Stats Chart Well

Some more constructive comments to talk about.

Dave has some interesting ideas, and not just the erroneous attribution to Sir Winston Churchill of the phrase "Lies, damn lies and statistics"!

He suggests that there is an argument not only for looking at statistics by division, but taking that a step further and breaking the season into three periods – early, middle and late.

This idea certainly has merit. Early season form is unreliable because close season changes, both managerial and playing staff, take time to bed in, and new teams to a division are something of an unknown quantity. Would Burnley have beaten Manchester United in January? Many late season matches can be harder to call if one or both teams have nothing to play for, (especially in Italy), so there’s an argument for restricting investments to games where both teams have an incentive.

Dave suggests that the early and late periods produce more goals than the middle period, which if true would suggest that not only unders, but also the draw would be more likely in that middle period. Food for thought, and another mini-project in the works!

Anonymous is confused about my 2 goal predicted wins, and I, in turn, am confused by his reasoning for turning a profit of 84 into a loss.

I'm a bit confused by your returns on bets on 2-0 and 3-1. 100 units returning 184 suggests, to me, a profit of 84? Not sure if that's what you mean or not. But, if so, then the investments presumably aren't profitable as you would make 84 and lose 100 (assuming level stakes) by covering 2-0 and 3-1.

A return of 184 to a stake of 100 means that for every 100 points staked you receive back 184 points, i.e. a profit of 84 points. Both the 2-0 and 3-1 scores would have been very profitable in the Premier League last season.

He goes on:

Of course, you also make no mention of other two goal scorelines which I find a little strange. From previous posts last season, I assume you would also be covering the "unquoted" or has that conveniently been left out because there were no 4-2s or 5-3s? (i don't know if that was the case or not)

It’s not really that strange. If you have ever looked at the Correct Score market on Betfair, you will know that 2-0 and 3-1 are the only correct two goal scores that can be backed. If a win is by 4-2, 5-3 etc., then yes, technically the prediction was correct, but from a betting perspective these fall into the less lucrative Any Unquoted bracket. Such results are included in the overall wins total along with the 1-0 and 3-0 wins etc.

Another Anonymous comments that

“A team expected to win by two goals can, on average, be expected to win 80% of their games (give or take not very much). 73% isn't a particularly good hit-rate (albeit that your results are relatively short-term).”

I suspect that Anonymous means “teams expected to win by 2 goals or more”, because the average price on the +2 selections was way higher than the 1.25 that would be expected if the probability is 80%. (Some examples from 27.Feb; Birmingham v Wigan was 2.0, Chelsea v Manchester City was 1.44).

My +3 and +4 selections are not included in the +2 results. When you factor this in, the 72.73% strike rate becomes very good requiring 1.38 as an average price to be profitable, albeit from a limited sample size.

He does however make a good point regarding the problem of supremacy margins having no "in-betweens" - supremacies of 0.5 to 1.49 are all being rounded to a +1 expectancy which is clearly not good. If the numbers were 0.49 and 1.5, they would be much different. It’s something I will be looking at this summer, as the spreadsheet evolves.

A major flaw in the way you approach things is that you seem to have no "in betweens". i.e. your methodology works on the simple basis that a team is expected to lose by two, lose by one, draw, win by one, win by two, win by three etc.

With a little work, you could achieve better results if you found a way to incorporate more precise expected (average if you like) supremacy margins a la the spread betting firms.

Incidentally, can anyone spot the relevance of the picture in this post?