source

A couple of disclaimers to start with, firstly I am a massive numbers geek so the idea of trawling through reams of data, pulling it all together and then trying to find some discernable pattern appeals to me no end. That being said, I am not a professional gambler so I have no idea if this model will provide any value or indeed if I have the discipline to stick with it and critique my own work to the extent that I can see whether I am just being lucky/unlucky over a short period of time.

Furthermore, this is not a model for predicting winners (in more ways that one), I see plenty of supposed tipsters on this platform who love to back the favourite. In fact, some of them like to back the favourites every time and then brag that they are predicting more winners than losers, to those people I deliver to you the Nancy Pelosi sarcastic hand clap

No, instead this model is looking for value in odds and specifically value in odds for the draw (X).

Why the draw?

It's an observable fact that punters tend to favour a result in either direction as opposed to the draw. In fact, you don't have to go any further than Scorum itself to see this phenomenon. I counted the number of tips and predictions posted this weekend under the soccer tag and found that of the 42 made on the result of a game only 3 forecast a draw (7%). The reality is that most major European leagues have a draw percentage of between 20-25% which suggests that the draw, as boring as it may be, is often undervalued in the odds market.

Features of the model

The model draws on general statistics for both sides and also factors in their relative league position as well as their propensity to draw games.

For example, the draw percentage for the Premier League going into this weekend of fixtures was just over 19% so about 1 in 5 matches would end in a draw. However, if you start to look at teams that are closer to each other in the league table this percentage rises. For example, in fixtures between the current top 6 of the EPL the percentage rises to 22% ( that will have increased following Spurs and Arsenal's draw this weekend). Equally, there are some teams that seem to draw more of these games than others. Liverpool, for example, have played 8 games against fellow top 6 members and drawn 50% of those matches. On the other hand, Manchester City have played 9 games against the same calibre of opposition and drawn just once. The model gives a higher weighting to fixtures that include evenly matched sides that have a tendency to draw games.

Ultimately once all the data has been crunched, what I am trying to do is find matches in which I believe the probability of a draw is 5% better than the implied probability of the odds on offer.

Early Trials

So far the model has suggested 9 games in which the draw offered better value than the implied probability by 5% or more and I have had success on just 2 of those lines. It is, of course, early days and the reality is that with the odds on offer for draws I probably only need to win 1 in 3 bets to turn a small profit.

The next game I will be betting on is the Leganes v Levante match in La Liga. You can bet on this game being a draw at odds of 3.59 by using @julienbh's bet bot which matches odds from pinnacle.

The implied probability of this match finishing in a draw is 27.85% but the model suggests a 33.23% likelihood of the game ending all square.

The real proof in the pudding will of course be to see how this model pans out across 100 or 1000 games when the likelihood of good or bad luck can be ironed out of the statistics - time to get betting!

Speaking of good odds, don't forget to follow @fortunabetting who offer odds boosts on a variety of lines every week. You could have taken Everton to draw with Liverpool yesterday at odds of 4.62 better than any other online bookmaker or exchange