Dropping the preseason games from the equation has seriously scrambled the ratings but now I will stop expecting the Vikings to win. I hope no one ordered Super Bowl tickets. The orange bars are based just on the regular season; the others have the preseason games included, with diminishing weight as games were added. Up is good; down is not, zero is average.
Here’s the big picture, etc., . . . Week 3
It was another 50% week, but now I have gotten the preseason out of my system.
This little experiment has now established definitively that the preseason means nothing. The correlation between the four preseason games and the first three weeks of the regular season is 0.08 with a standard error of 0.16, which of course means the correlation is not statistically different from zero.
My two Super Bowl teams based on the Preseason, Denver and Minnesota, have been advancing steadily toward mediocrity.
| Visitor | Home | Probability of Visitor Win | V Pts | H Pts | Logit Visitor | Logit Home |
| Tampa Bay | Atlanta | 0.48 | 14 | 56 | -0.47 | -0.39 |
| SUN, SEP 21 | ||||||
| San Diego | Buffalo | 0.68 | 22 | 10 | 0.18 | -0.56 |
| Dallas | St. Louis | 0.48 | 34 | 31 | -0.47 | -0.40 |
| Washington | Philadelphia | 0.56 | 34 | 37 | 0.38 | 0.15 |
| Houston | NY Giants | 0.56 | 17 | 30 | 0.16 | -0.07 |
| Minnesota | New Orleans | 0.69 | 9 | 20 | 0.63 | -0.16 |
| Tennessee | Cincinnati | 0.29 | 7 | 33 | -0.33 | 0.54 |
| Baltimore | Cleveland | 0.69 | 23 | 21 | 0.61 | -0.19 |
| Green Bay | Detroit | 0.501 | 7 | 19 | 0.183 | 0.181 |
| Indianapolis | Jacksonville | 0.57 | 44 | 17 | -0.22 | -0.49 |
| Oakland | New England | 0.42 | 9 | 16 | 0.00 | 0.31 |
| San Francisco | Arizona | 0.25 | 14 | 23 | -0.27 | 0.86 |
| Denver | Seattle | 0.57 | 20 | 26 | 0.90 | 0.62 |
| Kansas City | Miami | 0.53 | 34 | 15 | -0.26 | -0.39 |
| Pittsburgh | Carolina | 0.33 | 37 | 19 | -0.55 | 0.14 |
| MON, SEP 22 | ||||||
| Chicago | NY Jets | 0.41 | 27 | 19 | -0.49 | -0.13 |
Here is picture of the Week 2 train wreck
Week Three
The Vikings, on the basis of their week two performance, got their logit cut in half but still rank third, behind the Broncos and Cardinals and ahead of the Seahawks. Perhaps Rasch should stick to remedial reading tests and leave professional athletes along. I will forego the obvious sarcasm. For the second week, the Vikings have the honor of playing an underperforming team who have an elite quarterback and who lost last week. The Vikings definitely will not win by three touchdowns; in fact, no one will, statistically speaking. The Cardinals against the Forty-niners is the most likely.
The Broncos (#1) at the Seahawks (#4) will be the hot ticket. The Packers at the Lions may be the game to see. They are ranked 9 and 10 and the game is a mathematical toss-up to two decimal places. Actually the Packers have a logit of 0.183 to the Lions’ 0.181, which gives the Packers a probability of 0.501. Don’t bet the farm. As a point of emphasis, nothing you see or hear here should be used for anything resembling gambling.
| Visitor | Home | Probability of Visitor Win | V Rank | H Rank | Logit Visitor | Logit Home |
| Tampa Bay | Atlanta | 0.48 | 28 | 24 | -0.47 | -0.39 |
| SUN, SEP 21 | ||||||
| San Diego | Buffalo | 0.68 | 11 | 32 | 0.18 | -0.56 |
| Dallas | St. Louis | 0.48 | 27 | 26 | -0.47 | -0.40 |
| Washington | Philadelphia | 0.56 | 7 | 13 | 0.38 | 0.15 |
| Houston | NY Giants | 0.56 | 12 | 16 | 0.16 | -0.07 |
| Minnesota | New Orleans | 0.69 | 3 | 18 | 0.63 | -0.16 |
| Tennessee | Cincinnati | 0.29 | 23 | 6 | -0.33 | 0.54 |
| Baltimore | Cleveland | 0.69 | 5 | 19 | 0.61 | -0.19 |
| Green Bay | Detroit | 0.50 | 9 | 10 | 0.18 | 0.18 |
| Indianapolis | Jacksonville | 0.57 | 20 | 29 | -0.22 | -0.49 |
| Oakland | New England | 0.42 | 15 | 8 | 0.00 | 0.31 |
| San Francisco | Arizona | 0.25 | 22 | 2 | -0.27 | 0.86 |
| Denver | Seattle | 0.57 | 1 | 4 | 0.90 | 0.62 |
| Kansas City | Miami | 0.53 | 21 | 25 | -0.26 | -0.39 |
| Pittsburgh | Carolina | 0.33 | 31 | 14 | -0.55 | 0.14 |
| MON, SEP 22 | ||||||
| Chicago | NY Jets | 0.41 | 30 | 17 | -0.49 | -0.13 |
I have now reached the level of chimpanzees throwing darts. Three of the four teams that I gave three touchdown spreads won but none beat the spread but I sort of predicted that. The fourth was Minnesota and that’s been talked about enough but even I don’t think any one player would account for a 44 point swing.
