Darlington Raceway will host the first of three races (6pm ET, Sunday on the USA Network) that will eliminate four drivers from the Cup Series playoffs.
But which driver?
This is a question that is harder to answer this year than it was last year. Only 9 points separate the 5th and 16th seeds. Additionally, the changes wrought by next-generation vehicles cast doubt on the usefulness of historical statistics.
So I developed an algorithm to predict the first four playoff eliminations based solely on 2022 numbers.
How to build an algorithm
Predicting the first driver to leave the playoffs is a very different task than predicting Championship 4. The algorithm used the following metrics:
Top 10 Finishes: By definition, none of the first four drivers eliminated in the playoffs win a playoff race. So the algorithm should be more about running well and scoring points than winning. The playoff driver has 4 (Chase Briscoe) to his 17 (Chase Elliott) top 10 finishes.
Average final position: Drivers with lower average finishing places either have few bad finishes or enough good finishes to offset bad finishes. Briscoe and Denny Hamlin are tied for worst average finishing position at 19.0.
Average running position: Hamlin and Briscoe have the same average finishing position, but Hamlin has double the DNF. Average running position is a better measure of the strength of a driver who fails to finish. Elliott has the best average running for his position (10.2) and Austin Sindrick has the worst (17.9).
Stage win: Stage wins are included because they reflect playoff points. The number of stage wins for a playoff driver goes from his 0 (Austin Dillon) to his 5 (Elliot, Joey Logano, Ryan Blaney, Ross his Chastain).
Stage point: Stage points make a difference, as the points difference between drivers from the 5th seed onwards is small. The caveat is that some crew his chief chooses his position on the track over good stage finish.
Here is the evaluation from the first iteration of the algorithm:
However, as always, there is room for improvement.
accident multiplier
DNF is high this year, and there are many accidents and spins. Some drivers have many penalties. I have combined these into what I call accident multipliers. I didn’t want it to have too much of an impact, so I set it up for drivers who don’t have many accidents. That’s Logano, with 3 penalties and 10 attention incidents.
I set it to 0.8 for the driver with the most accidents. Hamlin holds the unfortunate distinction of 17 penalties and his 14 accidents/spins.
After scaling everyone else’s numbers, we subtracted 0.01 for each DNF.
Chastain dropped from 2nd to 4th and Hamlin dropped from 10th to 14th. However, Hamlin recently avoided a penalty. Chastain’s performance in the second half of the regular season was not as good as in the first half. Plus, every team knows more about their next-gen cars than they did at the start of the season.
Improvement factor
I compared the indicators in the first half and second half of the regular season. The results confirmed some things I had already suspected, but also brought some surprises.
Among the allegations confirmed are that Alex Bowman and Kyle Busch have been having a tough time lately.
Bowman’s finishing average is 7.7 places worse in the second half than in the first half of the regular season. His average running his position he is 5.7 positions worse. Eight of Bowman’s ten top ten finishes came in the first half of the regular season.
Kyle Busch’s average finishing position is down 7.3 spots, but his average starting position is down just 3.1.
Their drop is more than double the next largest drop in the average finishing position. Austin Dillon is down 3.3 places in average finishing position and 1.2 places down in average starting position.
Surprisingly, Daniel Suarez tops the list of improved drivers. His average finishing position increased his 4.1 places, and his average starting position increased his 6.1 places. Suárez has had six fewer penalties and five fewer accidents/spins in the second half of the regular season.
Hamlin and Sindrick are tied 3.92 places higher for the second improved average finish. However, Cindric’s average rank is 2.0 places worse, while Hamlin’s average rank is 3.7 places higher than he is. Hamlin’s team also reduced the number of penalties in the second half of the season.
I thought Kevin Harvick would be on the list of big improvers, but no. His average final position improved by 0.31 places in the second half.
Chastain’s results have been mixed. His average running position is 1.2 spots up, but his average finishing position is 2.2 spots worse. We have compiled all this data into an “improvement bonus” and incorporated it into our algorithm.
predict
Here are the final rankings:
Top 8 drivers have a small improvement bonus. Elliott is the closest to zero as he has been consistently good this season.
Again, the algorithm is designed to predict the first four drivers to drop out, not the final standings of the season.
Also, I’m new to algorithm development, so there’s a lot that could be wrong. For example, you may not be properly weighting different factors.
There is a big gap between the final ranking of the last four drivers and Sindric. Even Sindrick and Hamlin could end in Round 1 if any of the drivers below them perform unexpectedly well.
As for Darlington, Hamlin averaged a running position of 9.9 in the spring and led 42 laps, despite finishing 29th. Bowman, Briscoe, Sindrick and Dillon were all not leading a lap with an average running position of 18.5 or less than him.
