Ryan Blaney is the model's pick for Sunday at Phoenix, at 15.4% to win and 51.6% for a top five. Christopher Bell follows at 12.6%, then Denny Hamlin at 10.3%, William Byron at 10.3% and Kyle Larson at 9.2%.
Blaney is there on the strength of the track rather than the season. Across nine Phoenix races since 2022 he has eight top-five finishes, two of them wins, and the single exception is an engine failure in March 2025. That record gives him a Phoenix rating of 109.67, the highest of the 39 entries, and the model builds a driver's strength from 0.65 times his track rating plus 0.35 times his recent form. His form, 92.80, is second in the field by four hundredths of a point. Both numbers pointing the same way make him the strongest car in the field by the model's reckoning, and the only driver above 15%.
The uncomfortable part is what has happened to the other drivers who held that position. In the three races the forward test has scored, the favorite has lost every time. Kansas went to Larson at 15.2% while Hamlin was the favorite at 16.1%. Las Vegas went to Chase Briscoe at 2.23%, 13th in the model's order. Charlotte went to Briscoe again at 4.06%, eighth that week.
Three favorites, three losses, and almost nothing to explain. Those three were priced at 16.1%, 16.2% and 15.7%, which puts the chance of all three losing at 59%. The likeliest single outcome across three races was exactly the one that occurred, and the expected number of favorite wins was 0.48. Over the same three races the full set of predictions scored 18.1% better than a uniform guess, so the model has been adding real information while its most visible number kept coming up short. Worth saying plainly: three races is an anecdote, not a sample.
What makes Phoenix a sharper test than the three behind it is the kind of race the model thinks it is. The track catalog files Phoenix under Flat & Short Intermediate, and the settings fitted for that type differ sharply from Charlotte's Intermediate & Large Oval. Pace and finish are coupled at 0.477 at Phoenix against 0.274 at Charlotte, and higher coupling means the finish tracks the running order more tightly. The chance that a car finishes a race but loses track position on the way runs 13.2% at Phoenix against 21.6% at Charlotte.
Charlotte, in other words, was a race the model expected to come apart, and it duly did, with a winner it had eighth. Phoenix is one where it expects speed to convert. That cuts both ways for Blaney: it is the reason his number is the highest on the board, and it removes the excuse if he leads and finishes eighth.
One detail argues against reading too much into his position anyway. At 15.4% he is priced lower than each of the three favorites who lost, which is the model being slightly less sure of him than it was of Hamlin at Kansas or Larson at Las Vegas. The calibration behind that number is also pointed only at how often the favorite wins, not at the shape of the rest of the finishing order, which is a known weakness rather than a hidden one.
Blaney goes in fifth in the 2026 standings, 85 points behind Larson with four races to run, while Briscoe's two wins have carried him to fourth. Briscoe is eighth again at 3.6%, the same figure Monday's piece was built on, and it has not moved because the Phoenix prediction was logged within minutes of the Charlotte result. Two wins in three weeks lifted his form to the best in the field and left his Phoenix history exactly where it was.
None of this will settle whether the favorite is priced correctly. A fourth scored race cannot do that. What Sunday can say is whether the one kind of track where the model expects speed to hold up actually delivers it.