Auto Racing Analytics

So… at what point can we say the 100 races thing is unnecessary?

@CJS97531 Yes, not enough info on him for a fair comparison

@AR_Analytics Is there not enough data for the 5 or where was he running before the tire issue. He seemed to have a fast car

Had the ENTIRE 160 laps at Indy been based on if each driver achieved the result of their Optimal Lap Model for every lap, this is approximately where the fastest cars would've finished on track Highest ovals of 2026 for avg dirty air effect per lap Poc-0.188 Ind-0.103 Tex-0.086

Complete lists of where your favorite driver ranks this season for Optimal Lap Rank, Restarts, Passing, Defending and more is available for subscribers at AutoRacingAnalytics.com

Optimal Lap Model takes into account when drivers are in dirty air/clean air, tire degradation, and a driver's lap time inconsistency to estimate the expected best possible lap time of the car for each lap through a run

So Indy… Less Power (and prevent shifting) More Drag More Tire Wear Do we thing that creates a better product? To be clear I’m happy we’re on the oval. I just can’t help myself thinking if we could make it better

Complete lists of where your favorite driver ranks this season for Optimal Lap Rank, Restarts, Passing, Defending and more is available for subscribers at AutoRacingAnalytics.com

Pass Efficiency on Intermediate Tracks 2026 Passing is expected to be very difficult at Indy. Maybe some of these drivers will be the ones to over come it and work through the field Values are a percentage: Total Real Green Passes divided by Laps in Passing Range

@willowmillenniu @jeff_gluck Optimal lap isn’t trying to show speed as a summation of the whole race. That’s what median lap is for. Optimal lap is trying to show what a cars top end potential should be. It’s more of a performance metric than a results metric. (Thanks for reading if you made it here!)

@willowmillenniu @jeff_gluck It works the other way too. If someone has the fastest car all race, and then makes a bad adjustment, that won’t hurt their optimal lap model. It also is able to take into account dirty air so we can find cars that didn’t get a chance to show full potential. 4/…

Ooof man did Elliott get screwed by that caution

@willowmillenniu @jeff_gluck I always give a super basic explanation under the post so people can understand what they’re looking at, but not everyone wants a stats lesson so I keep it brief. I always am happy to explain in more detail those who ask. I don’t just post my code because I worked hard develop it

@AR_Analytics @jeff_gluck Then there perhaps needs to be more transparent cited evidence and explanation as to how the optimal lap model works because it all sounds very vague. And in the case of the 9 seems a bit far from reality, at least for someone who observed the lap times/in-cars/race itself.

@JoeAlt3 @jeff_gluck For NW I think it was around 0.015 per lap on average. Last year I made an update to hedge that value a little on the low side to prevent it from over fitting cars stuck in traffic as you described. There’s some other specific error checks too, but that’s the basics of it 2/2

@JoeAlt3 @jeff_gluck Sure. The model tracks when cars are within passing range of car ahead (varies by track). It then takes those lap time and compares them to what the model expects they should’ve run. It does this through the field to get an overall dirty air per lap value. 1/many

@JoeAlt3 @jeff_gluck I’d be happy to explain it in detail, but that is not how the model works. I would agree that would be a terrible analysis

@jeff_gluck Jeff, as someone that does this type of thing seriously, you can’t put any stock into auto racing analytics model. it spits out terrible results every week. “Optimal lap” seems to mean he takes their best lap times and multiples that over a race, which isn’t how racing works.

@260ksti @BravesFan1295 @willowmillenniu @jeff_gluck The reason it struggles on road courses isn’t the split strategy, but the length and number of tire runs. The model performs better on runs of roughly 15+ laps, and road courses don’t always have multiple of those

@260ksti @BravesFan1295 @willowmillenniu @jeff_gluck I used to use median lap back in 2022-2023. Then developed optimal lap model to replace it. It’s been tested and shown to be more accurate. Overall fastest lap is not something that plays into the model at all. 1/2