Saturday, April 7, 2012

Predictions that are never right

Right then, here's the second big table in my mystical magical spreadsheet. I've been tracking the bets from dred's excellent WPL game over on BigSoccer and trying to get a feel for how fans feel about how well each team should perform each week. It's not the most robust way to judge a teams expectations, but given 40 something players each weighting bets on matches, it's about as good a way as I could think to figure out who's overperofmring (or underrated) and who's been misfiring (or just isn't as good as everyone thought, LA, I'm looking at you). I look at it from both a weighted fraction of expected full points for each game as well as put together a little equation that takes peoples bets and tries to predict a winner (and the goal difference) for each match. Let's look at both.

At this point in the season, from a normalized fractional full point (FPP) (i.e. split 3 points between each team absed on weighted bets and then divide by 3), we get an expected seasonal performance that looks like this.

     Seattle - 0.846
     LA - 0.823
     Kansas City - 0.799
     Salt Lake City - 0.729
     Portland - 0.690
     Vancouver - 0.658
     Dallas - 0.655
     San Jose - 0.541
     New York - 0.537
     Tortonto - 0.466
     Chicago - 0.444
     Colorado - 0.428
     Columbus - 0.428
     Philadelphia - 0.388
     Houston - 0.309
     DC - 0.266
     Chivas - 0.131
     New England - 0.093
     Montreal - 0.057

Going into the season I actually think that might have been a good prediction, with the obvious cavet of wtf are you people thinking with houston, last year's Cup runner up!?! I guess people are thinking they'd pull a 2011 SKC and start the season on the road horrifically, but still... Anyway, that's all fine and good, but let's push it one step further and see where teams are at now compared to where we thought they'd be. And, for fun, let's try to see how well we've predicted wins, draws, and losses, using a totally made up and arbitrary excel equation. Basically, I tried to look at the weighted bets and find out who was favored for each match, and then looked at the weighting and tried to find out if there was a consensus. If 75% of the bet weight was for a team, they'd be assigned a win. If the average bet on that team was <150, they'd be assigned a 1 goal win. 150 < bet < 250 was a 2 goal win. Am average bet of more than 250 was considered a 3 goal blow out.

For the record, this model has about a 50% chance of picking the result exactly after 4 weeks. It's got a 30% "push" rate (i.e. predicted a win and got a draw, or predicted a draw and got a win). But really, I think 50% success rate is the conservative way to gauge it. So, at this point, here's how the table would look if fan betting controlled the league.

Team Predicted
LA 9 3 6
Toronto 4 0 4
Portland 7 4 3
Seattle 9 6 3
Chicago 6 4 2
Philadelphia 2 1 1
New York 7 6 1
Dallas 5 4 1
Vancouver 8 8 0
Chivas 2 3 -1
Montreal 0 1 -1
Kansas City 10 12 -2
DC 2 4 -2
 Salt Lake City 7 9 -2
Colorado 7 9 -2
Columbus 4 6 -2
San Jose 6 9 -3
Houston 1 6 -5
New England 0 6 -6

Obviously a few things jump out. LA and Toronto seem to be really underperforming expectations. I'm not sure Toronto is that big of a surprise, actually, but obviously people in LA are getting worried. Houston's overperformance I think is more a factor of them starting so many games on the road, but kudos to New England and San Jose for both stepping up. I think they might turn out to be the two surprise teams of the year.

At any rate, you've herad it here first. LA is the most overrated team in the league. You can prove it with science.

Finally, while speaking of predictions, let's make some for Week 5, shall we? We're already 2 for 2 as far as midweek games go...

Salt Lake City beats Montreal by 3 goals
Dallas beats New England by 1 goal
Montreal beats Toronto by 1 goal
Columbus and New York draw
Kansas City and LA draw
San Jose beats Vancouver by 1 goal
DC and Seattle draw
Salt Lake City beats Colorado by 1 goal
Portland beats Chivas USA by 2 goals

Like I said, 2 for 2 on results for SLC v Montreal and Dallas v New England, and even got the New England GD correct. Let's hope SKC beats the junk out of LA though...

No comments:

Post a Comment