Wagering in the SuperMargin market this season has been as exasperating to me as it's been interesting since I've not known how to feel about a particular score at any point in the game. All I've found myself doing as the game progresses is hoping that the score doesn't stray too far from the required bucket, feeling pleased if it's in the correct bucket sometime in the final term, then rooting for a behind-a-thon from that point onwards. It feels a little like closing my eyes for the first 3000m of the Melbourne Cup then barracking for a short half-head victory by the favourite.

I'd much rather know how my money's doing while I can pretend that my supporting it in the correct fashion can make a difference, so I decided that it was time to apply some statistical modelling to the issue, the full details of which are available in a blog over on the Statistical Analyses journal. 

For our purposes here, all we need is this summary.

(Please click on it for a larger version.)

To use this table, select the column applicable to the bucket we've wagered on. Let's assume for the purposes of explanation that we're on the Home team to win by 20-29 points. The first block of data in this column tells us that the range of scores that we should be happiest about at quarter time is a Home team lead by anywhere between 10 and 16 points. For any margin in that range the probability of our bet winning is a little under 11%.

Now if we've only secured $7 for our SuperMargin bet, an 11% likelihood of success means that the bet has a negative expected return. In fact, if we only secured $7 for our bet, there's no margin at quarter time that would mean our wager had a positive expected value. At this point, being in the preferred range merely minimises our expected loss. (Think of this as sneaking a quick peek as the horses pass the post for the first time, just hoping that the favourite hasn't been trailed off 8 lengths by the pack.)

Looking next at the middle block in the 20-29 points column we see that, at half time, our preferred range of Home team margins is now 14 to 20 points, which is slightly less than the final margin we desire. Any score in this range provides us with around a 14% probability of collecting. Better still, if the margin is in the range of a 15 to 19 point lead for the Home team, our $7 wager now has a positive expectation.

Finally, the lower block reveals that the preferred range of Home team leads at three-quarter time is 17 to 22 points, which provides a greater than 1-in-5 chance of success. In fact, any Home team lead in the range 6 to 34 points still provides us with a positive expectation.

Data in the other columns should be interpreted in a similar way, noting that for the "Draw" column the price at which the possibility of profit has been assessed is $51 rather than $7, since this is almost always the price on offer for this bucket (well thimble really).

Hopefully this table will make following SuperMargin wagers a little more enjoyable.

(For any of you who have read the blog on the Statistical Analyses journal and who are wondering which of the two models I've used in the construction of this table, I went with the multinomial logit formulation mostly because it seems to me to be the more conservative of the two.)