The model was obviously low on Phil this week after his win at the PGA (we had him 0.4% to win, books were in the 1.4% - 2% range).
If he plays well again this week we could be in the awkward situation where a high-ceiling, but slumping, player suddenly regains form. Recent examples of this are Spieth and Leishman 
Although as things stand thru 10 holes, it looks like Phil is back to his usual waysā¦
Smart money is that this performance was an anomaly but this is golf, anything can happen. I will be rooting for him to win but I will be betting on him to miss the cut. Iām guessing Iāll get ~+120 or so to miss the cut in the US Open and I will bet it. I already bet on him to miss the cut at +168 this week and Iām in good shape after round 1.
If he puts together a couple more Top 20ās and datagolf continues to insist that heās a -0.5 SG golfer then Iāll just ignore the datagolf simulations for Phil.
Recent examples with Spieth and Leishman are good ones. I lost a lot of money on matchups and miss the cut wagers as a result. After being burned like 3 times I just decided I wasnāt going to bet against them.
Who knows, next up might be guys like Keegan Bradley or Rickie Fowler. Only this time I hope to avoid losing money betting against them.
Yeah @matt_courchene has PTSD from tracking Leishman matchups over the past couple months.
You wouldnāt have to completely ignore DG sims, you can just run your own after sliding more weight towards Philās recent rounds!
Itās kind of wild how keeping an eye on Data Golf every week/day/hour like some of us do - we are able to see these trends happen week to week.
I try to follow the model 98% of the time but with Spieth and Leishman on the PGA Tour and with guys like Nienaber and Valimaki on the Euro Tour, you learn to make small adjustments before making certain bets.
Iām too lazy to make adjustments when guys like this are toward the top to see how it affects the odds of other players - I think itās pretty negligible when one player has slightly better actual odds than the model is showing.
The Spieth and Leishman improvements happened over a period of time - Iām guessing Phil wonāt be the next version of thisā¦
Ohhh Nienaber and Valimakiā¦ god that was a rough stretch.
The annoying thing is that they have both been pretty bad in 2021 (although Nienaber did win a couple weeks ago on the Challenge Tour). Their play last fall might end up being a minor blip in their performance trends. Whether the model should be able to better capture that is a separate questionā¦
The next question is whether Fowler will be the next version of this. T8 in the PGA Championship and T7 after Round 1 in the Memorial. Thereās a decent chance that heās ābackā.
I donāt want to be that guy betting on him to lose matchups to Brendan Grace or to miss the cut at +150 while heās finishing inside T20 every week.
Golf is a tough game to handicap because on the one hand you need long-term results to distinguish between noise and genuine improvement. On the other hand you can go from a +0.2 SG to +2.2 SG in the space of one 2-hour range session, or go from +1.6 SG to -2.0 SG overnight and quit golf 6 months later.
IMO it is obv good to note the players that fall in the āheās backā category but I feel like this more the exception to the rule not the norm. (Altho maybe the subset of āwas eliteā like Phil, Fowler, Spieth are just a different breed.) I just think back to guys like Brendan Todd in 2019 who were just on fire and then came back to earth. Benefit of using a model is that it uses a statistically significant sample that ignores recency biases that we all fall into.
Recency bias helps you earn a lot of money though. You can ride the hot hand with players like Higgo and score a couple of wins before he comes back down to earth. In the meantime you can avoid playing him to miss the cut at +400 and lose 7 times in a row while heās making you look like a fool.
The biggest flaw of any model is that it predicts the past, not the future. We often see these players gain or lose 2.0 SG overnight without any warning and play at that level for an extended period of time. It seems like luck but then you turn on the TV and theyāre not missing a shot. Last year DeChambeau had a 7 tournament stretch where he became a +3.0 SG golfer and it looked like he had broken the game. He eventually came back down to earth but those who insisted that he was a +1.5 SG golfer who got lucky and had a 2% chance of winning the US Open looked very silly (and did not make money).
Hindsight is 20-20 but looking back it seems like I made a lot of bad bets that were avoidable.
As you admit Daniel, hindsight is 20/20. For every Garrick Higgo, who goes on to play well for 4 straight weeks there are several players who have 1 good week and then revert back to their baseline the following week. How do you identify Higgo versus the others?
I also think the point about āluckā in golf is very interesting. In golf, very little comes down to luck in the true sense of the word, e.g. bounces off trees, favourable gusts of wind, etc, but rather āluckā is just performing way better than you āshouldā be. One model you could use to frame this is to think about the nuance in your bodyās biomechanics as it relates to hitting golf shots: some shots/rounds/weeks your timing is just better than others, and when that timing is on youāll be objectively better at golf (and we can call it āluckā, in the more familiar sense, when that timing is on).
