After another eventful race for the playoff field, it’s time to once again run the simulation and see which drivers are likely to make it through Bristol, and whose seasons are likely to end there. After being unsatisfied with some of the models’ predictive power, I’ve tweaked it to better account for current team performance, better estimate stage finishes/points, and include fastest lap points. You can see a full breakdown at the bottom of this post. Let’s get into the breakdown of probabilities and actual vs expected performance now that we’ve updated the playoff simulation with results from Gateway.
Denny’s win makes what was a virtual lock into the next round real and the extra 5 playoff points give him a nearly 5% bump in advancing out of both the 2nd and 3rd rounds. Larson, Blaney, Byron, Reddick, and Bubba are all now virtual locks on advancing while Bell and Chase are nearly certain to advance; If Bristol becomes another tire-strategy race, it could open the door to exploit a strategic play and secure valuable stage-win playoff points. Their probabilities of advancing out of the 3rd Round have all now jumped up as well: Denny, Larson, Byron, Blaney, and Bell all cluster around 75% odds of advancing, while Reddick and Elliott are in the 60% range.
The biggest mover in probability is Logano, his chances have shot up by nearly 30%; he’s now in the “nearly guaranteed club“, while Chastain remains on the cusp of the club. After another disasterous day for both Berry and Bowman their chances of advancing have shrunk to nearly zero; they’ll either need to win or hope their luck befalls multiple drivers well above the cutline. Dillon’s odds remained unchanged, while SVG hangs on with an outside chance to advance. Let’s take a look at how the drivers on the bubble need to perform to advance.
About the Chart
These charts represent the probabilities a driver is likely to advance given a range of finishing positions. The colors indicate their likelyhood of advancing, while in cell both the probability of advancing (P Adv) is given and the Probability that this result occurs (P Occ) is given. Drivers probabilities of advancing are contingent on both the stage points they earn and how all of the other playoff drivers finish.
Above the Cut Line
Logano enters Bristol in a relatively safe place; as long as he finishes in the top 20 he’s guarenteed to advance, and as long as he finishes within the top 30 he has a very strong chance of advancing. All but a DNF (with no stage points) would see Logano into the 2nd Round.
Chastain is in a very similar position to Logano, as long as he has a solid showing (which is expected 75% of the time) he should be free and clear into the 2nd Round.
Cindric has a reasonably wide path to advancing, he must finish above 20th to have a high probability of advancing; with a 56% chance of that occuring, he should feel comfortable going into the race. He still has a reasonably strong chance of advancing if he finishes between 25th and 21st, but must avoid running outside of the top 25 to feel comfortable throughout the race.
Below the Cut Line
Dillon must run well to have confidence in his advancing into the 2nd Round. Finishing outside of the top 5 gives him only slighlty more than a coin-flips chance of advancing, and a finish anywhere outside of the top 10 means he’s unlikely to advance. With a 25% chance of Dillon running in the top 10, don’t expect to see him make it past the first round. His biggest opportunity to advance is to hope one of the bubble drivers above the cut line DNFs, and that he can capitalize and perform above expectations.
SVG needs to finish inside the top 5 to have any reasonable chance of advancing, with only a 5% chance of occuring, he’s going to have to get lucky to continue his underdog run in the playoffs. If multiple bubble drivers above the cut line were to fall out early, he may have a chance to advance with a finish in the top 15.
Bowman and Berry have to win to advance, it’s that simple, there are no other finishing positions where either are likely to advance besides locking in with a win. Will Bowman finally convert a strong starting position into a strong finish? Or will Berry’s short-track roots give him the skills he needs to advance?
So What changed between the two models?
The RaceSimulator has shifted from treating races like a regression problem to modeling them as a series of ranking contests. Instead of estimating absolute performance with simple driver and track effects, the new system uses a Plackett–Luce architecture that predicts stage results, finishes, and fastest laps together, anchored by a Bradley–Terry likelihood that better reflects how small performance gaps translate into big swings in results. It applies smarter temporal weighting that emphasizes recent races, calibrates stage noise differently than full-race outcomes, and adds a skewed noise component to capture the very real chance of bad luck or mechanical failure. The result is a model that doesn’t just say who’s “fastest on average,” but simulates the messy, uncertain dynamics of actual races.
Some of the biggest changes:
- Briscoe: The model now more heavily weights JGR’s performance in his simulations, boosting his advancement odds significantly
- Byron & Bell: [tk tk]
- Hamlin & Larson: [tk tk]
- Reddick & Cindric: [tk tk]
- Blaney: [tk tk]
- Wallace & Chastain:
- SVG: Trackhouse’s historical performances, and his recent uptick in form bumps up his probabilities
Leave a Reply