I started playing the NFL games on FanDuel and DraftKings last season. It’s needless to say that, as a first-timer, I didn’t finish in the Top 1.3%. I did, however, have much less failure on FanDuel (FD) than on DraftKings (DK), and that got me thinking — and digging into the data — about why. The answer, it turns out, is one of our oldest statistical friends: variance.
As an intermediate-level poker player to this day, the idea that variance influences game selection (and game outcomes) is familiar. The concept of variance should also be familiar if you know anything about NFL game theory. In (almost) all cases, the more inherent variance in a game (i.e., the larger the point spread and the more disparate the head coaching track records), the worse favorites perform, and the better underdogs perform. If I’m a poker favorite, I should mostly play cash games, which have lower variance than sit-n-go’s and multi-table tournaments. If I’m an NFL favorite, I should adopt a conservative game plan, which has a lower variance than a more-aggressive game plan. And if I’m an above-average daily fantasy sports (DFS) player, I should mostly play head-to-head games or 50/50s, which have lower variance than guaranteed prize pool tournaments.
As it turns out, if I’m an above-average DFS player, the concept of variance also suggests I should play on FD more than DK. The rest of this post is a basic statistical explanation for why.
FanDuel Variances vs. DraftKings Variances
I’m just going to come out and say it: At every major lineup position, the week-to-week variance in DK points is higher than the week-to-week variance in FD points. My evidence comes from all regular season NFL games since Week 1 of 2012 for quarterbacks, running backs, wide receivers, and tight ends.1 My only other inclusion criteria for this sample were that players (a) played on the same team for the duration, as is my wont on I//R, and (b) ranked among the Top X at their position as follows:
- For both FD and DK, 20 quarterbacks and 20 tight ends.
- For FD, 40 running backs and 60 wide receivers.
- For DK, 50 running and 70 wide receivers because of having a flex position and no kicker.
Given the above criteria, the table below provides the evidence showing that, all else equal, DK games are more susceptible to variance than FD games:
|Pos||FD Avg||FD SD||DK Avg||DK SD||FD CV||DK CV|
As you can see, the standard deviation (SD) at all four positions is higher for DK than for FD. And because variance is defined as SD-squared, all four positions have a higher variance in DK games than in FD games.
And for statistically savvy readers, even if we use the coefficient of variation (CV) to account for the fact that larger averages tend to produce larger SDs, it’s still the case that scoring for quarterbacks and running backs is more variable on DK than on FD.
So, if you’ve been wondering like I have why you’re losing on DK despite being an above-average season-long fantasy football player and a winning player on FD, it’s because higher-variance DK tends towards more-random results while lower-variance FD tends towards less-random results.
DT : IR :: TL : DR
- Whether we’re talking about the NFL, poker, or DFS, game theory says that favorites should adopt low-variance strategies, while underdogs should adopt high-variance strategies.
- Since 2012, the four major positions on FanDuel have produced lower week-to-week variance than on DraftKings. This is especially true for quarterbacks and running backs.
- Taken together, both of the above suggest that better (or more veteran) DFS players are better off focusing on FanDuel, while worse (or more novice) DFS players are better off focusing on DraftKings.
I’ve included the first two games of 2015, but their impact is negligible considering their proportion of the total sample. ↩