Friday, 20 August 2010

Superiority Is Complex

Some mostly positive comments to my last post.

Anonymous said...

I can confidently state a team has never been a 2.5 goal favourite away from home in the Premiership. And nor are Chelsea this weekend.
I’m not sure about this. As I mentioned in my post, Wigan are currently the bottom rated team and in poor form while Chelsea are currently top rated and in good form. Admittedly the form carries over from last season, but if there is any game where my ratings will predict a 2.5 goal superiority for the away team, this is it. The 8-0 win at Stamford Bridge, and other extreme results, are all smoothed out, and the 8-0 win isn’t in itself a big reason why the calculation comes out at -2.5.
Also, something is seriously flawed if you have a team with an expected one goal superiority at 2.27 to win.
The average for teams with a one goal superiority is actually at 2.29 right now, seriously flawed owing to two factors – a low sample size and the start of a new season, meaning the Elo ratings do not accurately reflect the strengths of teams. I highlighted the number simply to illustrate why Elo ratings can’t be used at this stage of the season with any expectation of being profitable.

Danny makes an excellent point with his comment about the problems of ...
teams have very different home/away personalities; like Villa last year who were great on the counter attack when away but struggled to break down teams at home - they ended 3rd in the Away league table, 12th in the Home league table
I did think about maintaining two different ratings for each team, and may look to do this in the future, but for now I simply use an additional factor in an attempt to adjust for teams that perform significantly better either at home (think Mainz, Mallorca, Udinese, Lille, Fulham or Blackburn for example) or away (e.g. Racing Santander, Lazio, Werder Bremen, Wolfsburg, Nancy, Aston Villa).
How do you adjust for home/away form in your ratings? I'm guessing an away win is worth more?
Yes. A win away against an opponent results in more points being gained than a home win against the same opponents.

Thanks to Talkbet for his kind wishes. Crystal Palace’s signing of Edgar Davids is a little strange given that he’s not played in two years, but a pay-as-you-play deal is not a huge financial obligation for the club.

At worst, his presence should generate some interest and put some ‘bums on seats’ and at best, he is as good as ever and inspires Palace to promotion, and on to a Champions League place in 2012. I’m not investing too much cash on that possibility though.

1 comment:

Anonymous said...

Ok. I don't have records but let me know if you can find an Asian handicap price in the whole history of the premier league for an away team at -2.5 evens or lower. (Evens isn't actually quite the correct price for a superiority of 2.5 but it's near enough for this debate).

It would take fairly extraordinary relative team news, an enormous motivational gulf or a fix for such a price to be close to appropriate. Even Derby County (or Sunderland in their horrific season) wouldn't (shouldn't) have managed it at home against anyone.

That said, the Asian market wasn't always as efficient as it is now but I still don't think it would have been sufficiently inefficient in the past to price a game as above.

The price you mention of 2.29 for a one goal favourite is actually very close to the correct price for a home team v a team of equal capability. And home advantage sure as hell isn't worth a whole goal.

A one goal favourite should normally be priced in the 1.65-1.70 region.

You may find it interesting to search for something like the "detech football predictor" to get an idea of more accurate supremacies for different match-ups. They're not necessarily correct individually but will give a fairly reasonable assessment.