value by odds is the away team, but my system shows a home win or draw as projected outcome i won't select it.) For all selections 2,192 it showed a 0.4% positive yield. (I also agree with above statements saying over 14-16% yields could have other factors other than stats influencing odds (Injuries, derbies, Manager change etc) so i also tested on each 2% edge as suggested by Lardonio. For selections under 10% edge it showed a 7.39% yield from 911 games. For selections under 8% edge it showed a 8.73% yield from 662 games. For selections under a 6% edge it showed a 10.32% yield from 439 games.

[Vilamoura04]

Re: How big a margin So if I understand it correctly, the closer the odds are to the implied probability, the higher the yield? If so, TheGooner (sorry to keep banging on about it - Im just trying to help :ok)also found the same correlation with his sheet. The DV PK line filters out selections where the 'value' (odds>probability) is more than ~110%. Basically, if the price is too good to be true...something is up dont trust it.

[KThom]

Re: How big a margin

So if I understand it correctly, the closer the odds are to the implied probability, the higher the yield? If so, TheGooner (sorry to keep banging on about it - Im just trying to help :ok)also found the same correlation with his sheet. The DV PK line filters out selections where the 'value' (odds>probability) is more than ~110%. Basically, if the price is too good to be true...something is up dont trust it.

Pretty much how i've assessed it too. I've done the same testing for Italian, Spansih, League 1 and League 2. The Italian league produces very poor results. I don't know whether i should just discount thios league due to poor prior results or look for a logical reason why. Also, Away value bets seem to give a much higher yield. Again this could be down to chance - or a specific reason behind it.

[Vilamoura04]

Re: How big a margin Oddly - when putting his ratings together gooner was very succesful in Italy. The league he had most difficulty with was the La Liga. I dont know what he eventually concluded but I remember his hypothesing that a lot of teams in Spain have a generally ordinary squad but one or two key (often foreign) players..if those key players were missing then you may as well throw the form book out of the window. I wonder if the same could now true in Italy? Without wanting to get into an argument with anyone - many have opined that Serie A's financial troubles have certainly affected squad quality in the league - which suggests increased reliance on key (ie expensive) players. However the italians generally enjoy teamwork and organisation at the expense of entertainment and flair - so Im not convinced. In terms of away value - I have seen this come up in various ratings. When i put my own ratings together I think I had simply underweighted for home advantage. But for others the conclusion I have come to is that there may be value in a long shot away team in a given fixture but the strike rate is so poor that it is not a viable strategy. For example, Sunderland may be value when playing at Old Trafford as statistically they will win once in ten visits...and the odds imply they will only win once in twelve visits. However you might have to wait for them to play each other ten times before Sunderland actually get that win - and most people arent going to stick with strike rates like that.

[Lardonio]

Re: How big a margin I don't think strike rates of 1 in 10 or 1 in 12 is a problem, that is more of a rule if you play for example the correct score markets. I agree it would be boring to only do betting on Man United vs Sunderland, one bet each year and wait 10-15 years for it to come in ;) but if you have enough bets then you will get regular winners and it will be easier to avoid losing patience. If you find value then the winners take care of themselves, some professional gambler said, and it's certainly true. I think the often forgotten key to doing a successful value betting strategy is using a system that produces a lot of bets, that way it's a lot easier to evaluate if the system is actually working or not. If you have 50 bets a year with odds in the 8-12 range you will have to keep doing it for years to produce any reliable conclusions on the performance of the system. I'd rather do 100 bets a week, it helps keep things more interesting.

[KThom]

Re: How big a margin I agree with you Lardonio. If you have done sums correctly - it wont matter how frequent or infrequent a 10/1 chance offered at 12/1 comes round - it will be profitable in the long run. I'm still pumping French and German scores through my test machine - that way i can back analyse without the wait of 2/3 years. I'm also interested in considering the Draw no Bet market. However finding previous odds will be a challenge. Is there a recognised formula to convert standard odds for H/A into DnB odds? For example a 2.00 (evens) bet on normal match outcome will be more or less a 1.60 on DnB?