@matt_courchene why does the model have a 98% probability of a +3 cut when it seems like +3 may get through?
That’s what the cut line means, dawg.
Yeah, just interpretation. A cutline of +3 means that everyone at +3 or better gets in. You are probably thinking these are “probability of being cut” at each score, which, is fair enough.
On a related note… we had the probability of +4 all the way down to 0.5% earlier in the day, and then it rose up to 35% or so. So pretty big model fail there, but I think I know what the issue is. Predicting cutlines is one of the harder things to do well.
Basically it has to do with changing course conditions.
Matt had an article about that in the 2021 PGA Championship. It was also an issue in the 2021 Palmetto Championship.
In the PGA Championship Mickelson shot the equivalent of a +2 SG in the final round but got credit for a 0 because the wind made the conditions way tougher for the last 5-10 groups compared to all the other golfers.
It was even weirder in the Palmetto Championship because there were 5-6 players in contention that all melted down in the back 9. There was a wind switch and there were several high scores in the back 9 by the closing groups.
You can’t predict the future but what you can do is add a fuzzy “randomness” factor for future scoring conditions. I’m sure Data Golf will think of something.
With that said Bryson was just dumb when he flew home. If he looked at the Live Prediction Model he would’ve realized that his chances of making the cut never fell below 30%.
We do account for uncertainty in future course conditions; started doing that a few years back. If you don’t allow for that your cut probabilities will be much too certain on Friday morning. The issue at Abu Dhabi today was likely that we were projecting (accounting for the weather) a scoring average of like +1.5-2 this morning, but instead it ended up being +3.7. From our model’s perspective this was (very) unexpectedly high, so for the afternoon we were probably projecting something like +3. And the afternoon ended up playing +4.3 or so. That was why a cutline of +4 had such a low probability in the early afternoon. We were still allowing for uncertainty in conditions, but if you are projecting +3, and a scoring avg of +4.5 to +5 is required for the cut to move to +4, then it’s going to be unlikely that hits.
It’s possible on the European Tour we don’t have enough uncertainty in future conditions, as it seems like things can change quicker and more dramatically there. But this is all stuff that can be gleaned from data. It’s also possible that once the wind gets above 20mph the course gets substantially harder (i.e. moving from 10 to 11mph typically yields a smaller change in course difficulty than 20 to 21mph).
Re the Mickelson example, we now correct for this.
If the weather forecast holds up players will be dying on the Stadium Course in Round 3. This could be fun
Yeah. You think that course will be made relatively more difficult by the wind than the other two? 17 will be playing really tough.
Wind is a tricky beast though, in Hawaii if the wind blows at 20-25mph the players average -3. I guess it all depends on the setup and the wind direction.
The most amusing situations are when golf courses go from being roadkill to being impossible in the space of 1 day. 1999 Bob Hope Invitational and 1997 Las Vegas Invitational are two of the more infamous examples.
Looks like the tournament officials left the sprinklers on overnight, the greens were unusually soft in round 3. Certainly softer than what you expect when it’s sunny and breezy.
Just once I want to see how the Stadium Course plays when the course is rock hard. It’s designed for bone dry conditions, with the slopes, bunkers, and uneven lies. This course doesn’t really need rough, just extremely hard and fast fairways and some desert brush.