I think thatās why short-term form can be so alluring to bettors, itās not that rare for a mediocre player to play like the #1 player in the world for a couple weeks (obviously they canāt gain 20 yards, but with improved accuracy, iron play, putting, etc). Itās easy to look at that and think itās clearly not due to luck in the traditional sense, so why shouldnāt it continue?
Iām not trying to defend the model too vigorously, as I know it can miss on guys like Higgo and we feel the betting frustration of that first-hand (especially because we never deviate from model predictionsā¦). But, I do think it ends up being correct some of the time on these guys: Westwood, Leishman, and even Bryson the model had some big wins betting against them. Bryson at the November Masters was good for the model, for example. And Westwood reverted very quickly after his 2 good weeks, which was mostly profitable for us.
Actually Higgo was one of the easier cases to foresee since he was a hot prospect from the start. He turned pro at the age of 20 and won in he Big Easy Tour, the Sunshine Tour, the Challenge Tour, and the European Tour. At this point he was widely considered to be a young star who had already arrived as one of the top players on the European Tour.
By the start of 2021 he was an established player and the bookies had already pegged him as one of the top 5 favorites to win every time he teed it up. He was going off at 10-1 odds every week so I should have regarded the Datagolf win odds with some skepticism. By the time he rattled off 3 straight top 10ās, including his second European win, warning bells should have been going off and I should have played the hot hand.
Instead I played him to miss the cut that week. He shot -27 and won by 6 strokes.
I mean, I have no one but myself to blame here and hindsight is 20-20 but that was a colossally stupid bet.
Iām not asking DataGolf to try to incorporate everything into the model, because thatās impossible. But what I would like is some for meaningful conversation among Datagolf members on which projections are likely to be accurate and which projections Datagolf simply whiffed on. That way I can avoid laughing off McIlroyās chances and then see him win by 8.
Just to play devils advocate, you could say something similar about Matteo Manassero. Huge star all his life, amazing amateur career, really strong start to his professional career and then all of a sudden it just REALLY fell off and he hasnāt been relevant since, really
Not sure what Higgoās future trajectory is but if statistical modeling showed that Manassero had a 3% chance of winning and a 75% chance of making the cut in an average strength of field European Tour event in the middle of 2013, I should be very skeptical. (And he won at a lower rate than Higgo, by the way)
Higgo could become a superstar or fall off the map, but that doesnāt change the fact that I should have been more skeptical of the Datagolf projection in this instance.
I want DG users to discuss where the model is missing, too. Re Higgo, itās one thing to rhyme off accolades, but itās another to pinpoint why the model was off. We had the data from all those wins in the model. If you look at his career true strokes-gained, heās really only had these 3-4 recent good weeks. Youāll also notice we have a lot of data on him as an amateur (he played in the US system) and he wasnāt too special, never playing better than the avg. player in the D1 NCAA Championship (i.e. about -2.3 true SG / round) for a sustained period of time. Didnāt come close to cracking our top 200 all-time amateur rankings. Perhaps his game is built for the professional ranks, as I have heard other pros say he is an impressive ball-striker.
Again, I think Higgo is a similar case to others where it seems like the model canāt update quick enough on guys who have a string of really strong performances. I personally donāt think Higgo has shown much evidence yet that he can be a solid PGA Tour professional; those European Tour events he won were weaker than KFT events.
FYI, I played Mickelson to miss the cut at the US Open at +120. I donāt think itās money in the bank though, he has shown signs of good play even apart from the PGA Championship so I think this one is 50-50.
@matt_courchene I was just gonna post something about Higgo and how the model is evaluating him this week. Nearly every matchup is showing value against him and the outrights and top5/10/20 markets are WAY overvaluing him if you look at DG numbers. Hes so young its hard to accurately forecast how great of a player he will become. For betting purposes this week how would you guys treat him? Seems logical to estimate his skill as somewhere between where DG and books have it (altho that is quite a large gap!)
The eye test says heās a top level player with a totally complete game.
Heās like a human cheat code. His last 6 results are T4, 1, T8, 1, T64, 1. The dude is 22 and anecdotal reports indicate that he is improving very rapidly.
Not certain whether he can sustain this level of play but if you erased the name āGarrick Higgoā and penciled in the name āViktor Hovlandā he would be perceived very differently.
Higgo might end up being goodā¦ but to compare him to Hovland is pretty ridiculous. Hovland has been more than 2 strokes/round better than Higgo for the last few years, going back to their amateur days. Higgo has now won 3 events that were a lot closer to Korn Ferry quality (last week was better obviously, but still the worst event on Tour this year apart from opposite field events). There is no reason to put him near the highest echelon of young players yet (Morikawa, Hov, Scheffler, Niemann). To be a top player in the world you need to average 1.5-2 SG / round consistently. Higgoās last 6 events, where heās playing out of his mind, is in that range. He has a lot more to prove.
Just saw you started a Higgo thread. Could quote this and move it over if you reply.
No problem, feel free to paste in your comment in the other thread.
