Thursday, September 30, 2010

Points Line Rankings System

You may notice that up top among the pages is now my Points Line Rankings System. While not as theoretically sound as my normal rankings system, it does have the advantage of predicting score. It's all explained under How It Works, but here's an example looking at Saturday's Iowa game.

Rank Team Total Off Def
11 Iowa 62.31 47.603 -14.707
21 Penn State 55.007 33.395 -21.612

Adding offensive rating to the other team's defensive rating gives expected points. For Iowa 47.6-21.6 means they should score about 26 points. Penn State should score about 19 points. To account for home field advantage (about 4 points) we add two points to the home team and take two from the away team. Since the games at Iowa, that means the predicted score is:

Iowa 28 Penn State 17

Here are a few other predicted scores for this weekend:

Stanford 33 Oregon 32
A high scoring battle for front runner rights in the Pac-10 race.

Northwestern 36 Minnesota 13
Northwestern is looking to prove that it really is a top 25 team. Minnesota is on the other end.

Alabama 29 Florida 19
Alabama looks to hold on to it's top ranking by making full use of the second best defense in the nation.

USC 45 Washington 26
Jake Locker may be able to put up points, but the Washington Huskies have less defense than Reggie Bush does Heisman Trophies.

Oklahoma 31 Texas 24
The Sooners look to make it back to back losses for the underachieving Longhorns.

Colgate 27 Georgetown 25
Although both rankings systems have Georgetown as the better team, the home field advantage puts Colgate over top. By both rankings. Yet another chance for the Hoyas to defy my predictions.
Reaction:

NFL Rankings 9-28-10

Three weeks into the NFL season I am debuting my NFL rankings. It's all posted under the page links at the top, but for the lazy here are the top 10.
Team W L Rating SOS
1 Pittsburgh 3 0 8.402 2.352
2 Atlanta 2 1 4.653 1.687
3 Green Bay 2 1 3.524 1.393
4 Tennessee 2 1 3.427 1.109
5 Chicago 3 0 2.969 1.658
6 Philadelphia 2 1 2.108 0.931
7 NY Jets 2 1 2.067 1.101
8 Kansas City 3 0 1.456 0.539
9 Indianapolis 2 1 1.454 0.587
10 New Orleans 2 1 1.356 1.122

What stands out about these rankings? First, they do a good job of noting that even though teams like Chicago and Kansas City are undefeated, they're not necessarily among the three best teams in the league.
Among the teams themselves there are a couple trends of note. First, the torch appears to be passing, as former great teams like the Colts and the Patriots (almost exactly average now) give way to new top tier teams such as the Falcons and the Packers. Manning and Brady are giving way to Ryan and Rodgers. While it's early, that may be a trend worth following.
It's also astounding to note the parity among NFL teams. Flipping a coin you'd expect 4 teams to be undefeated, but there's only three right now. There's also very few teams that have yet to win a game. Pittsburgh is significantly better than the average team right now, but other teams might just be lucky to be up there. Either way, don't expect too many teams to run away with their division this year.
Reaction:

Sunday, September 26, 2010

College Football top 25 9/26

After four weeks worth of games, the rankings are starting to take shape.

Rank Team W L Rating SOS
1 Stanford 4 0 1166.98 51.56
2 Ohio State 4 0 1008.73 46.01
3 Alabama 4 0 696.3 49.46
4 LSU 4 0 617.13 132.77
5 Arizona 4 0 594.87 71.66
6 Texas A&M 3 0 469.06 40.9
7 Miami FL 2 1 410.3 54.64
8 Iowa 3 1 396.35 38.94
9 TCU 4 0 389.2 37.08
10 Boise St 3 0 382.89 67.83
11 California 2 2 322.6 69.88
12 Virginia Tech 2 2 281.77 120.5
13 Florida 4 0 277.94 32.22
14 Oregon 4 0 277.87 14.49
15 Oklahoma St 3 0 256.03 25.75
16 Nevada 4 0 221.64 24.71
17 James Madison 3 0 215.5 37.02
18 Kansas St 4 0 210.65 83.24
19 Nebraska 4 0 210.1 16.67
20 South Carolina 3 1 208.71 62.38
21 Arkansas 3 1 198.81 38.04
22 Missouri 4 0 194.11 22.61
23 Michigan 4 0 193.42 40.23
24 Northwestern 4 0 191.12 42.95
25 Mississippi St 2 2 190.13 102.51

