clock menu more-arrow no yes mobile

Filed under:

A Less Deterministic View of DFS

“My optimal lineup.” We seem to hear that phrase a lot in daily fantasy sports, right? Many players create projections and values, then formulate what they consider to be an optimal lineup-a group of players that will maximize profitability over the long run-based on their values.…

“My optimal lineup.”

We seem to hear that phrase a lot in daily fantasy sports, right? Many players create projections and values, then formulate what they consider to be an optimal lineup-a group of players that will maximize profitability over the long run-based on their values.

Whether or not there’s really one true “optimal” lineup, it seems naive to think that we would have the ability to consistently identify the exact combination of players that could be considered “optimal” in some sense. That’s the goal, of course-we’re always striving for that-but we also need to plan for the inevitable fact that, hey, we’re just going to be wrong from time to time.

The ability to “optimize” a lineup is only as strong as the ability to accurately project players. And guess what? Projecting players is difficult. Some daily fantasy users are better than others, of course, but anyone would be foolish to expect perfection every time they create a lineup. It’s my belief that we can become the best daily fantasy players that we can be only after we accept and account for the fact that our knowledge is limited and we’re going to be wrong sometimes. I’ve been thinking about this topic for quite some time because I believe that many daily fantasy players, even some of the best ones, sometimes think and act in a very binary manner-as if decisions are either right or wrong and there’s no gray area. Once we move away from that style of thinking-once we embrace our own limitations as players-the way we approach the game should change.

Greater Diversification

The first inevitable outcome of embracing uncertainty is utilizing a greater range of players. Whether or not there’s one “optimal” lineup in a vacuum is irrelevant; all we need to be concerned about is our ability to identify that lineup. The less certain we can be that we’ve found it, the more player diversification makes sense.

I liken this idea to how we view injury-prone players. Are there players who are more likely to get injured than others? Probably. Are there some that we’ve already identified? I think so. However, the “injury-prone” label is a very black-and-white characteristic that probably isn’t so binary in real life. Further, we so often see players labeled as injury-prone after what could be a fluke injury or two, with no accounting for sample size issues or simple noise.

For the most part, I think it’s really difficult to identify injury-prone players, which subsequently means that I’m bullish on players who have been injured in the past. Maybe they’re injury-prone, maybe they’re not, but the fact that I can’t really tell means I’m generally going to take a chance on them if everyone else believes they can’t stay healthy. This idea is related to the second outcome of embracing uncertainty.

Limits the Value of $/Point

Some daily fantasy players create $/point values, projecting players and then calculating how many dollars they need to spend on a player for each point he can be expected to score. The problem is that there are inevitably always players near the top of the lists who are extremely close in value. Should we really be starting one WR over the other because of a difference of 0.2 points in our projections? At a certain point, we need to implement a margin for error such that we don’t rely on a system of picking players that is inherently fragile-very susceptible to wild fluctuations based on small changes in measurements.

Let me be clear that I don’t think $/point (or other value systems) are useless. They help you immediately identify a group of potentially valuable players and also eliminate guys who aren’t anywhere near the top values. We just can’t blindly follow the $/point values because we don’t have that kind of certainty in our projections. When we build in a margin for error, it again expands the pool of usable players.

Being Contrarian

A contrarian strategy is one that purposely goes against the grain. It’s popular in tournaments because hitting on a player who is low-usage is a really big advantage. The idea behind a contrarian strategy is that you’re giving up a little bit of value to create a unique lineup. Of course, most people calculate value as if it’s a certainty and then generally stick with their highest-value players. The merits of a contrarian strategy-one that emphasizes balancing value with expected player utilization-are rosier if everyone is treating their projections and values as flawless when they aren’t.

Less Weight Placed on Matchups

Matchups matter. In general, we want players who have perceived favorable matchups, especially in cash games. It would be a big mistake to simply start the best players in a vacuum, regardless of which team they’re facing.

However, just as our player projections can be flawed, so can our assessments of defensive strength. Sometimes the numbers are skewed one way or another, and sometimes good defenses just don’t play well in a given game. Note that this idea is more applicable in daily fantasy football than other sports. In football, there are sample size issues that limit our ability to completely trust what we observe. Compared to the 162-game MLB season and 82-game NBA season, the 16-game NFL slate makes it challenging to not only make overarching claims about players or defenses, but also to place a high level of confidence in those conclusions.

Emphasizing Long-Term Trends

I had some success playing daily fantasy baseball this year, due primarily to understanding that predicting how hitters will perform in a given night is really difficult. There are all sorts of ways to tilt the odds in your favor as a daily fantasy baseball player when selecting bats, including batter vs. pitcher data, recent performance, Weighted On-Base Average, and so on, but I’ve been the most profitable when I emphasize long-term trends.

The idea is that predicting the short-term is difficult enough that we’re probably better off siding with the long-term data in most situations. Instead of starting a lefty bat because he was crushing it lately and the wind was blowing out lightly to right field and whatever other information I could gather, I’d do it because he was facing a righty and historically crushed right-handed pitching. There’s a lot more data supporting that claim such that I could place more confidence in it, knowing I wasn’t getting fooled by something that might or might not be a factor in a player’s projection for the night. I think this transition from placing a ton of importance on very specific factors to generally seeking out long-term trends is a natural extension of embracing variance in daily fantasy sports.

Playing More Money

The gist of this article has been “We need to account for our own fallibility as players.” You’d think that might mean taking a more cautious approach to daily fantasy sports, but I actually think the opposite is true, for three reasons.

The first reason is that variance doesn’t equate to unpredictability. Even though baseball is a very volatile sport from night to night, it’s also extremely predictable over the long run; career .300 hitters have a very high probability of hitting near that number over the course of a season. Again, that’s why emphasizing long-term stats is vital; it helps overcome short-term swings.

Second, variance also doesn’t equate to less of an edge as a daily fantasy player. In fact, I find that the most volatile sports are also subsequently the most beatable because others have a tendency to get fooled by noise. Even though sports outcomes are far from random, even a random event can be beatable if there are exploitable inefficiencies in the way people approach it; a coin flip will come up heads 50 percent of the time over the long run, but you could still find an edge in predicting coin flips if you could get the right odds, i.e. better than even money.

Finally, and most important, you can play more money by embracing uncertainty because it opens up the value of hedging. Remember, accounting for your fallibility opens up the pool of available players, which in turn improves the expected value of more lineups. The key is that your lineups are +EV, i.e. they will be profitable over the long run. As long as that’s the case, hedging with multiple lineups can make a lot of sense because it reduces your downside.

Let me give you an example. If you were to play only one lineup and place 10 percent of your bankroll on it, that would be much riskier than playing two lineups-one with 7.5 percent of your bankroll and the other with 5 percent-even though you have more money in play in the second scenario. In the latter case, you’d be trading in exposure to your “optimal” lineup in exchange for safety, i.e. reducing your risk of ruin.

Well, the greater the pool of available players and the less confidence you can place in what you believe is your “optimal” lineup, the more exposure you can have to what you think are your second and third-best lineups. Ultimately, embracing uncertainty allows you to ironically play more money (without increasing risk) and still maintain a positive expected value.