At the start of the season, I published the results of a simulation model that represented my best analytical prediction of how the 2026 season would go. We’re now roughly a quarter way through the season, so it feels like a good time for an update those predictions.
A lot has happened since I published my original set of predictions on March 1: The Indy 500 field has filled out - about as predictably as possible. We have a track layout for the Freedom 250. My Twitter has gained several followers1. And most importantly, four races have actually happened.
Updated Predictions
All of those things affect the model, (except the twitter followers) so let’s see what the updated predictions look like with those changes incorporated:
| Driver | Average Season Points | Race Finish Rate | Average Starting Position | Average Finishing Position | Wins per Season | Poles per Season |
|---|---|---|---|---|---|---|
| Palou, Alex | 656.6 | 82.7 | 5 | 6.9 | 5.5 | 3.7 |
| O’Ward, Pato | 565.1 | 88.4 | 6.2 | 7.6 | 2 | 2.7 |
| Lundgaard, Christian | 506.9 | 88.3 | 10.5 | 8.8 | 0.9 | 0.8 |
| Kirkwood, Kyle | 506.2 | 88.2 | 10.3 | 9.2 | 1.8 | 0.5 |
| Malukas, David | 489.9 | 88.4 | 6.5 | 9.9 | 1.2 | 3.2 |
| McLaughlin, Scott | 485.1 | 88.4 | 10.5 | 10.1 | 1.6 | 1.8 |
| Newgarden, Josef | 453.2 | 88.2 | 10.5 | 10.9 | 1.8 | 1 |
| Armstrong, Marcus | 429.7 | 88.3 | 8.5 | 10.8 | 0.5 | 0.6 |
| Dixon, Scott | 420.6 | 82.8 | 15.5 | 11.1 | 0.3 | 0 |
| Rossi, Alexander | 394.5 | 88.2 | 12.3 | 11.9 | 0.4 | 0.3 |
| Ericsson, Marcus | 385.7 | 88.2 | 10.9 | 12.3 | 0.2 | 1.2 |
| Rosenqvist, Felix | 377.9 | 88.4 | 10.2 | 12.5 | 0.2 | 0.9 |
| Ferrucci, Santino | 344.4 | 82.7 | 15.9 | 13.9 | 0.3 | 0.4 |
| Simpson, Kyffin | 343 | 88.4 | 15.3 | 13.8 | 0.2 | 0 |
| Rasmussen, Christian | 334.6 | 82.8 | 14.6 | 14.5 | 0.3 | 0.1 |
| Power, Will | 323.2 | 82.7 | 15.4 | 14.6 | 0.1 | 0 |
| Grosjean, Romain | 317.5 | 76.8 | 12.8 | 15 | 0.2 | 0.1 |
| VeeKay, Rinus | 317.1 | 88.4 | 15.9 | 14.9 | 0.2 | 0.1 |
| Siegel, Nolan | 302 | 82.7 | 13.4 | 15.6 | 0.1 | 0.2 |
| Rahal, Graham | 294.2 | 88.3 | 15.8 | 15.7 | 0 | 0.1 |
| Robb, Sting Ray | 278 | 88.2 | 18.9 | 16.6 | 0.1 | 0 |
| Foster, Louis | 258.1 | 82.3 | 16.4 | 16.9 | 0 | 0 |
| Hauger, Dennis | 237.2 | 86.6 | 20.2 | 18 | 0 | 0 |
| Collet, Caio | 236.7 | 87 | 20 | 18.1 | 0 | 0 |
| Schumacher, Mick | 224.3 | 82.1 | 16.6 | 18.8 | 0 | 0 |
| Sato, Takuma* | 20.2 | 77.3 | 15.8 | 14.8 | 0 | 0 |
| Carpenter, Ed* | 19.1 | 76.9 | 11 | 16.1 | 0 | 0.1 |
| Hunter-Reay, Ryan* | 18 | 77.8 | 14.1 | 16.2 | 0 | 0 |
| Daly, Conor* | 15.3 | 76.8 | 24.9 | 18 | 0 | 0 |
| Castroneves, Helio* | 14.7 | 78.1 | 22.2 | 18.6 | 0 | 0 |
| Harvey, Jack* | 11.6 | 77.1 | 27.6 | 21.2 | 0 | 0 |
| Driver | Average Championship Standing | Total Championships | Total Indy 500 Wins | Best Standings Finish | Worst Standings Finish |
|---|---|---|---|---|---|
| Palou, Alex | 1.3 | 8152 | 2447 | 1 | 9 |
| O’Ward, Pato | 2.8 | 1220 | 1663 | 1 | 15 |
| Lundgaard, Christian | 4.6 | 160 | 764 | 1 | 16 |
| Kirkwood, Kyle | 4.6 | 139 | 345 | 1 | 15 |
| Malukas, David | 5.5 | 157 | 911 | 1 | 18 |
| McLaughlin, Scott | 5.7 | 143 | 491 | 1 | 19 |
| Newgarden, Josef | 7.2 | 20 | 122 | 1 | 20 |
| Armstrong, Marcus | 8.4 | 7 | 371 | 1 | 21 |
| Dixon, Scott | 8.9 | 0 | 212 | 2 | 21 |
| Rossi, Alexander | 10.6 | 2 | 499 | 1 | 23 |
| Ericsson, Marcus | 11.2 | 0 | 128 | 2 | 24 |
| Rosenqvist, Felix | 11.8 | 0 | 163 | 2 | 24 |
| Ferrucci, Santino | 14.4 | 0 | 378 | 2 | 25 |
| Simpson, Kyffin | 14.5 | 0 | 166 | 4 | 25 |
| Rasmussen, Christian | 15.2 | 0 | 338 | 2 | 25 |
| Power, Will | 16.2 | 0 | 51 | 6 | 25 |
| Grosjean, Romain | 16.7 | 0 | 70 | 5 | 25 |
| VeeKay, Rinus | 16.7 | 0 | 243 | 5 | 25 |
| Siegel, Nolan | 18 | 0 | 79 | 6 | 25 |
| Rahal, Graham | 18.7 | 0 | 2 | 8 | 25 |
| Robb, Sting Ray | 20 | 0 | 77 | 9 | 25 |
| Foster, Louis | 21.8 | 0 | 0 | 13 | 25 |
| Collet, Caio | 23.1 | 0 | 11 | 10 | 25 |
| Hauger, Dennis | 23.2 | 0 | 1 | 11 | 25 |
| Schumacher, Mick | 24 | 0 | 2 | 15 | 25 |
| Sato, Takuma* | 27.9 | 0 | 169 | 26 | 31 |
| Carpenter, Ed* | 28.1 | 0 | 141 | 26 | 31 |
| Hunter-Reay, Ryan* | 28.2 | 0 | 93 | 26 | 31 |
