Compiling Match Odds (Part II)

In the first part of this series (HERE), we looked at Goal Supremacy. This produced match odds based on the goal difference of each team in question over the last six matches. Taking a match at random (and it really was random), the compiled odds weren't too far away from the odds that the bookmakers were offering. So this would indicate that we've made a decent enough start on the path to generating our own match odds.

This was all fine and dandy, but it's not too difficult to see some problems with this first method. Chief amongst these is that bookmakers are really pretty good at pricing up matches (the bastards), and if we perform our own pricing just based on goals scored and conceded, we'll probably not gain any real ground. Relying only on such an obvious and basic method may not throw-up any value bets for us to get our teeth into.

Also, as discussed in the comments after the initial posting, there is the issue with the type of goal scored. Should a goal scored against Man Utd or Man City be given the same points value as a goal scored against Wolves or Bolton? And not discussed but just as relevant is whether to use all the goals for all the past six matches. By that I mean should we use all the last six matches, home and away? Or should we just use the last six home matches for the home team and the last six away matches for the away team? Yes, yet another weighting issue to consider. Sorry about that.

In essence, we can say that Goal Supremacy was a good start at trying to price up a match, but if we're serious about our pursuit of finding value, then we probably need to delve just a little bit further.

The next method, detailed below, sadly does not address all the issues just outlined, but does provide another way of looking at football data and another way of compiling match odds. This can only help us to build-up an overall calculation. Later methods should begin to address more of our concerns.

Dyads and Triads

You'll be pleased to know that this is nothing to do with gang culture. Unfortunately I cannot remember where I stole this method from, so am not able to give due credit and reference to the originator. This doesn't really matter. I'm sure whomever I stole it from probably stole it themselves. That's how these things work. I do seem to remember, however, that this was a system devised by some professor or other when he was trying to formulate an effective way to play the Pools.

So what is a dyad and a triad? Nothing too dramatic: a dyad is a sequence of two matches and a triad is a sequence of three matches. If we look at dyads first, we can see that from the three possible results, we can acquire six possible outcomes depending on whether we're dealing with the home team or the away team:

1. Home Win    (HW)
2. Home Draw   (HD)
3. Home Loss   (HL)
4. Away Win    (AW)
5. Away Draw   (AD)
6. Away Loss   (AL)

A dyad is a combination of any of these two, and the triad is the predicted third result. These groupings have been looked at in detail over the course of several seasons, and back-tested to predict the result of the third match in the sequence. 

Now my stolen... erm I mean acquired, data for this method was taken from 10 years worth of English football league results, and looked at the percentage of home wins, draws and away wins based on every single dyad combination. For example, the triad result for the dyad "HW, AL" showed the following:

   Home Win = 47.38%, Home Draw = 26.19%, Home Loss = 26.43%

Of course, this is not too far from the long-term average anyway, but it's these relationships and the results recorded for each combination that going to be is the basis of our predictions, allowing us to draw-up probabilities for each outcome. Opportunities will arise when the dyads throw-up odds away from the norm.

Right, so when looking at Goal Supremacy, we used the upcoming Arsenal v Spurs match. So let's now use the same match for this method.

   Arsenal's current dyad (EPL only) is: HW, AW

   Spurs' current dyad (EPL only) is     : AD, HW

Note, that the latest match always comes last in the dyad (this is important). Using the data available, this equates:

   Home = 49.30%, Away = 25.10%, 25.60%

And if we divide 1 by all these percentages, we get match odds of:

   Home = 2.03, Away = 3.98, Draw = 3.91

It's about now that you may be groaning and saying to yourself, "Well, how the hell am I going to generate all this data that he's talking about? I might as well stop reading now."

That's your decision, of course, but I would like to just mention that I will be providing all this data for you in a spreadsheet. Please find the link to download the file at the end of this post. The data comes in the form of an Excel workbook. Inside the workbook are the following worksheets:

• "Config" - contains each combination to be looked-up with a lookup number (defined as "dyad_lookup" name)
• "Triad HW" - Every dyad combination that resulted in a home win triad over the last ten years
• "Triad AW" - Every dyad combination that resulted in an away win triad over the last ten years
• "Triad D"     - Every dyad combination that resulted in a draw triad over the last ten years
• "Analysis"   - contains an example of how I perform the lookup. Examine the H9, I9 and J9 for the lookup itself. If you're an Excel wiz, then you may have a superior method to the one I've used.

And that's it. This method, even if used on it's own, is still good enough to raise any predictions well above random. Good luck.

Download the data in Excel spreadsheet format =>  HERE