Tyler Reddick won the Freeway Insurance 500 on Sunday having led 18 of 312 laps. The model gave him 5.2%, seventh on its board.

Ryan Blaney, the published favorite at 15.4%, started on pole, led 51 laps, and finished second. So the model's top pick has now failed to win four races running, and this is the version of that result it can live with: the car it rated highest ran at the front all afternoon and lost to one that spent 294 laps somewhere else.

Those four favorites were priced at 16.1%, 16.2%, 15.7%, and 15.4%. The chance that none of them won is 50.2%, and the expected number of wins across the four was 0.63. Four races is an anecdote, and this is the anecdote you would expect to get.

The result worth more than that one was called here a week ago. The argument then was that Chase Briscoe's form rating led the field and the model still had him eighth at 3.6%, because a driver's strength is 0.65 times his rating at the track and 0.35 times his recent form, his Phoenix number was poor, and the model has no concept of a driver changing equipment. Phoenix was named as the test.

He started third, led 13 laps, and finished eighth.

Take that for what it is. The model publishes a win probability, not a projected finishing position, so a driver ranked eighth finishing eighth is consistent with the call rather than proof of it. One race settles nothing about whether a hot streak should outrank a thin record at a specific track. It is the outcome the reasoning pointed at, though, and the reverse would have counted against it.

What the model had no read on was the shape of the afternoon. The three drivers who led the most laps were Christopher Bell with 85, Joey Logano with 84, and Bubba Wallace with 52. They finished fourth, sixth, and seventeenth. The track catalog files Phoenix under Flat & Short Intermediate and rates its variance low, and the race still went to the driver in front at the end rather than any of the drivers in front all day.

There is a pattern forming in the misses, and it is a flattering one. The winners of the four scored races came second, thirteenth, eighth, and seventh in the model's order. Las Vegas was a genuine shock. Phoenix was a car the model had inside its top ten and priced above 5%, which is roughly one race in twenty. The favorite keeps losing while the board keeps getting closer to the winner, and those two things are easier to hold together than they look.

Blaney's other published number came in, for whatever a single race is worth: he was given 51.6% for a top five and ran second.

The championship is tighter than any of this. Kyle Larson leads with 2,370 points, Bell is five back, Briscoe sits 24 behind in third, Denny Hamlin is 31 behind, and Blaney is 40 behind in fifth. Logano, who led 84 laps on Sunday and finished sixth, is 44 out. The full standings are here.

Three races are left, and the next one is Talladega. The model has no demonstrated skill at drafting tracks, a limitation it states rather than buries, so it has nothing useful to offer about Sunday beyond the observation that the race is close to a lottery. Five points covering the top two, with a pack race next, is a poor arrangement for anyone hoping this season gets decided on merit. The one instrument this site has for sorting a field is going to sit that one out, and say so.