Week 3 should be the last week that the preseason has any weight at all but in the meantime, the matrix of coefficients is still singular without it. And here’s what I thought last week and what happened on the field.
| Visitor | Home | Probability of Visitor Win | V Pts | H Pts | Logit Visitor | Logit Home |
| Pittsburgh | Baltimore | 0.22 | 6 | 26 | -0.76 | 0.49 |
| Miami | Buffalo | 0.66 | 10 | 29 | -0.27 | -0.94 |
| Jacksonville | Washington | 0.38 | 10 | 41 | -0.41 | 0.06 |
| Dallas | Tennessee | 0.31 | 26 | 10 | -0.67 | 0.12 |
| Arizona | NY Giants | 0.79 | 25 | 14 | 1.05 | -0.29 |
| New England | Minnesota | 0.18 | 30 | 7 | -0.25 | 1.24 |
| New Orleans | Cleveland | 0.56 | 24 | 26 | -0.01 | -0.26 |
| Atlanta | Cincinnati | 0.31 | 10 | 24 | -0.25 | 0.56 |
| Detroit | Carolina | 0.68 | 7 | 24 | 0.39 | -0.38 |
| St. Louis | Tampa Bay | 0.62 | 19 | 17 | -0.20 | -0.67 |
| Seattle | San Diego | 0.69 | 21 | 30 | 0.88 | 0.07 |
| Houston | Oakland | 0.42 | 30 | 14 | 0.08 | 0.39 |
| NY Jets | Green Bay | 0.36 | 24 | 31 | -0.13 | 0.45 |
| Kansas City | Denver | 0.21 | 17 | 24 | -0.26 | 1.06 |
| Chicago | San Francisco | 0.39 | 28 | 20 | -0.65 | -0.22 |
| Philadelphia | Indianapolis | 0.57 | 30 | 27 | 0.03 | -0.26 |
Think of the preseason as a prior distribution; it’s still in there but not weighted as heavily as last week. The basic idea is rather simple; some would say rather simplistic: teams that play strong opponents, score lots of points, and allow very few will be rated highly in logits. Because we are using the log ratio of points for to points against, it really likes strong defenses. We pay no attention to home versus away, injury reports, weather forecasts, or police blotters; just points scored and points allowed and against whom. I don’t even know what a blotter is.
Here’s what you need to know to ravage the next office pool; or at least to not do anything really dumb. It may be really dumb on my part, but the Vikings, Broncos, Ravens, and Cardinals by three touchdowns. Big spreads are dangerous because one team or the other may forget to play the fourth quarter.
| Visitor | Home | Probability of Visitor Win | V Rnk | H Rnk | Logit Visitor | Logit Home |
| Pittsburgh | Baltimore | 0.22 | 31 | 6 | -0.76 | 0.49 |
| Miami | Buffalo | 0.66 | 24 | 32 | -0.27 | -0.94 |
| Jacksonville | Washington | 0.38 | 27 | 13 | -0.41 | 0.06 |
| Dallas | Tennessee | 0.31 | 30 | 10 | -0.67 | 0.12 |
| Arizona | NY Giants | 0.79 | 3 | 25 | 1.05 | -0.29 |
| New England | Minnesota | 0.18 | 20 | 1 | -0.25 | 1.24 |
| New Orleans | Cleveland | 0.56 | 15 | 23 | -0.01 | -0.26 |
| Atlanta | Cincinnati | 0.31 | 19 | 5 | -0.25 | 0.56 |
| Detroit | Carolina | 0.68 | 8 | 26 | 0.39 | -0.38 |
| St. Louis | Tampa Bay | 0.62 | 17 | 29 | -0.20 | -0.67 |
| Seattle | San Diego | 0.69 | 4 | 12 | 0.88 | 0.07 |
| Houston | Oakland | 0.42 | 11 | 9 | 0.08 | 0.39 |
| NY Jets | Green Bay | 0.36 | 16 | 7 | -0.13 | 0.45 |
| Kansas City | Denver | 0.21 | 22 | 2 | -0.26 | 1.06 |
| Chicago | San Francisco | 0.39 | 28 | 18 | -0.65 | -0.22 |
| Philadelphia | Indianapolis |
What’s up with the home field?
Nine out of 16 correct, which is not statistically better than a chimpanzee throwing darts. The biggest surprises were the Colts, Texans, and Jets. And nine of the visiting team exceeded expectations. Only five of them managed to outscore the opponent but a moral win is a moral win.
I’ll have something to say about week 2 before anyone takes the field, which will include the updated logit ratings and ranks for everyone. The Vikings may be number one.
HTML and I still aren’t on the same page with table formatting.
Visitor Home Probability V H Correct Exceed
– – – – – – – – – – – – – – – Of V Win Points Pick Expected
Green Bay Seattle 0.41 16 36 1 Seattle
New Orleans Atlanta 0.60 34 37 0 Atlanta
Minnesota St. Louis 0.76 34 6 1 Minnesota
Cleveland Pittsburgh 0.78 27 30 0 Pittsburgh
Jacksonville Philadelphia 0.48 17 34 1 Philadelp
Oakland NY Jets 0.80 14 19 0 NY Jets
Cincinnati Baltimore 0.36 23 16 0 Cincinnati
Buffalo Chicago 0.33 23 20 0 Buffalo
Washington Houston 0.67 6 17 0 Houston
Tennessee Kansas City 0.54 26 10 1 Tennessee
New England Miami 0.54 20 33 0 Miami
Carolina Tampa Bay 0.52 20 14 1 Carolina
San Francisco Dallas 0.59 28 17 1 San Fran
Indianapolis Denver 0.10 24 31 1 Indianap
NY Giants Detroit 0.26 14 35 1 NY Giants
San Diego Arizona 0.34 17 18 1 San Diego
If you are reading this on Monday, you may be sensing that predictions based on Pre-season results may not be the best indicator of regular season. The accuracy of these predictions should be enough to convince you that I did the arithmetic before Sunday. I did it a couple hours before the Packer-Seahawk game and wrote this text on Friday so I’ll take credit for getting that winner right. However, the Seahawks should only have won by about 10 points, not 20.
The “Correct Pick” is yes if the team with the higher probability wins; the team in the “Exceed Expected” column is the team for which the observed logit is greater than the predicted logit. For those who were watching football rather than doing homework on weekends, that means they beat the spread (as I not Las Vegas define the spread.)
Right now I am saying that the Broncos and the Vikings are the two best teams in football but it may not be time to order your Super Bowl tickets. The regular season should be a convincing demonstration of regression to the mean, probably for the Vikings, maybe for the Broncos, and hopefully, for some, for the Bills and the Steelers, who are defining the other end of the scale.
I will update this on Tuesday and try again on formatting the table.
| Week | Visitor | Home | Probability of Visitor Win | V Pts | H Pts | Correct Pick | Exceed Expected |
| 1 | Green Bay | Seattle | 0.41 | 16 | 36 | Y | Seattle |
| 1 | New Orleans | Atlanta | 0.60 | ||||
| 1 | Minnesota | St. Louis | 0.76 | ||||
| 1 | Cleveland | Pittsburgh | 0.78 | ||||
| 1 | Jacksonville | Philadelphia | 0.48 | ||||
| 1 | Oakland | NY Jets | 0.80 | ||||
| 1 | Cincinnati | Baltimore | 0.36 | ||||
| 1 | Buffalo | Chicago | 0.33 | ||||
| 1 | Washington | Houston | 0.67 | ||||
| 1 | Tennessee | Kansas City | 0.54 | ||||
| 1 | New England | Miami | 0.54 | ||||
| 1 | Carolina | Tampa Bay | 0.52 | ||||
| 1 | San Francisco | Dallas | 0.59 | ||||
| 1 | Indianapolis | Denver | 0.10 | ||||
| 1 | |||||||
| 1 | NY Giants | Detroit | 0.26 | ||||
| 1 | San Diego | Arizona | 0.34 |
NFL on Rasch 