And a couple other teams of interest:
51 Texas 3 1 81.08 36.67
96 Washington 1 2 23.85 41.41
329 Georgetown DC 3 1 1.21 0.71

Stanford's still number 1?! After beating up on Notre Dame, it's starting to make sense. And pollsters are realizing it, as the Cardinal moved up to 9 in the AP poll.
Texas showed why they weren't ranked last week against UCLA. It may be a long week for the Longhorns.
Iowa is flying high again, despite the loss to Arizona last week, due largely to it's 3 other dominating wins.
Northwestern has broken into the top 25 for the first time this season. Their early success may continue into the Big Ten season.
Overall, this week just confirmed what we thought earlier. It'll be exciting to watch next week as conference play, and the upsets it brings, get in full swing.
Reaction:

Friday, September 24, 2010

Games of the 25th

I have a new alternative ratings system in the works, and while probably not being quite as good, it does have the advantage of predicting actual scores. While it's not ready to debut totally yet, here are some forecasts for tomorrow.

Stanford 50 @Notre Dame 25
Stanford has the best offense in the nation, and is gonna put it on display again tomorrow against a mediocre Notre Dame defense.

@Iowa 38 Ball State 10
Another very uneven battle between with an Iowa team eager to avenge it's loss.

Central Michigan 10 @Northwestern 9
Two good defenses could put on a show tomorrow.

Holy Cross 12 @Georgetown 0
Might be a rough homecoming for Georgetown.

Alabama 31 @ Arkansas 12
Alabama shows why it's in the national title hunt.

With just a few games to rely on, and just a few hours of work on the new ratings, these picks are kinda rough still. I don't believe the Northwestern one, or that Holy Cross's defense is that good. Still maybe the numbers know more than I do.

Reaction:

Thursday, September 23, 2010

Massey Comparison

Since the debut of my college football rankings I've been lucky enough to be included in Ken Massey's ratings comparison site. Unfortunately my stats aren't so good, but that's largely due to being a predictive system rather than one that just rewards wins and losses (that ranking of mine will debut later on this year). Still, it's pretty cool to see my name near those of Sagarin and Massey (not to mention the AP poll).
In honor of being included I have added some links near the top of the blog to various ratings pages. These will make looking up rankings easier, and include a new explanation of how they work. Enjoy!
Reaction:

500 Home Runs

ESPN the Magazine ran an article recently on players shooting for certain milestones, and the ability to tell statistically. They covered 300 wins, and 12,000 yards (this is a milestone?) but they missed the biggest milestone in all of sports: 500 Home Runs.
Can it be shown statistically that MLB players hang on to get 500 home runs? Lets look at the breakdown for retired players home run totals, grouped by hundreds:

Reached Actual Expected
700 3 2.1
600 3 4.9
500 16 11.3
400 16 26.3
300 69 61.1
200 168 141.9
While 500 has more than expected (Calculated with exponential trend lines), and 400 in noticeably less, assuming Poisson distribution neither is statistically significant.
Breaking it down into bins of 25 though gets interesting:

Reached Actual Expected
600 1 1.42
575 2 1.76
550 3 2.19
525 3 2.72
500 8 3.37
475 4 4.18
450 3 5.19
425 7 6.44
400 2 8
500 home runs is much higher than expected, and strangely 400 home runs in much lower. Both are now statistically significant at the 95% level. What does this mean? It appears that some players do purposefully try to reach 500 home runs (You think Eddie Murray plays to 41 without that?). Oddly, it also appears that there's a barrier at 400, nobody wants to just barely make 400 home runs. I suppose we may get a few more years than expected out of Carlos Delgado (473) and Vlad Guerrero (434).
Reaction:

Sunday, September 19, 2010

College Football Top 25 9/19

After three games, it's time to break out the first top 25 of the year. It's not great yet, but it's surprisingly accurate for just 3 games. Also, I swear this is not rigged, the math actually gives our number 1.







