In my last article, I considered FanDuel golf scoring by exploring the concept of first- vs. second-half golfers, introducing the idea that if there are distinct differences in how golfers play between periods of a tournament then we should weight our stats and information accordingly to parse out those differences. In this piece, we’ll 1) go through how to do exactly that and 2) figure out which stats are more important for finding good first- vs. second-half golfers.
Weighting the Factors
Using something like Player Models for FD’s format takes some getting used to: In both cases you’re using the same data (Long-Term Adjusted Round Score, weather, odds, etc.), but you’re trying to predict two separate aspects of a player’s game. Confounding the problem is that these aspects aren’t unrelated: To put up good scores in either half, a golfer still needs to play well, and isn’t that what we’re trying to find generally with Player Models anyway? You can wrap yourself in a circle at this point, believe it’s too confusing, and just use the same weights for both the first and second halves. If you do, however, you’ll be forfeiting a massive advantage.
I pulled first- and second-half FD scores for all tournaments in 2016, and then fit the fields in Player Models to see which stats are most predictive for each type of score. I put the resulting weights on the same scale and took the difference of the first- and second-half weights to show which stats are more or less important for each type of scoring. Here’s how that differential plot looks:
The Three Key Stats
Here are my biggest impressions from that graph:
- Odds just straight up don’t matter for predicting first-half scores. I knew odds would be less important, but I wasn’t expecting them to drop off this much. Odds still capture some fundamental information regarding a player’s ‘overall’ ability, so I was expecting them to get some weight, but it appears that information is still captured by the more granular stats like LT Adj Rd Score. Odds are absolutely vital for second-half scores, since cut-making is much more important there, but I think there’s an edge in not incorporating odds for first-half stats.
- Weather becomes even more important. It’s not surprising at all to see wind and temperature on the left side of the graph. We already use weather to predict cut-making to some degree, but when points are accumulated only in the half that determines the cut it makes sense that this angle is magnified.
- Course experience matters more in the first half. This is the most intriguing finding. The second-highest differential comes from the number of times a player has played the course: Not how well he did or how often he made the cut, but just how many times he’s played. I can get behind this the more I think about it: In the early rounds, if a golfer is not familiar with the course then he’s still familiarizing himself with it throughout the tournament. By the time he’s figured out the quirks of a course, he might have already missed the cut — but he would have that experience to use for the next couple rounds he played there. That’s a pretty intuitive explanation for why course sample size matters. This may be my favorite hidden edge to exploit in the new scoring format.
The rest of the weights don’t stand out too much, and I don’t see much that I would disagree with, but the three factors above are more than enough to get you started as you build your own models for finding first- vs. second-half golfers.
Doing Your Own Research
Clearly, it’s a good idea for you to do your own research on this. When we incorporate FD data into our suite of Tools, you should experiment for yourself — particularly with our Trends tool — so that you can find the factors that yield the highest Plus/Minus values for first- vs. second-half golfers. Next, you should adjust the weights in your personal models, create tons of rosters via our Lineup Builder, and then tweak your models based on the first- vs. second-half golfers you’re able to fit into your lineups. Whenever FD golf starts, you should consult our DFS Ownership Tool to get a sense of which factors the DFS community is valuing most.
Next week, we’ll go through the full fundamentals of roster construction for this new format, including ways to work out our findings.
