Wednesday, April 21, 2010

Early 2010 NFL Season Projections

The NFL schedule came out yesterday which means I can now plug it in to my computer to see what it projects for the regular season ending records.  The full projections are below, but first a little explanation and commentary.

First, these projections are based on the final ratings from the 2009/10 season.  So they assume the teams performance will on average be unchanged, but clearly some teams will improve, others won't, and injuries will be a factor as well.  But these projections are still useful as last years rating is a good starting point and the projections do take into account the specific schedule of each team.

Second, how do these projections work?  For each game, the computer predicts the chance of a team winning that game based on last years rating, the opponent, and where the game is played, and then looks at every permutation of how a team could achieve every record from 0-16 to 16-0 and the chance of that occurring, then it identifies the record that has the highest probability of occurring.  This is what appears in the Record column along with the chance of that record in the % chance column.  Since it does compute the probability of every record, I also include the probability of a team winning and losing one more game but have the full range of probabilities if anyone is interested.

With that out of the way, what does it project?  It says that New Orleans and San Diego will lead the league at 11-5 with 6 teams close behind at 10-6.  Even more (7) teams will be just behind at 9-7 making it 15 teams vying for 12 playoff spots.  At the bottom, St. Louis and Detroit will still be the cellar dwellers, but again, this doesn't take into account them presumably improving with their top picks in the draft.

You might ask, why is New Orleans only projected to go 11-5 after going 13-3 last year?  And why a 4 game drop-off for Indy after they went 14-2 last year?  Part of it is tougher schedules as these teams get the (in theory) toughest being the Super Bowl teams from last year, and part of it, particularly in Indy's case, is that they eked out a bunch of their wins last year and thus overachieved a bit and so they could easily have been a few games worse last year.  Look for another blog entry on schedule strengths soon.

But also, by looking at the table below, there is quite a good chance that New Orleans will go 12-4 (20.8%) and even a 13 win season has a 14% chance so still very well within the realm of possibility.  An undefeated season you ask?  That comes in at only 0.23%.

Will the records below happen as is?  Highly unlikely, but it is still fun to look at them as a baseline.  If you think your team will improve more than the rest of the league (or their schedule), then add a game or two.  If not, perhaps subtract a game or two.

And while the records are interesting, in the end what matters is who makes the playoffs.  So I've put the AFC teams in red and NFC teams in blue with the division winners also in bold.  Where the record is the same, I break the tie by giving it to the team with the higher chance of finishing one game better.

If things play out this way, the week 17 games that could be especially meaningful are:

  • Carolina at Atlanta - Both projected at 9-7 but only one likely to make the playoffs
  • Dallas at Philly - Dallas projected at 10-6 against Philly's 9-7, and Philly might not get in at 9-7 so could be a swing game
  • Arizona at San Francisco - Both projected at 9-7 and one likely the division winner, but that team also possibly so only one to get in the playoffs

What do you think?  Will your team do better?  Worse?  Leave a comment!

TeamRecord% chance% +1% -1
TeamRecord% chance% +1% -1
New Orleans11-522.120.817.4
San Diego11-521.116.420.2
NY Jets10-620.718.118.2
Green Bay10-621.317.519.6
New England10-620.316.519.2
San Francisco9-720.916.919.7
NY Giants8-820.416.319.6
Tampa Bay7-921.416.820.5
Kansas City6-1020.519.516.6
St Louis4-1222.720.518.2

Tuesday, April 6, 2010

NCAA Tournament Final System and Poll Comparison

And Duke wins it all.  And with no poll or system picking them, there aren't any big scores for any system/poll, but I am pleased to see that my computer eked out the best score beating Sagarin by 2 points.

System/Poll Points
Schmidt Computer89

The difference between my computer and Sagarin ultimately was my computer having Butler getting to the sweet 16 and Sagarin having them losing in the second round.  The big difference against the rest was the my computer and Sagarin both had Baylor getting to the Elite Eight (AP, Coaches, Seeds, Vegas had Villanova) and Sagarin and my computer had Duke getting to the final.

I'd like to think this proves my computer did a better job ranking and ratings teams than the polls or Vegas, but it is just one year.  But I'll do this again next year and share it all again.