I have been browsing thru the excellent Smarkets prediction markets on soccer (a.k.a. association football). You have 3 possibilities for betting:
- the home team wins;
- the home team loses;
- there is a draw.
The soccer rules (which allow for a draw) make me uncomfortable for betting on the outcome of a match. How can you assess the probability of a draw? I don’t get it. The Soccer Federation (FIFA) should get rid of “draws” in soccer. As for forecasting and betting, it is annoying.
If you take matches that did not result in a tie, divide them into equal groups according to whether their odds for a tie were relatively high or low, is error for winning-odds higher in the high tie-odds matches? In other words, do the tie-odds contain some information about confidence? And/or do they reflect the expected number of total goals? Someone who actually bets on this must have looked at this before..
I mean error when you adjust for different tie-odds. Of course higher tie-odds matches will have more error in non-tie outcomes.
I’d guess that tie-odds are inversely proportional to expected scoring.
I’m not clear on your objection to betting on ties, Chris. What is particular problem with assessing the likelihood of a draw? Lots of data exist on wins/losses/ties. If the data is at all useful in assessing the likelihood of a win, wouldn’t it also be useful for assessing the likelihood of other possible outcomes?
Mike,
I blogged my sentiment (not a rational investigation), so as to see if it would prompt some comments from readers. Thanks for your participation.
Any soccer bettors reading this?
Trying to predict one out of three is more complex than 1 out of 2.
>> Trying to predict one out of three is more complex than 1 out of 2.
You don’t have to predict anything. You want to find instances or conditions where the market is underestimating or overestimating the likelihood of a draw.
For instance, I found at some point that at the start of the season (first 6 to 8 weeks) teams have a bit more problem with scoring. This more often results in a draw or < 2.5 goals (there is a correlation between these two) than the market anticipated. That was based on 5 years of data and something like 30 leagues.
And I had a method to bet on a draw or NOT(draw) depending on results of prior matches from both teams.
In general, a draw is very hard to predict as there is a lot of luck involved. A lot of factors you can't control or predict. So profit margins tend to be very low in soccer when betting on win/draw/lose and that is the problem when using a systematic approach and when you're on 3-4% commission. You end up working for betfair. That's where I lost interest.
Another approach would be to detect instances where both teams would benefit from or be happy with a draw. This requires a lot of experience/knowledge and devotion. Although I expect profit margins to be higher (so that you can beat commission) the problem of this non-systematic approach is that you don't get to bet on a lot of matches. And that's risky because you become more dependent on luck in the short term. Or you would have to play with limited stakes. In the end it is all about the risk that your bank will collapse due to a bad run. It's the same in poker, btw.
Overall, it is very exciting to study these markets but soccer represents the most mature market and toughest to beat after commission. You need to be on 1% commission.
As for calculating odds and using this as a method to detect potential value, I don't think it is the right approach. Well, we wouldn't need betting exchanges if one person could do it.
This is all perfectly standard in soccer betting, and in other sports that allow for ties. All the earlier comment are good.
I hate betting when a draw is one possibility.
No, You hate sports where a draw is a possibility. And, so do most people. That’s why the NHL now has shootouts instead of ties. It’s much more exciting.
People, for the most part, are happy to bet on just about anything.
Daniel, I am starting an international campaign right now to suppress draws from sports.