| Daly, Conor* | 28.7 | 0 | 50 | 26 | 31 |
| Castroneves, Helio* | 28.7 | 0 | 13 | 26 | 31 |
| Harvey, Jack* | 29.4 | 0 | 0 | 26 | 31 |
Woah, Alex Palou, who’d have thought?
But seriously, Alex Palou is now being given an 81% chance to win the championship. I know that to us humans, his chances feel closer to 100%, but 81% is still really high for a statistical model. The model is also predicting him to win 3.5 more races this season, which with 5 road courses still on the schedule, actually feels pretty low. It’s not just road courses, either, as he is predicted to win the the Indy 500 about 25% of the time. I know this is shocking, but apparently my model thinks we should expect a good rest of the season from Palou.
Further down the (predicted) championship standings, we find Kyle Kirkwood. Giving Kirkwood the 4th-best championship odds behind both (competitive) McLarens feels a little harsh for the only guy to lead Alex Palou in the championship standings in almost 2 years. My computer, if it were petty enough to argue back, might point out that “we were all saying this last year” and “remember how that turned out?”2 This year certainly feels different from last year, but it’s hard to ignore that we’ve already raced on 2 street tracks — which is Kirkwood and Andretti’s strongest track type — but only one of everything else. We have a lot of ovals and road courses to get through where Kirkwood has historically not shined, and while top 5 finishes in both Phoenix and Barber are promising, the model will need more than two data points to change its mind on Kirkwood’s ability on non-street courses.
Blaming track types is more difficult for Pato O’Ward, who my model still prefers to 4 drivers above him in the standings. O’Ward is better on ovals, but he’s still plenty strong on road and street courses. The model still gives him the 2nd-highest chances of winning this season, and predicts that he is 8 times more likely to win the championship than Kirkwood, Lundgaard, or Malukas — all of whom currently lead him in the standings.
Change Since Start of Season
If you remember my first post (or just clicked on the link), however, you’ll realize that Pato’s championship odds are way down compared to where they were a month ago. This brings up a good point: we can learn a lot by seeing how the model has changed since before the first race. The below table gives the difference between today’s predictions and those from before St. Pete:
| Driver | Average Season Points | Average Finishing Position | Wins per Season | Average Championship Standing | Total Championships | Total Indy 500 Wins |
|---|---|---|---|---|---|---|
| Kirkwood, Kyle | 65.1 | -1.9 | 0.9 | -3.5 | 117 | 104 |
| Ericsson, Marcus | 40.2 | -1.7 | 0 | -4.6 | 0 | 42 |
| Malukas, David | 39.5 | -1.5 | 0.1 | -2.2 | 109 | 313 |
| Hauger, Dennis | 17.4 | -1.1 | 0 | -1 | 0 | 10 |
| Newgarden, Josef | 16 | -0.7 | 0.8 | -1.3 | 3 | 23 |
| Palou, Alex | 12.8 | 0 | 1.1 | -0.2 | 1386 | 405 |
| Armstrong, Marcus | 10.7 | -0.9 | 0 | -1.2 | 1 | 78 |
| Collet, Caio | 10.4 | -0.8 | 0 | -0.5 | 0 | -10 |
| Lundgaard, Christian | 4.2 | -0.5 | -0.3 | -0.2 | -89 | -115 |
| Rahal, Graham | -0.3 | 0 | 0 | -1.2 | 0 | -1 |
| Rossi, Alexander | -5.8 | -0.4 | -0.1 | -0.5 | 0 | -31 |
| Power, Will | -6.2 | 0.1 | 0 | -0.9 | 0 | 3 |
| Schumacher, Mick | -9.5 | 0.4 | 0 | 0.6 | 0 | -1 |
| Foster, Louis | -11.9 | 0.4 | 0 | 0.2 | 0 | 0 |
| Rosenqvist, Felix | -14.5 | 0.1 | -0.1 | 0.1 | 0 | -13 |
| McLaughlin, Scott | -14.8 | 0 | -0.3 | 0.6 | -195 | 108 |
| Grosjean, Romain | -19.4 | 0.7 | 0 | 0.2 | 0 | 17 |
| O’Ward, Pato | -22.4 | 0 | -0.5 | 0.5 | -1286 | -384 |
| Veekay, Rinus | -26.4 | 0.5 | -0.2 | 0.8 | 0 | -97 |
| Ferrucci, Santino | -34.8 | 0.9 | -0.2 | 1.6 | -1 | -172 |
| Rasmussen, Christian | -42.1 | 1.5 | -0.1 | 2.2 | 0 | -25 |
| Robb, Sting Ray | -44.9 | 1.8 | -0.1 | 2.4 | 0 | -55 |
| Dixon, Scott | -45.5 | 0.9 | -0.4 | 2.4 | -45 | -202 |
| Simpson, Kyffin | -47 | 1.2 | -0.2 | 2.6 | 0 | -55 |
| Siegel, Nolan | -58.5 | 2.2 | -0.1 | 3.6 | 0 | -69 |
