Tuesday, August 31, 2010


In an effort to expand into the DC area metro audience, here's a look at Georgetown's projection.
Last year they were the 389th best college football team in the nation (at 0-11, not very good for a school it's size). The Hoyas have 14 returning starters, didn't play in a playoff game, and didn't have a player lost in the NFL draft. However, I would caution that all the trends were set for teams in the top 50, not at 389th.
Still there's a projected 35% improvement, moving G-town up to 356th in the nation. More importantly, with the improvement, assuming a similar schedule, Georgetown projects to have 1.84 wins. 2-9? Definitely better than 0-11. Hoya Saxa!

Pre-Season College Football Top 25!

In preparation for the 2010 football season, I've gotten caught up in the pre-season polls and started wondering how I might do that myself. After a couple of different attempts, I've finally settled on 4 roughly independent factors that can change a team's performance from season to season: Offensive returning starters, Defensive returning starters, Bowl game result from the previous season, and loss of star talent (I counted two points for a player chosen in the first round of the NFL draft, one for a player in the second). After studying how these variables related to changes in ratings last year, I took the weights and applied them to changes this year. Here's one example:
By my rankings Northwestern last year was the 70th best football team in the nation, with a power ranking around 51 (yes, their record was better than that, see the article about luck from last week). For this year they have 8 returning starters on O and 6 on D. 6.5 is roughly the average for each, so they'll probably improve a little because of the returning players. They lost their bowl game last year, which actually means they'll likely improve, and they had no players taken in the first two rounds. Overall, Northwestern will likely be about 1.6 times better this year than last year, with a power ranking around 80, making them about the 55th best team in the nation.
What? Bowl losses mean improvement? That surprised me too, but it was a definite trend at least last year. One possible explanation is that teams that win become too satisfied, and don't work hard enough during the off-season. It also may be an auto-correction for including bowl game results in the power ranking.
And now, the debut of my Pre-Season top 25:
2 Oregon
3 Texas
4 Boise St.
5 Ohio State
6 Virginia Tech
7 Miami FL
8 North Carolina
9 Nebraska
10 Alabama
11 Arkansas
12 Stanford
13 Texas Tech
14 Cincinnati
15 LSU
16 Georgia Tech
17 Georgia
18 Wisconsin
19 Washington
20 Oregon St.
21 Penn State
22 South Carolina
23 West Virginia
24 Texas A&M
25 Clemson

Some interesting things to note: Three teams that are projected to make big jumps are Stanford(!), North Carolina, and Stan Kamande's Washington Huskies. All three of these teams are actually expected to be better this year than last year (just maybe not as much better in Washington's case). Non-BCS schools TCU and Boise St. are highly ranked, but they both have a lot of returning starters from pretty good teams (each actually lost some star talent in the draft though). Also, Alabama is expected to drop from other people's preseason polls (They have only 1 returning defensive starter!).
The most astute observer might note two highly ranked teams are missing: the Florida Gators and the Iowa Hawkeyes. Florida amazingly drops to 26th in my projections, easily the biggest fall, but it kinda makes sense. They only return 9 starters, second lowest of the top 50 teams, they won their bowl game, and they lost the most star talent (3 1st rounders, 3 2nd rounders). Everything is working against them. The same thing happened last year to USC, and they dropped from 2nd to 24th. So Florida should maybe be a little worried.
What about my Iowa Hawkeyes? They are actually only expected to get a little worse this year (they lost some star talent, and won their bowl game), but because they played so many close games last year my ratings system only had them at 22nd in the nation. A preponderance of teams bunched around 30 meant that the small projected drop of the Hawks moved them from 22nd to 36th in the ratings.
Is the projection system perfect? Absolutely not, Florida and Iowa will still probably be top teams this fall. But it does provide some food for thought: maybe these pre season polls are not as accurate as we imagine.

Friday, August 27, 2010

Top NBA players of the last 10 years

A few days ago I posted the top 5 NBA players last year, according to an average of various measures. What happens over the last decade? Here are the last ten years, ties broken first by eliminating PPG, then by counting who appeared in the most categories (with adjustments to roland rating, switched to adjusted +- for 02-03 to 04-05 and dropped before that because neither existed):

2009-10 2008-09 2007-08 2006-07 2005-06
1 James James James Nowitzki James
2 Durant Wade Paul Duncan Bryant
3 Wade Bryant Bryant Bryant Nowitzki
4 Howard Paul Garnett James Nash
5 Bryant Howard Stoudemire Nash Wade

2004-05 2003-04 2002-03 2001-02 2000-01
1 Nowitzki Garnett Duncan Duncan O'Neal
2 Duncan Duncan McGrady O'Neal Iverson
3 O'Neal Stojakovic Garnett McGrady Duncan
4 Garnett McGrady Bryant Kidd Webber
5 Nash Bryant O'Neal Nowitzki Carter

A couple interesting things to note: First, it splits up into the Duncan first half and the Lebron second half (and it surprises me at least how early Lebron was the best). Also interesting is that Dirk Nowitzki had a three year stretch where he may have been the best player in the NBA (robbed of a championship only by some curious 06 officiating, and a sudden collapse in 07). Finally, it's amazing to note that Kobe was not once the best player in the NBA. He's been fairly steady top 3 recently, which is really good, but no where near MJ territory (who was the best player in 10 straight seasons he played, but we'll get to that). So that argument needs to be put to rest.
What does this mean for the decade? Who's the best player?
Giving a player 5 points for first, 4 for second, and so on, here's the top 8:

Rank Player Points
1 Duncan 25
2 Lebron 22
3 Bryant 17
4 Dirk 14
5 O'Neal 13
6 Garnett 12
7 McGrady 9
8 Wade 8

Best player of the decade? Still Tim Duncan (there you go George). Yeah, Kobe's third. But the player who really stands out is Lebron James. At just 25 years old, he's already been the best player in the league 3 straight years. Who else has been the best three years running? Jordan of course, Kareem Abdul-Jabbar, Wilt Chamberlain, Larry Bird, and Karl Malone (end of the Jordan era, believe it or not). That means Lebron James has a serious shot at being the best basketball player of all time. And that I'm gonna get tickets when Miami visits the Bay Area.

