Thursday, 25 November 2010

Form Fitting

Perhaps SAF doesn't read this blog, or if he does, he forgot to mention in his team-talk that Manchester United needed a three goal win to move ahead of Chelsea, but the ratings only predicted a 0.68 goal win, so they inch slightly closer with their 1-0 win.

In the other games where I have ratings for both teams, Schalke '04 were rated .84 better than Olympique Lyonnais and won 3-0, and Tottenham Hotspur were rated 2.74 goals better than Werder Bremen and won by 3-0 for a good evening of results.

Speaking of Lyon, if anyone is looking for a book to add to their list for Santa this Xmas, you could do a lot worse than ask for Soccernomics, by Simon Kuper and Stefan Szymanski. Among a number of fascinating topics, it details the reasons between the rise of Lyon in the past few years from a nothing provincial club in France, to a major European power (last night's result notwithstanding). Bloomberg News describes it as "a blend of Freakonomics and Fever Pitch, bringing surprising economic analysis to bear o the world's most popular sport". If you enjoyed Moneyball, by Michael Lewis, then you will certainly enjoy this one. It's not betting related, but it's a very well written book on a subject that I suspect most readers of this blog would enjoy.

I recently linked to Mark Iverson's post on his football system, and found myself in the role of Anonymous by questioning Mark on some of the details. Some of my questions, and Marks's answers are:

Why 4 matches? My system only takes into account Premier League matches (not cup games etc.) so if you go back more than 4 games then you could be talking about 5-6 weeks ago. That's too long ago for me.

Do you weight the most recent game heavier? Not in the rating, but it's something I consider when looking at the overall trend of the 4 games.

Do you account for the games being at home or away? Yes. I use different scorecards for each but a 20 rating at home should represent a similar 20 rating away so you wouldn't know the difference.

Do you account for the strength of opposition? Yes.

Why 4 to 34 and not 0 through 30? There's some free points on offer. Maybe I should change this to 0-30! lol! Won't do it for this season but maybe next year.
Mark also posted up his teams ranked by form, and I compared them with my form figures. The results were a surprise, in that there was very little correlation at all. As Mark says in his post "Some unexpected teams are towards the top" and I agree. Everton third on form? Well, I have Everton third too, only I have them third from bottom! Mark goes back four games, and I go back six, with the more recent games given more weight, but Everton's last six games are, most recent first:
A Sunderland D 2-2; H Arsenal L 1-2; H Bolton D 1-1; A Blackpool D 2-2; H Stoke City W 1-0; A Tottenham D 1-1
I have Everton underforming expectations in all but the Sunderland and Tottenham games, and the Tottenham game is about to drop off. How are Everton ranked third on form? For the record, Everton are ranked seventh in my overall ratings, having started the season in fifth place. We do agree on West Bromwich Albion being ranked 19th though, and we have Bolton at the top or second, but for the most part we have some big differences. But if we all had the same opinion, we would never find value.


Mark Iverson said...

Thanks for the mention Cassini.

I think the differences must be in the amount of weighting we assign to the results themselves.

It's a factor for me but not a huge part of my performance ratings. I set-out to identify if 'the best team won' rather than just take the final score as the true picture.

If WBA were to beat Chelsea 1-0 it doesn't necessarily mean they were the better side.

Good luck as always.


Anonymous said...

I've been waiting for ages to see if Cassini would ever twig that a quality ratings system is not based purely on results.

It's not as if the best brains in the business - the guys who both set and move the markets - simply input final scores into a spreadsheet and then put their feet up. But thousands of wannabee pros do just that and expect to have a magic answer. And will come up witn nonsense like in the latest post where Spurs were expected to win by nearly 3 goals at home to Bremen.

How you move it on from simply pure results is the tricky bit. There's no single right answer but there is, or should be, a single aim, and that aim is key.