Monday, December 13, 2010

Ratings Contest

A website called Kaggle have regular contests where they give people a data set, and have them to predict something about the data set. They have done it for a variety of topics (e.g., tourism prediction, grant application success) and, of course, they just finished one for chess ratings.

Spurred on by Jeff Sonas (the Chessmetrics guy), there were a lot of entries, and suffice it to say, someone came up with a method that is 6.6% better than standard Elo ratings.

You can read all about it here, and wait for the follow up contest in January.

Thursday, October 28, 2010

USCL Power Rankings: Final Week

The final week of the United States Chess League regular season has come and gone, and the playoffs are set. The power rankings only are for the regular season, so the rankings below are the final ones for this year.

The last week really shook up the rankings, with St. Louis plummeting four places and other teams jockeying for a better position. In the end, the newcomers, the New England Nor'easters are far above the remainder of the field. Interestingly the bottom three places of the rankings are filled by Eastern Division Teams.

Before the 2010 list, let's look at where teams stood in the 2009 list.

Power Rankings 2009 (End of Regular Season)
1 - New Jersey (8.0 - 1st East)
1.000
2 - Seattle (7.5 - 1st West)
0.953
3 - Boston (7.5 - 2nd East)
0.929
4 - Arizona (6.0 - 3rd West)
0.901
5 - San Francisco (6.5 - 2nd West)
0.894
6 - Miami (6.0 - 4th West)
0.879
7 - New York (5.0 - 3rd East)
0.853
8 - Chicago (4.0 - 5th West)
0.764
9 - Philadelphia (4.0 - 5th East)
0.735
10 - Dallas (3.5 - 6th West)
0.730
11 - Queens (3.5 - 6th East)
0.726
12 - Baltimore (4.5 - 4th East)
0.724
13 - Tennessee (2.0 - 7th West)
0.647
14 - Carolina (2.0 - 7th East)
0.622

How some teams fortunes have changed in a year, especially New Jersey which went from the top in 2009 to near the bottom this year.

Therefore, here is the final power rankings for the 2010 season.

2010 USCL Power Rankings
End of Regular Season


Rank Team Score Change
1 - New England 1.000 0
2 - Boston 0.886 +1
3 - Arizona 0.857 -1
4 - Chicago 0.793 +2
5 - New York 0.793 -1
6 - Baltimore 0.787 +1
7 - Miami 0.771 +1
8 - Seattle 0.760 +2
9 - St. Louis 0.743 -4
10 - Los Angeles 0.740 +3
11 - San Francisco 0.708 -2
12 - Philadelphia 0.697 +2
13 - Dallas 0.686 -2
14 - New Jersey 0.677 -2
15 - Manhattan 0.638 0
16 - Carolina 0.607 0

The top eight teams in the power rankings are also the top eight teams in the playoffs. Good luck to all of them.

Tuesday, October 26, 2010

USCL 2010 Power Rankings Week 9

A little late... okay a LOT late, but I was away.

Notable about this is that New England has completely pulled away from the pack.

Rank Team Score Change
1 - New England 1.000 0
2 - Arizona 0.886 0
3 - Boston 0.857 0
4 - New York 0.793 0
5 - St. Louis 0.793 +2
6 - Chicago 0.787 0
7 - Baltimore 0.771 -2
8 - Miami 0.760 +1
9 - San Francisco 0.743 -1
10 - Seattle 0.740 +1
11 - Dallas 0.708 -1
12 - New Jersey 0.697 +2
13 - Los Angeles 0.686 -1
14 - Philadelphia 0.677 +1
15 - Manhattan 0.638 -2
16 - Carolina 0.607 0

Thursday, October 14, 2010

USCL 2010 Power Rankings Week 8

The power rankings show that New England is clearly the class of the league, with a huge lead in strength over second place Arizona. Baltimore and Dallas were the big gainers this week, while San Francisco took a big tumble.

Only a whisker separates Boston and New York, and when these teams likely play in the playoffs in the first round, it should be a close match.

Not even visible to the human eye (only the fourth decimal place) separates Chicago and St. Louis!

Rank Team Score Change
1 - New England 1.000 0
2 - Arizona 0.914 0
3 - Boston 0.846 0
4 - New York 0.844 0
5 - Baltimore 0.810 +2
6 - Chicago 0.799 0
7 - St. Louis 0.799 +1
8 - San Francisco 0.794 -3
9 - Miami 0.769 0
10 - Dallas 0.728 +3
11 - Seattle 0.703 -1
12 - Los Angeles 0.692 -1
13 - Manhattan 0.679 -1
14 - New Jersey 0.658 +1
15 - Philadelphia 0.651 -1
16 - Carolina 0.632 0

Remainder of Season Simulation: Post-Week 8

Chance to be at least tied for 4th in match/game points, based on 100,000 simulations.

Western Division
100.00% Arizona
_73.15% St. Louis
_72.37% Chicago
_63.15% San Francisco
_51.96% Miami
_26.43% Dallas
_13.62% Seattle
__4.72% Los Angeles

Eastern Division
100.00% New England
_99.65% Boston
_94.25% New York
_92.55% Baltimore
_13.47% Manhattan
__0.78% New Jersey


Monday, October 11, 2010

USCL Simulation after Week 8 Monday

UPDATED! Now includes Western Division!

Chance of making playoffs (100,000 simulated Wednesday Week 8s and Weeks 9 and 10)

Eastern Division
New England 100.00%
Boston 98.85%
New York 97.44%
Baltimore 92.29%
Manhattan 11.27%
New Jersey 0.79%
Philadelphia 0.00%
Carolina 0.14%


Western Division
Arizona 99.99%
Chicago 85.45%
San Francisco 69.28%
St. Louis 35.04%
Seattle 30.48%
Miami 18.73%
Los Angeles 16.68%
Dallas 49.12%

Note that Arizona made the playoffs in 99,991 of the 100,000 simulations!




















Sunday, October 10, 2010

Remainder of Season Simulation

I have expanded on my prior projects to determine the likelihood that a team will make the playoffs this season. As I said previously, the ability to be comprehensive in all possibilities of match outcomes eats into years of computing time, so simulation is an obvious answer.

I created a simulation that attempts to model all the matches in the remainder of the USCL season.

The goal was the same -- what is the chance that a team will make the playoffs. This time, however, I included by match points and the game points tiebreaker. (If a team was tied for a playoff spot with the same match and game points, I said they both would make the playoffs -- I didn't go beyond the first tiebreaker.)

To do this, I made several other assumptions.
  1. All teams have the same strength. Therefore, they have the same chances in each match against their opponents.
  2. The game point results of the match follow the historical distribution of USCL matches (which is here). That is, 2.5-1.5 is more common than 3-1, which is more common than 3.5-0.5, etc.
I ran 100,000 simulations for both the Eastern and Western division. Here are the results, where the percentages indicate a tie for fourth place or better.

Eastern Division (Chance to make playoffs)

New England 100.00%
Boston 98.43%
New York 97.49%
Baltimore 70.27%
Manhattan 24.79%
New Jersey 6.92%
Philadelphia 2.69%
Carolina 0.61%

Western Division (Chance to make playoffs)

Arizona 97.18%
Chicago 86.54%
San Francisco 78.07%
St. Louis 45.24%
Seattle 35.50%
Miami 27.39%
Los Angeles 23.44%
Dallas 12.92%


I also included the chance that a team will receive draw odds in the first round of the playoffs (tied for second place or better).

Eastern Division (Chance have draw odds first round of playoffs)

New England 98.37%
Boston 55.19%
New York 41.63%
Baltimore 4.90%
Manhattan 0.23%
New Jersey 0.00%
Philadelphia 0.00%
Carolina 0.00%

Western Division (Chance have draw odds first round of playoffs)

Arizona 80.94%
Chicago 50.92%
San Francisco 40.50%
St. Louis 13.13%
Seattle 7.55%
Miami 6.01%
Los Angeles 3.60%
Dallas 1.25%

Any scenarios you want me to run? Things like: If NJ wins the remainder of their matches 4-0, what's their chance to get in the playoffs... things like that....

Let me know in the comments...

Friday, October 08, 2010

USCL Western Division: Playoff Hunt

In my previous post, I posted the chances of at least tying for fourth place in the USCL Eastern Division. You can read about the method I used in that post.

I ran the same thing for the Western Division. Here is the percentage that each team will finish in at least a tie for fourth in match points.

99.14% Arizona
95.70% Chicago
82.20% San Francisco
61.70% St. Louis
35.96% Seattle
35.69% Miami
34.61% Los Angeles
15.31% Dallas

USCL Eastern Division: Playoff Hunt

There are three weeks to go in the USCL regular season. What are the chances that any particular team will be in playoff contention?

Let's assume for the moment that all teams are equally matched, and that all teams have an equal chance to win, draw, or lose a match with every other team. Therefore, each match can have three possible results.

In the Eastern Division of the USCL, there are twelve matches left and therefore there are 3^12 (three to the twelfth power) possible results (in terms of match results). That's 531, 441 possibilities.

If we make the criterion for "possibly in the playoffs" to be tied for fourth or better, then the percentages below indicate what a team's chances to possibly be in the playoffs.

100.00% - New England
99.55% - Boston
98.27% - New York
74.75% - Baltimore
41.28% - Manhattan
10.32% - New Jersey
3.05% - Philadelphia
1.21% - Carolina

Now, if we take game points (the first tiebreaker) into consideration, that increases the number of possibilities to 9^12 power, which is actually 531,441 times bigger than the last time. Since the simulation presented here took about 4 seconds to run on my computer, the code for including game points is about 50% more complex (increasing time to completion by about 50%), it would take over 553 days for the simulation to run!

I think I'll wait for the actual results.

Anyone interested in me doing this for the Western Division too?

Thursday, October 07, 2010

USCL 2010 Power Rankings Week 7

Only three weeks left in the season, and the races are tightening.

New England keeps the top spot in the rankings, while Arizona jumps four places to number two. Miami also makes a big jump up to number nine. Several pairs of teams are almost indistinguishably close in their score. San Francisco and Chicago are five and six; Miami and Seattle are nine and ten; Manhattan, Dallas, and Philadelphia are twelve, thirteen, and fourteen.

Will New England keep their top spot in the rankings at the end of the regular season? Only time will tell.

Rank Team Score Change
1 - New England 1.000 0
2 - Arizona 0.912 +4
3 - Boston 0.889 -1
4 - New York 0.878 -1
5 - San Francisco 0.849 0
6 - Chicago 0.845 -2
7 - Baltimore 0.803 0
8 - St. Louis 0.771 +2
9 - Miami 0.756 +4
10 - Seattle 0.755 -2
11 - Los Angeles 0.739 -2
12 - Manhattan 0.686 +3
13 - Dallas 0.685 -2
14 - Philadelphia 0.683 -2
15 - New Jersey 0.668 -1
16 - Carolina 0.607 0

Wednesday, September 29, 2010

USCL Power Rankings Week 6

Because it was interdivisional week, we are able to combine the Power Rankings into a complete list. There are some interesting things to note.
  • The top three and the bottom three teams are all from the Eastern division.
  • Chicago, San Francisco, and Arizona are all very tightly bunched for 4th, 5th, and 6th.
  • Dallas, Philadelphia, and Miami are extremely tightly bunched as well.
Rank Team Score
1 - New England 1.000
2 - Boston 0.927
3 - New York 0.913
4 - Chicago 0.905
5 - San Francisco 0.901
6 - Arizona 0.894
7 - Baltimore 0.856
8 - Seattle 0.766
9 - Los Angeles 0.739
10 - St. Louis 0.727
11 - Dallas 0.718
12 - Philadelphia 0.712
13 - Miami 0.710
14 - New Jersey 0.643
15 - Manhattan 0.628
16 - Carolina 0.562

Can New England hold the top spot next week? Will any of the Western teams break away from the pack?

Only time will tell...

Thursday, September 23, 2010

USCL 2010 Power Rankings Week 5

Only changes at this top this week in the United States Chess League Power Rankings.

Despite that they are in second place in the match standings, the New York Knights have passed the New England Nor'Easters, and now are atop the Eastern Division Power Rankings. The Knights' game points were key in this.

In the Western Division, an surprising tie occurs at the top of the standings. Arizona and San Francisco share the top spot. Both have played the same five opponents, and they have identical match and game points.

Next week is the inter-divisional week, and afterwards, the power rankings will be combined into a single list.

Note to the commissioner! The "color choice" for the USCL finals ought to be decided by the records of the divisions in the inter-divisional matches. If the Eastern Division wins more matches than the Western Division, the Eastern Division champ should get color choice in the USCL finals, and vice versa. Game points would be the next tiebreaker. What do you think, Mr. Commish?


Eastern Division Change
1 - New York 1.000 up 1
2 - New England 0.974 dn 1
3 - Boston 0.897 n/c
4 - Baltimore 0.809 n/c
5 - Philadelphia 0.738 n/c
6 - New Jersey 0.702 n/c
7 - Manhattan 0.621 n/c
8 - Carolina 0.579 n/c





Western Division Change
1 - tie Arizona 1.000 n/c
1 - tie San Francisco 1.000 up 1
3 - Chicago 0.962 n/c
4 - Los Angeles 0.814 n/c
5 - St. Louis 0.790 n/c
6 - Seattle 0.748 n/c
7 - Dallas 0.739 n/c
8 - Miami 0.668 n/c

Wednesday, September 15, 2010

USCL 2010 Power Rankings Week 4

Week 4 of the power ratings of the United States Chess League.

In the East, New York is only a hair behind New England for the top spot. Boston overtakes Baltimore, and Philadelphia overtakes New Jersey.

In the West, Arizona is clearly in the lead. San Francisco overtakes Chicago, while Los Angeles makes the biggest jump, vaulting over both Saint Louis and Seattle.


Eastern Division
1 - New England 1.000
2 - New York 0.998
3 - Boston 0.895
4 - Baltimore 0.881
5 - Philadelphia 0.790
6 - New Jersey 0.749
7 - Manhattan 0.624
8 - Carolina 0.579




Western Division
1 - Arizona 1.000
2 - San Francisco 0.895
3 - Chicago 0.814
4 - Los Angeles 0.783
5 - St. Louis 0.718
6 - Seattle 0.688
7 - Dallas 0.674
8 - Miami 0.596

Until next week!

Saturday, September 11, 2010

USCL 2010 Power Rankings Week 3

Power rankings for this week reflect pretty closely the standings for each division. That's not too surprising...

What to notice, however, is the closeness of some of the teams in the rankings.

In the east, New York, Baltimore, and Boston are about equidistant from each other, while New Jersey and Philadelphia are close to one another.

In the west, Chicago is only a hair behind Arizona, despite being a 1/2 point behind in the match standings.


Eastern Division
1 - New England 1.000
2 - New York 0.922
3 - Baltimore 0.883
4 - Boston 0.835
5 - New Jersey 0.739
6 - Philadelphia 0.720
7 - Manhattan 0.618
8 - Carolina 0.522




Western Division
1 - Arizona 1.000
2 - Chicago 0.990
3 - San Francisco 0.902
4 - St. Louis 0.761
5 - Seattle 0.747
6 - Los Angeles 0.685
7 - Dallas 0.660
8 - Miami 0.556

Soon, there will be the return of the USCL rating list, with an updated algorithm to calculate more precise ratings!

Wednesday, September 01, 2010

USCL 2010 Power Ratings Week 2

Power rankings for week 2. Remember, these are done in each division, but after the interdivisional week (week 6), they will be combined into one list.

Highest score a team can get is 1.000, and the difference between teams' power is reflected in their score.

Eastern Division
1 - New England 1.000
2 - Baltimore 0.867
3 - New York 0.842
4 - Philadelphia 0.788
5 - New Jersey 0.783
6 - Boston 0.729
7 - Manhattan 0.703
8 - Carolina 0.571

Western Division
1 - San Francisco 1.000
2 - Arizona 0.875
3 - Dallas 0.757
4 - Chicago 0.749
5 - Los Angeles 0.674
6 - Miami 0.606
7 - St. Louis 0.589
8 - Seattle 0.538

Thursday, August 26, 2010

USCL 2010 Power Ratings Week 1

Here are the inaugural power rankings for the USCL, Week 1, 2010. Because of the lack of inter-divisional play this year, they will be in two divisions, until after week 6, where they will be combined into a single list.

Note, after only one week, these power rankings mean close to nothing, but I suppose they are fun anyway. Ties were broken by where the team's stood at the end of Week 10 last season.

Week 2 should have more meaningful results, and by Week 4, they should be pretty close to reality.

Eastern Division
1 - New York 1.000
2 - Philadelphia 0.942
3 - Manhattan 0.913
4 - New England 0.913
5 - Carolina 0.625
6 - Baltimore 0.625
7 - New Jersey 0.596
8 - Boston 0.538

Western Division
1 - San Francisco 1.000
2 - Arizona 0.941
3 - Chicago 0.941
4 - Los Angeles 0.941
5 - Seattle 0.644
6 - Miami 0.644
7 - St. Louis 0.634
8 - Dallas 0.584

Until next week!