Wednesday, August 25, 2010

Giants in the Playoffs?

For weeks now I've been waiting for the Giants to overtake the fluke Padres and to give me a chance to watch the MLB Playoffs in person. But as the season goes on, it's still not happening. Why? Turns out they've actually been a little unlucky, their run differential indicates they should have 76 not 75 wins. What are the chances that the Giants can overtake them?
Assuming pythagorean winning percentages (RS^2/(RA^2+RS^2) estimates winning percentage well, and predicts future success) play out for the rest of the season, it turns out that the Padres will probably win 7 games more than the Giants, and the Giants only have a 3.8% chance of winning the NL West.
Can I take Solace in the Wild Card? The Giants are tied for first place there, but the Cardinals who are a couple games back have been very unlucky this year. Doing a three team simulation between the Cardinals, Giants and Phillies the following Wild Card odds appear:
SFG: .337
PHI: .323
STL: .341
Turns out first place is only a slight advantage, it's actually a tight three way race. Which unfortunately means that I have at best a 37% chance of watching the Giants in the playoffs. At least it's better than the Cubs.

Tuesday, August 24, 2010

Luckiest Big Ten Football Team

The Iowa Hawkeyes were notorious last year for their multiple fourth quarter comebacks. Some people call it clutch, but in sports statistics we call it luck. This would make the Hawks a very lucky team. Last night though Alex Potter and Andy Meisner tried to tell me Northwestern (jNU) was the luckiest team in the Big Ten. So who is it really?

The best way to find a team's luck is to compare it's point differential to it's wins. If a team scores more points than it's opponents do, usually it wins most of it's games. The Big Ten has data for the conference season. By graphing both Big Ten wins and Big Ten point differential, then adding a trendline, we can find each point gained is about 1/30th of a win, and a point differential of 0 is worth roughly 4 wins. Then we can use the team's point differential to predict it's wins (Illinois scored 58 points less then it's Big Ten opponents, meaning it should've won 2.07 games. It won two, so that's pretty good.)

Just the Big Ten schedule takes out a couple of games though, so I also did the comparison for the whole schedule (called the simple rating). On top of that, I used last year's power rankings and opponent strength to predict wins and compared it (called the complex rating). The following table is the results (positive means a team was lucky, won more than it should've).

Big Ten Simple Complex
Ohio State -1.038 -1.426 -0.06
Iowa 0.497 2.039 2.02
Penn State -1.17 -1.487 0.124
Wisconsin -0.402 0.15 0.62
Northwestern 1.7 1.555 0.978
Michigan State 0.566 -1.211 -1.775
Purdue 0.8 0.013 -0.357
Minnesota 0.167 1.273 0.739
Illinois -0.065 -0.372 -1.152
Michigan -0.031 -1.09 -1.402
Indiana -1.032 0.627 -0.405

What does that mean? Yes, Northwestern won more games than it should've (probably just 1 more), so they were lucky. But the luckiest team in the Big Ten was definitely the Iowa Hawkeyes, who won about 2 games more than they should've. Which bodes poorly for the future because luck doesn't carry over years so the Hawks might not live up to the hype this year. Going the other way are the Michigan teams, both of which were very unlucky last year. Here's hoping the Hawks go with skill instead of luck this year.

NBA's Top Five

Who are the top five players in the NBA? There are multiple ways look at it. On voting, you could use the All NBA first team (5 players), or MVP vote shares. Statistically, you could use PER, Win Shares, Roland Rating, or even Points Per Game (weak, but the point of basketball is to score points). For explanation, Win Shares is an estimate of how many wins a player adds to his team (click here for more info) and Roland Rating is basically a hockey style plus minus rating (with adjustments for teammates, or for out preforming opponents, explained here).
What happens this year?

All NBA-Lebron, Kobe, Dwayne Wade, Dwight Howard, Kevin Durant
MVP votes-Lebron, Durant, Kobe, Howard, Wade
Win Shares-Lebron, Durant, Howard, Wade, Dirk Nowitzki
PER-Lebron, Wade, Durant, Chris Bosh, Tim Duncan
PPG-Durant, Lebron, Carmello Anthony, Wade, Kobe
Roland Rating-Lebron, Durant, Wade, Howard, Kobe

Say we give 5 points for first, 4 for second, and so on (All-NBA is 3 points each). For Durant, for example, he would get 3, 4, 4, 3, 5, and 4 points, giving him a total of 23 points. What are the final standings?
1. Lebron James - 27 points
2. Kevin Durant- 23 points
3. Dwayne Wade- 15 points
4. Dwight Howard- 10 points
5. Kobe Bryant- 8 points
Kinda makes sense right? Some might say Kobe should be higher (not in the regular season he wasn't), but it's a fairly accurate assessment of the NBA's top five players.