I often find myself jumping on players whose price has recently fallen—both batters and pitchers, although mostly the former—due to poor recent play. It’s not that I think that some form of streaky play is impossible—whether it’s due to changes in confidence, health, or whatever. I think players’ anticipated “mean” production shifts due to a variety of factors—but that doesn’t mean we can use that information in any predictive way.
It’s really, really easy to get fooled by randomness in baseball. Is that player who is 9-for-12 in his last three games really “hot?” The answer is I have no freaking clue—I don’t think I can separate a signal from the noise in a very high percentage of situations—so I’d prefer to pay as cheap of a price as possible given the inability to make accurate predictions.
Hot and cold streaks are a vague sort of thing very similar to “momentum” in the NFL—there until it’s gone and gone until it’s there. But I do think we can and should make an attempt to quantify short-term fluctuations in randomness. If we can do that, we’ll have a better idea if a player has been lucky or not—if he’s hitting a ton of home runs because he’s really crushing the ball or if he’s just gotten lucky with placement, for example.
BABIP—Batting Average on Balls in Play—is a player’s batting average…on balls…that he hits into play. Shocker.
There are a few factors that can cause BABIP to fluctuate—talent level and defense among them—but the most important factor in determining a player’s BABIP is just luck. BABIP is a decent proxy for the luck a player has experienced because it can capture how often those cheap bloopers have fallen in for hits or how frequently hard-hit line drives have been straight at a defender.
Again, not every player’s long-term BABIP will be the same, but most players’ BABIP regresses toward .300. When a hitter’s BABIP is well below .300, it means that he’s probably been unlucky with batted ball placement—something that’s extremely difficult to control—and will probably see an improvement in overall numbers in the future.
Thus, for our purposes, a high BABIP is a bad thing, suggesting a decline in numbers in the near future. In general, it makes sense to target sluggers that we know can mash but just have a low BABIP. Once that variance levels out, they’ll post superior fantasy stats, even if they aren’t actually performing any better on the field.
HR/FB—home runs per fly ball—is similar to BABIP in that we’d expect players to have different numbers, even over the long run, but the short-term numbers are difficult to separate from randomness. That means we can use HR/FB, in conjunction with other numbers, to help determine how much of a player’s home run total is due to variance.
I like to compare a player’s HR/FB ratio in a given season to his personal average from the prior year or two. That way, we can adjust for his power level (to an extent), to help capture how lucky he’s been.For the average player, around 9-10 percent of fly balls result in home runs. But again, it differs based on the player. Whenever a batter has hit a home run on more than 20 percent of his fly balls for any significant period of time, though, we can expect some sort of regression moving forward.
Predictive Over Explanatory
There are a number of similar stats I like to use—Weighted Runs Created, Batted Ball Profiles, Plate Discipline, and so on—but the concept behind them is always the same: make accurate predictions. There are lots of interesting stats that can effectively explain past events, but not predict future ones.
In football, a team’s run/pass balance is one such stat. In general, running the ball is associated with winning. How many times have you heard “Team X has won Y straight games when running back Z gets 25 carries?” The problem is that run plays explain the past, but don’t help us predict the future; teams run because they’re already winning, not the other way around, and being bullish on truly run-heavy teams, as a rule, probably isn’t that smart.
Similarly, a batter’s average or home run total will help us determine how effective he’s been in the past, but they aren’t the most effective tools we can use to predict his future performance.
Ultimately, emphasizing the predictive over the explanatory reduces your risk. If you’re bullish on a hitter who is rolling—say, four straight games with a dong—you’re typically going to be overpaying for him in a worst-case scenario and getting what you pay for (long-term) in a best-case scenario. Bulk stats like home runs are priced into DraftKings salaries.
Meanwhile, if you buy low on a struggling player who has shown signs of improved play—a hitter who has been on the wrong side of variance—then you’re getting what you pay for as a worst-case scenario and finding sensational value in a best-case scenario.
If beating the daily fantasy baseball market is about maximizing the difference between anticipated production and price, then it will always make sense to emphasize predictive stats (a component of anticipated production) over explanatory stats (a major factor in determining price).