Now its clear that Kirkwood actually is getting a lot more “respect” from the model, and Pato’s stock has fallen a lot. In fact, Arrow McLaren’s stock as a whole seems to have fallen, with O’Ward and Nolan Siegel3 towards the bottom of the list, and Christian Lundgaard losing an expected 90 championships despite his solid start to the season.
This brings us back to the discussion of what the model has “learned.” Four new races represents just 100 data points for the statistical prediction models to learn off of. Some of our data points are set aside as test data, meaning we are actually closer to 80 new data points since the start of the season. In reality, that’s not a lot of new data to really sway the model vs the entire DW12 chassis era.
The simulation piece of this model is more sensitive to this new data, however, as it is taking into account the results from those first 4 races in each of its 10,000 iterations. This means that whatever actually happened in those 4 races is going to move the predictions a lot compared to what the model expected would happen in those first 4 races.
So why has Pato O’Ward lost basically 1300 simulated championships (out of 10,000) to Palou in the last month? The model might be knocking Pato and especially the McLarens down slightly, but Pato has underperformed his expectations in those first four races while Palou has slightly overperformed4. Clearly, the 25% of championships that the original model had Pato winning required a lot of good luck in Pato’s favor and bad luck in Palou’s favor. So far that hasn’t happened, and now Pato needs even more luck to not just score more points than Palou, but to make up 50 points in the standings — all in four fewer races than we had to work with at the start of the season.
To put things another way, the model hasn’t changed its mind about O’Ward’s talent or Palou’s talent, or even the quality of their respective cars. But Pato O’Ward now needs a lot more luck in his favor on a per-race basis to win the championship, and so the odds of that happening are now much lower.
Conclusion:
So is it time to give up on your hopes of a competitive season where someone besides the World’s Happiest Spaniard5 wins a championship? Well, yes and no. 81% is a high probability, but a model like this is a lot slower to learn from more recent data. You and I6, on the other hand, tend to react very strongly to recent data. The truth is probably closer to the model’s view than the human view, but it’s almost certainly somewhere in the middle. So if the rest of the season does follow the human view more than the model’s view, we might actually have a fight on our hands between Palou and Kirkwood or Lundgaard.
And what if the computers are right? Well, it might be a pretty boring season again for those who aren’t in the Witness Greatness camp. If that’s the case, however, we can all just start taunting Max Verstappen that he won’t quit F1 and come to IndyCar because he’s too scared of Alex Palou. At least then there’d be hope for a more competitive 2027 season.7
Cover photo by Bill Brine, Wikimedia Commons (CC BY 2.5)
All bots, obviously. ↩︎
Dang, my computer is kind of a jerk. ↩︎
All of the pay-to-play drivers (Siegel, Simpson, Robb) have fallen significantly since the start of the season. This could be due to talent really shining through this season, or it could be that the model doesn’t quite adequately capture driver effects vs team effects and so less-talented drivers will sift to the bottom as more “real” data comes in, or it could just be a coincidence - which would help explain why Dixon of all people is sandwiched between these three at the bottom. ↩︎
I don’t know how it’s possible to overperform already sky-high expectations while also DNFing one of four races. In video game terms, I think Palou is over trying to beat the game and has moved on to trying to unlock every achievement. Someone needs to tell him the real world isn’t supposed to work that way… ↩︎
Complimentary, obviously. ↩︎
Assuming you are one of my three human readers and not my thousands of bot followers. ↩︎
Alex would clean the floor with Max too probably. There really is no hope. ↩︎