THE MODEL DUEL
Basic stats vs. engineered metrics

Two models. Ten Chase races. One question: do fancy metrics actually predict better?

Every race of the 2026 NASCAR Chase, a Basic model built on box-score stats and an Advanced model built on engineered loop-data metrics each project the finishing order. After the checkered flag, both get scored against reality.

Live from Supabase

Basic model

Leading

Box-score stats: avg. start, win %, top-10 %, finish %, success rate

3/ 6 races won
Avg MAE
7.40
Avg R²
0.251
Top-10
36

Advanced model

Engineered metrics: speed score, pass-gain, cPOMS, wAvg finish, pos. vs. expected

3/ 6 races won
Avg MAE
7.44
Avg R²
0.248
Top-10
30

“Race won” = lower mean absolute error that weekend. Tiebreak on season average MAE. Feature importances are aggregated into the end-of-Chase report.

How the models are tracking

avg race MAE
0.05.310.6DarlingtonWorld Wide TechnologyBristolKansasLas VegasCharlotteBasic 7.40Advanced 7.44

Running average of each model's per-race mean absolute error. The line to watch is which one settles lower as the Chase goes on.

Chase schedule

6 of 10 races scored

Latest: Charlotte Motor Speedway

Chase Briscoe won the Oct 11, 2026 race; both models had Denny Hamlin on top (finished 5th).

Basic
MAE 7.95 · R² 0.108 · top-10 6/10
Advanced
MAE 7.84 · R² 0.147 · top-10 6/10
Full breakdown

Chase standings swing

Season chart →

Chase points behind the leader

Watch the gap swing across Charlotte Motor Speedway. Toggle to replay it.

  1. 1Kyle LarsonLeader
  2. 2Denny Hamlin−40
  3. 3Christopher Bell−55
  4. 4Joey Logano−71
  5. 5Ryan Blaney−80
  6. 6Chase Briscoe−84
  7. 7Tyler Reddick−106
  8. 8Ty Gibbs−110
  9. 9Austin Cindric−128
  10. 10Bubba Wallace−137
+6 more on the race page