Rank Team W L Rating SOS
1 Stanford 3 0 2747.92 86.74
2 Alabama 3 0 2105.8 69.92
3 Ohio State 3 0 1026.06 61.19
4 LSU 3 0 745.57 117.27
5 South Carolina 3 0 717.6 115.75
6 Texas A&M 3 0 598.07 52.15
7 Nebraska 3 0 522.71 31.37
8 Boise St 2 0 454.84 57.69
9 Arizona 3 0 449.15 27.78
10 TCU 3 0 416.63 28.64
11 Miami FL 1 1 365.66 85.49
12 Penn State 2 1 351.73 119.16
13 Oklahoma St 3 0 350.45 35.24
14 Fresno St 2 0 338.65 87.82
15 James Madison 2 0 335.87 37.68
16 Oregon 3 0 310.78 7.99
17 Virginia Tech 1 2 262.8 190.02
18 Arkansas 3 0 259.51 25.35
19 Georgia 1 2 256.15 129.46
20 Temple 3 0 255.19 122.36
21 Michigan 3 0 237.37 96.71
22 Oklahoma 3 0 210.47 66.82
23 Kansas St 3 0 184.55 61.41
24 Florida Int'l 0 2 182.6 320.49
25 North Carolina 0 2 182.51 302.88

Stanford?! Really?! Stanford!! Watch the replay of what they did to Wake Forest tonight, or UCLA last week. Both solid teams, both destroyed. Also up top as expected are Alabama and Ohio State among others. Amazingly Virginia Tech is still at 17, so it's possible expectations weren't too wrong.
Where to other teams relevant to this blog place?

33 Iowa 2 1 153.29 25.43
41 Northwestern 3 0 114.88 18.04
107 Washington 1 2 22 38.21
442 Georgetown DC 2 1 0.42 0.3

A rough loss for Iowa tonight against Arizona knocked them out of the top 25. Northwestern is 3-0, but is gets burned by a weak early schedule, and a close Vandy game. Georgetown has already improved on last year's win total, but with a ranking lower than last season more wins might not come easily.
All this goes to prove two things. First, even after only 3 weeks the rankings are surprisingly good. Second, Stanford may be a team to watch for more than just a Pac-10 title this year.

Reaction:

Friday, September 17, 2010

All Time Top 12

A couple weeks ago I posted my top 5 players of the past decade (and made a sporcle quiz!). I've since gone back all the way through 1955 (when data becomes scarce) with results that will get posted at some point. I began wondering if I could use the numbers to create a list of the top players ever to play. Just adding up totals like in the players of the decade post gave an unfair advantage to long careers like Karl Malone and Kareem. So I combined total top 5 points and average position in years ranked (I had a minimum of 5 years, so Kevin Durant who's only been ranked for 1 year has an average of .8). The rankings aren't great, but they bring up interesting ideas.

1 Jordan
2 Abdul-Jabbar
3 Chamberlain
4 Robertson
T5 K. Malone
T5 O'Neal
7 James
8 Pettit
9 Robinson
10 Duncan
11 Bird
12 West

There's also a top 12 using just years being the best in the NBA

Player Years Consecutive Ties
1 Jordan 10 7 2
2 Kareem 7 4 0
3 Chamberlain 7 3 0
4 James 4 3 0
5 Bird 3 3 0
6 K. Malone 3 3 2
T7 M. Malone 3 2 0
T7 Pettit 3 2 0
T9 Robinson 2 2 0
T9 Shaq 2 2 0
11 Duncan 2 2 1
12 Nowitzki 2 1 0

There are a few interesting things to notice here:
First, it's clear the Michael Jordan is the best player of all time, followed by Kareem Abdul-Jabbar then Wilt Chamberlain. This is a very believable thing without numbers to back it up, and it's solidified here.
Second, Bill Russell and Magic Johnson are missing from both lists. Each suffers from playing in a loaded era (Russell was always stuck behind Robertson and Chamberlain, Magic had Bird and Kareem/Jordan). Both probably also suffer from a lack of plus/minus data for their years because they set a unique style of play for their team. What their absence does indicate though is that both may be over-rated. While each was definitely a top-20 player top 5 may be a little bit of a reach.
Finally, Lebron James is ranked impressively high on both lists, 4th and 7th. It puts him in serious danger of challenging Karl Malone for best player ever to not win a championship ( something which could change this year). Just halfway through his career this is an amazing achievement. He had the highest average score of any player not named Jordan, and is on his way to possibly dethroning our current top 3. It will be an interesting next few years to watch.

Also for funsies, here's where a couple other top-40 players rank in the first ranking:
Tie 16th- Garnett and Nowitzki
18th - Magic
19th - Kobe
Tie 27 - T-Mac
29 - Bill Russell
30 - D-Wade
Tie 34 - Chris Paul
Reaction: