The computer's view of the 3 upsets today

The first big weekend that had some of the top teams playing each other resulted in some important results.

Computer #4 Georgia visited #8 South Carolina with South Carolina favored by both Vegas and the computer.  The computer had the winner right, but it was by far more than expected and South Carolina should move up.

Computer #1 Florida hosted #14 LSU and LSU was favored in the game.  The computer has not been impressed with LSU and picked Florida big and they won, but not by quite as much as picked.  Nevertheless, the computer got the upset right and pick against the spread.  Because they won by less than expected though, Florida could lose a few ratings points, but it depends on all the other games too of course.

Computer #10 Florida State state visited #56 NC State and was favored, but the computer has also not been impressed with FSU having them only #10 while the polls have them #3.  So the computer did not pick the upset, but did pick Clemson against the spread and got that right.

Interesting College Football Week 6 Upset Picks - Florida, TTU, Syracuse, Purdue, Northwestern

The computer has several teams rated a bit differently than the polls and apparently different than Vegas as there are several key games where the computer is picking the upset.

LSU visits Florida, and frankly LSU has not impressed in some of their games while my computer has Florida #1.  But Vegas has LSU as the favorite.  The computer says pick the upset.

Oklahoma visits Texas Tech as a road favorite, OU having a loss but being the only one ranked in the polls.  But the computer really likes Texas Tech to win.

Syracuse hosts Pitt in a similar scenario to the above, although the computer's opinion on Syracuse pulling the upset is not quite as strong.

Similarly, Purdue hosts a disappointing Michigan squad as the underdog but the computer likes Purdue to win.

Staying in the B1G, Northwestern goes on the road to Penn State as an underdog but the computer likes them to win.

These 5 games will each go a long way to determining what happens in their respective conferences and have the traditional or recently better team favored, but the computer, looking at just the data, likes the underdogs.  Will tradition prevail or is the computer right?

College Football Week 5 BCS Algorithm Ratings and Rankings - A first look at a BCS compatible algorithm

I post my regular ratings starting with week 1 of the season, but also calculate a set of ratings using a BCS compatible algorithm.  The latter does not use margin of victory, which makes it a far less accurate predictor of games, and also makes the ratings frankly pretty silly looking the first few weeks of the year, so I haven't posted them until this week.

The top-25 is below, and you will see that Florida is #1 here as they are in the regular ratings.  LSU makes a big move to #2 and they are followed by K-State, Georgia, and Oregon State.

Where is Alabama?  They are down at #9.  Understanding why this is can be educational as to why the BCS handicaps the computers.

Alabama has played a 2-loss Michigan, a 2-loss Mississippi, 4-loss Arkansas and Florida Atlantic, and a 1-loss Western Kentucky.  Because the computer can't take into account the margin of victory, while undefeated, the schedule isn't that strong and no single team is very strong.

LSU on the other hand just made a big leap and is rated higher because they beat a 1-loss Washington who also beat a 1-loss Stanford who beat a 1-loss Southern Cal.  So even though their average schedule is weaker, they have a very good quality win.

And Florida just has a much stronger average schedule.

What about Oregon?  They are way down at #13 and haven't played a team with fewer than 2 losses.

Most of the computers the BCS uses don't publish ratings early in the year because of the challenges with rating a team properly when you don't use the score, but also because most of the algorithms haven't been tuned to do so.  The result is some pretty fishy looking rankings.

Kenneth Massey does publish his and presently has a top-3 of Oregon State, Notre Dame, and Texas Tech.  Alabama is #8, Oregon #16, and LSU #13.  Not completely far fetched, but also likely not that accurate.  To be fair, Ken does have an algorithm that does use the score and it has Alabama #1, Oregon #4, and LSU #6.

Billingsley also publishes, but he also assumed starting rankings from the end of last year which would seem to be a violation of the rules for a BCS computer.  This is how he manages to have reasonable results.  Of course, he still has Michigan in his top-20 (#19) and Arkansas in the top-50!

Colley has rankings published and similar to Massey has Oregon State and Notre Dame at the top.  Jeff Sagarin also publishes and has the most reasonable results.

Rank	Team	Rating	Record	Schedule	Change
Rank	Team	Rating	Record	Schedule	Change
1	Florida	83.036	4-0	67.454	+0, -3.685
2	LSU	81.692	5-0	59.142	+8, +2.887
3	Kansas St	81.491	4-0	58.859	-1, -2.570
4	Georgia	80.949	5-0	61.141	+1, -0.981
5	Oregon St	79.623	3-0	71.840	+14, +3.534
6	Notre Dame	79.521	4-0	66.180	+0, -2.399
7	Texas Tech	79.076	4-0	62.836	+9, +2.333
8	Florida St	78.634	5-0	55.935	-5, -4.855
9	Alabama	78.249	5-0	62.625	+4, +0.341
10	South Carolina	77.874	5-0	65.503	-1, -1.941
11	Texas	77.385	4-0	64.832	-4, -4.534
12	Washington	76.486	3-1	64.891	+33, +6.256
13	Oregon	75.643	5-0	56.819	-9, -7.198
14	West Virginia	75.546	4-0	60.318	+15, +2.378
15	Ohio State	75.471	5-0	61.891	+26, +3.904
16	Stanford	75.351	3-1	69.083	-8, -5.113
17	Mississippi St	74.773	4-0	55.802	+8, +0.910
18	Texas A&M	74.754	3-1	63.281	-7, -3.588
19	Clemson	74.563	4-1	65.003	-5, -2.618
20	Iowa St	73.663	3-1	65.960	+2, -1.191
21	Rutgers	73.548	4-0	60.380	+18, +1.832
22	Missouri	73.080	3-2	69.552	+1, -1.147
23	Oklahoma	72.952	2-1	61.389	-3, -2.914
24	Purdue	72.641	3-1	65.204	+3, -0.746
25	Tennessee	72.460	3-2	63.555	-13, -5.599

NFL Week 4 "What If" Ratings, Rankings, and Projections

Below you will find my computer's ratings, rankings, and projections calculated using the week 3 Green Bay at Seattle game as a 12-7 win by Green Bay.

For the ratings and rankings, the change column is the difference from the actual week 3 ratings.  You'll see that the teams would basically swap positions 4 and 8, but a few other teams move slightly as well.

Rank	Team	Rating	Record	Schedule	Change
Rank	Team	Rating	Record	Schedule	Change
1	Houston	94.419	4-0	80.812	+0, +0.019
2	Atlanta	90.417	4-0	79.622	+1, +0.005
3	Arizona	90.251	4-0	82.601	-1, -0.492
4	New England	88.725	2-2	84.404	+0, -0.149
5	San Francisco	88.271	3-1	83.746	+2, +0.665
6	Baltimore	87.957	3-1	80.701	-1, -0.094
7	Denver	87.861	2-2	84.014	-1, +0.015
8	Chicago	86.544	3-1	79.991	+0, +0.533
9	Green Bay	86.041	3-1	80.811	+2, +2.170
10	San Diego	83.990	3-1	78.151	+0, +0.010
11	Minnesota	83.800	3-1	79.858	+1, +0.243
12	Seattle	83.583	1-3	83.751	-3, -1.973
13	Cincinnati	83.413	3-1	81.460	+0, -0.017
14	Miami	82.461	1-3	85.242	+0, -0.020
15	Pittsburgh	82.262	1-2	82.387	+0, +0.050
16	Philadelphia	81.638	3-1	83.475	+0, -0.179
17	NY Jets	81.340	2-2	82.905	+0, +0.121
18	NY Giants	80.661	2-2	77.784	+0, -0.207
19	Dallas	79.424	2-2	81.667	+0, -0.460
20	St Louis	79.287	2-2	81.660	+0, -0.292
21	Washington	78.710	2-2	78.857	+0, -0.081
22	Buffalo	78.625	2-2	79.384	+0, +0.006
23	Jacksonville	78.145	1-3	84.211	+0, +0.093
24	Detroit	77.803	1-3	81.535	+0, +0.088
25	Tampa Bay	75.881	1-3	78.247	+0, -0.184
26	Indianapolis	75.212	1-2	81.829	+1, +0.211
27	Cleveland	75.030	0-4	82.908	-1, -0.067
28	Oakland	74.959	1-3	84.143	+0, -0.001
29	Tennessee	74.783	1-3	86.234	+0, +0.001
30	Carolina	74.193	1-3	79.452	+0, -0.106
31	Kansas City	72.443	1-3	80.970	+0, +0.004
32	New Orleans	70.847	0-4	77.847	+0, +0.078

And in the projected records, Green Bay would go from being 9-7 and tied with Minnesota for the last spot to being 11-5 and tied with Chicago for the division, and since they currently have the tie-breaker would be projected to win it.  Seattle would fall from 9-7 and tied for the last wildcard spot to 7-9 and out of the playoff hunt.

Team	Record	% chance	% +1	% -1
Team	Record	% chance	% +1	% -1
Houston	15-1	40.1	24.1	25.9
Atlanta	14-2	30.9	27.6	19.9
Arizona	13-3	26.6	17.7	24.9
Denver	11-5	30.7	24.3	21.9
Baltimore	11-5	26.2	22.9	19.8
Chicago	11-5	25.5	16.9	24.6
Green Bay	11-5	25.0	16.1	24.9
Philadelphia	11-5	24.7	16.3	24.5
San Francisco	11-5	24.7	19.2	21.8
San Diego	11-5	24.1	16.4	23.7
New England	10-6	24.8	24.2	17.4
Cincinnati	10-6	23.7	20.4	19.5
Minnesota	9-7	26.2	24.5	17.9
Pittsburgh	9-7	23.7	18.0	21.9
Dallas	8-8	24.8	18.7	22.4
Miami	8-8	24.4	17.3	23.3
NY Giants	8-8	24.4	19.0	21.7
Buffalo	7-9	24.9	17.3	24.0
Washington	7-9	24.6	17.6	23.3
Seattle	7-9	23.8	21.4	18.7
NY Jets	7-9	23.0	22.2	16.9
Oakland	6-10	26.1	19.8	22.5
Jacksonville	6-10	25.8	20.8	21.3
Tampa Bay	5-11	29.8	27.1	18.1
Indianapolis	5-11	26.5	20.4	22.4
St Louis	5-11	25.7	24.3	17.4
Carolina	5-11	25.2	20.5	21.1
Kansas City	4-12	28.8	18.9	26.3
Tennessee	4-12	26.6	23.8	18.9
Cleveland	4-12	25.5	17.7	24.5
Detroit	3-13	31.8	28.3	17.3
New Orleans	2-14	30.8	23.7	22.4

I'll continue to calculate these each week so we can see how things would be different.

NFL Week 4 Ratings and Rankings - Houston stays #1

The ratings after week 4 are now posted and listed below as well.

Houston stays #1 while Arizona and Atlanta swap spots behind them.  Seattle drops but still has 3 NFC West teams in the top-9.

New Orleans stays the worst team and interestingly Cleveland is also winless but is ranked 6 spots ahead of them.

Rank	Team	Rating	Record	Schedule	Change
Rank	Team	Rating	Record	Schedule	Change
1	Houston	94.400	4-0	80.790	+0, +1.369
2	Arizona	90.743	4-0	83.182	+1, -0.042
3	Atlanta	90.412	4-0	79.641	-1, -0.762
4	New England	88.874	2-2	84.549	+2, +1.241
5	Baltimore	88.051	3-1	80.805	+0, -0.484
6	Denver	87.846	2-2	83.996	+2, +1.087
7	San Francisco	87.606	3-1	83.091	+0, +0.639
8	Chicago	86.011	3-1	79.584	+4, +2.773
9	Seattle	85.556	2-2	83.519	-5, -3.153
10	San Diego	83.980	3-1	78.148	+0, -0.178
11	Green Bay	83.871	2-2	80.986	-2, -1.957
12	Minnesota	83.557	3-1	79.594	+7, +3.472
13	Cincinnati	83.430	3-1	81.498	+5, +3.339
14	Miami	82.481	1-3	85.331	+3, +1.332
15	Pittsburgh	82.212	1-2	82.342	-1, -0.616
16	Philadelphia	81.817	3-1	83.690	-1, -0.946
17	NY Jets	81.219	2-2	82.730	-4, -1.952
18	NY Giants	80.868	2-2	78.016	-2, -0.518
19	Dallas	79.884	2-2	82.125	-8, -3.521
20	St Louis	79.579	2-2	82.018	+5, +2.222
21	Washington	78.791	2-2	78.961	+1, +0.165
22	Buffalo	78.619	2-2	79.407	-1, -0.754
23	Jacksonville	78.052	1-3	84.097	+0, -0.154
24	Detroit	77.715	1-3	81.381	-4, -1.712
25	Tampa Bay	76.065	1-3	78.460	-1, -1.761
26	Cleveland	75.097	0-4	82.979	+2, +0.615
27	Indianapolis	75.001	1-2	81.540	+3, +1.444
28	Oakland	74.960	1-3	84.130	-2, -0.036
29	Tennessee	74.782	1-3	86.242	-2, +0.186
30	Carolina	74.299	1-3	79.528	-1, +0.088
31	Kansas City	72.439	1-3	80.945	+0, -0.790
32	New Orleans	70.769	0-4	77.350	+0, -0.608

Estimated Dynamic NTRP summary reports available for the Pacific Northwest

Those of you that follow my blog and estimated DNTRP ratings may have wondered why I announced summary reports for the Denver Metro Area but not the Pacific Northwest given that it is the PNW where I started doing the ratings.  There was no slight intended and in fact I can generate summary reports for the PNW for those that are interested.

An example report is shown below and will include results from regular season matches played in Men's and Women's Adult, Senior, Super Senior, Super Duper Senior, and One Doubles leagues for the 2012 league season.  It won't include Mixed league results, as those are not used for computing year end ratings unless you are a mixed-exclusive player.  But contact me if you are interested in a report on Mixed.

John Doe
Current NTRP: 4.5
Estimated DNTRP: 4.47
Match Record: 12-2
Sets Won-Lost: 24-5
Games Won-Lost: 155-72
Best Match Result: 4.92 on 2012-08-25
Worst Match Result: 3.80 on 2012-06-26
Highest Estimated DNTRP: 4.55 on 2012-09-19
Lowest Estimated DNTRP: 4.28 on 2012-06-26

This chart shows both how the Estimated DNTRP changes with each match, and the specific rating computed for an individual match.  This gives a nice visual view of how each match is rated and the trends in your rating.

If you are interested in purchasing a report for a small fee, please contact me.  And if you aren't in the PNW section or the Denver area but are interested, let me know and I'll see what I can do.

Update! Estimated Dynamic NTRP for Denver Metro Area - Now with charts

I wrote earlier that I can now generate summary reports for women in the Denver Metro area.  I have an update on what is now included in the report.  The big news being a chart to show match by match performance.  Here is an example:

Jane Doe
Current NTRP: 4.5
Estimated DNTRP: 4.47
Match Record: 12-2
Sets Won-Lost: 24-5
Games Won-Lost: 155-72
Best Match Result: 4.92 on 2012-08-25
Worst Match Result: 3.80 on 2012-06-26
Highest Estimated DNTRP: 4.55 on 2012-09-19
Lowest Estimated DNTRP: 4.28 on 2012-06-26

This chart shows both how the Estimated DNTRP changes with each match, and the specific rating computed for an individual match.  This gives a nice visual view of how each match is rated and the trends in your rating.

If you are interested in purchasing a report, please contact me.