NBA Postseason Best Bets (5/14)
Alex Christenson
Now four weeks into this article, we can confidently say that we have the best PGA ownership projections in the industry. We’ve magnified this advantage by introducing our new GPP scores, which have resulted in back-to-back weeks of tremendous success at Pebble Beach and TPC Scottsdale, and more recently an enormously important fade at The Genesis Invitational. As you’ll see, this week was as much about fading Justin Thomas as it was landing on Hideki Matsuyama – and it was a hell of a lot easier to do the former than the latter.
Key Differences from the Industry
How much does ownership really matter? This is a question often discussed in our industry, so allow me to take a stab at answering it. I believe ownership matters only a little bit… until it matters a ton. More specifically, small differences in ownership from one source to another is close to meaningless (like 12% vs. 13%). Large differences, however, are enormously important, in large part because they affect the decisions we actually make on who to play or fade.
It therefore makes sense to look at the instances where our ownership projection deviated most from the industry average and compare those projections to the actual ownerships observed in contests.
All actual ownership percentages for this section will come from the large $33 Single-Entry contest. This tournament is large enough to represent the entire field of DFS players, yet the single-entry aspect makes it more about the psychology of clicking on each individual player as opposed to what an optimizer spits out.
Our three largest differences were:
Justin Thomas
Industry: 22.4%
FTN: 33.5%
Actual: 42.4%
Rory McIlroy
Industry: 17.0%
FTN: 8.7%
Actual: 8.1%
Adam Scott
Industry: 18.5%
FTN: 26.0%
Actual: 37.2%
We continue to be correct (whether absolutely or directionally) on the guys we deviate the most from industry. We will continue to share the r^2 and RMSE results for transparency, but this is what having the best ownership projections means to me – being right on the guys who are fundamentally the most important to the slate.
It doesn’t always happen this way, of course, but we again saw the two guys who were even chalkier than people realized (Thomas and Scott) underwhelm (especially Thomas!). It’s a shame that we didn’t get more from the super-contrarian McIlroy, but that’s a play I’d make 100 times out of 100.
Accuracy Analysis
We can compare overall accuracy by looking at r-squared and RMSE. For those unfamiliar with these metrics, you can focus on just r-squared. If we were exactly right about every single player’s ownership to the exact decimal place, our r-squared would be 100%. The closer to 100%, the better. The lower the RMSE, the better.
Takeaways
People were surprised to hear last week that Thomas had our lowest GPP Score, and he did so again this week. Let’s dive into why that was the case in both weeks.
The short answer is that people over-estimated him. The better answer is that people over-estimated him in two distinct ways, both of which contributed to his poor GPP Score. The best answer takes it a step further, identifying the fact that not only was he a bad play, but playing him also meant missing out on the opportunity to fade him, which was one of the easiest ways to gain an edge in both weeks.
The worst way in which people over-estimated him can be seen through our projections, where this past week he was:
- 9th in our expected strokes gained projection
- 17th in our top-5 odds
- 9th in our made-cut odds
I, of course, have no definitive proof that our projections were better than the industry’s, but the second part is the more important part of the argument anyway. Viktor Hovland was fourth in our projected strokes gained, fourth in our top-5 odds and fourth in our made-cut odds. If he had been JT’s price and ownership, would he have been good chalk?
No. Even Hovland only had a 20% chance in our simulations to finish top 5, and included a 12% chance to miss the cut (which would have been closer to if not higher than 20% if it were a regular cut instead of the top-50 & ties + within 10 strokes of the lead). This is the second way in which people over-estimated JT – they just don’t realize how wide the range of outcomes for golfers can be! You can quite easily make the case that no golfer ever, with the exception of Scottie Scheffler, whose probabilities “break the model” on a weekly basis, should ever be over 30%… let alone the 42% number he reached in the $33 single entry.
The GPP Scores perfectly reflect this concept, comparing ownership to odds of a ceiling and floor performance. In JT’s case, we had him at 10% for a top 5 (ceiling performance) and 16% to miss the cut (floor performance). This made him a bad play. What made him the worst play on the slate (by GPP Score and in my opinion) was the fact that if you played him, you obviously lost the opportunity to fade a massively owned bad play – something that wasn’t true of anyone else on the slate.
