MLB All-Star - Lesson 02 - Hot and Cold Streaks

Jonathan Bales is the author of Fantasy Baseball for Smart People—a 197-page book created to help you win playing daily fantasy baseball on DraftKings. Making predictions is difficult. There are all sorts of variables that go into selecting players in daily fantasy baseball, and your ability…

By DK Playbook May 31, 2020, 9:00pm PDT

If you buy something from an SB Nation link, Vox Media may earn a commission. See our ethics statement.

Jonathan Bales is the author of Fantasy Baseball for Smart People—a 197-page book created to help you win playing daily fantasy baseball on DraftKings.

Making predictions is difficult. There are all sorts of variables that go into selecting players in daily fantasy baseball, and your ability to assess those variables to make accurate predictions will determine your success.

Get the best DraftKings Daily Fantasy Deposit Bonus from the DK Playbook – 100% Deposit Bonus up to $600

A solid daily fantasy baseball strategy to increase your accuracy rate is to identify predictive stats. Not all stats are predictive. Some explain past events quite well, for example, while some actually aid in making future projections. Our concern shouldn’t be with how well stats explain particular phenomena, but rather with to what degree they can aid us in predicting baseball performances.

I believe that hot and cold streaks, for the most part, are explanatory, not predictive. That is, it’s certainly possible—perhaps probable—that players go through actual streaky play that isn’t just randomness, i.e. they’re “seeing the ball well” or “throwing lights out” or whatever. However, it’s one thing for streaks to exist and another for us to use a past streak to predict future play.

Understanding Streaky Play

Let’s say that you have a lot of time on your hands (like me) and you decide you want to roll a die a few thousand times. You’re going to do that to simulate a batter’s fortunes at the plate: roll a 1 or 2 and that’s a “hit,” roll anything else and it’s an “out.” Within those few thousand rolls, you’ll see all kinds of “weird” occurrences that resemble streaks, but are due to nothing but randomness.

I actually did this 25 times to give you a sample distribution of 25 completely random “plate appearances.” Rolls of 1 or 2 were hits (H), and everything else was an out (O). This would reflect a hitter with a .333 batting average.

H, O, O, O, O, O, O, H, H, O, H, O, O, H, H, H, H, O, O, O, O, H, O, H, O

This might be five or six games for a batter. Even though we’d expect one-third of all rolls to be a “hit,” this batter actually hit .400 over the 25 rolls. If this were real life, we’d probably hear something about how the batter is “hot” because he’s hitting .400 over his last 25 at-bats. Even within the 25 rolls, we see an early string of six straight outs—a mini-cold steak—as well as a string of four consecutive hits.

I’m not saying baseball is entirely random—it’s not—but if it were, we’d still see long “streaks” that we’re inclined to say are hot or cold play, even if they’re due to nothing but good or bad luck. With the amount of data we have in baseball, it’s inevitable we’re going to see strings of weird results for no reason at all.

So the question is “Can we separate the signal from the randomness?” And I’m not sure we can. Lots of studies in the past have shown that past hitting streaks don’t predict future production very well.

Again, that’s not to say that streaky play can’t exist—just that it might not be useful for fantasy purposes. It’s very similar to the idea of injury proneness. We’d expect that certain players are naturally more injury-prone than others, but it’s very challenging to use past injuries—which are difficult to identify as the result of being injury-prone or just poor luck—to predict future ones.

Identifying Actual Cold Streaks

As far as streaky play goes, I don’t think we need to put too much research time into trying to identify it or use it to predict future fantasy points. However, I think there’s probably a better chance that cold streaks exist (and are predictive) than hot streaks. The reason is because of injuries. When players are hurt but try to play through the pain, we’d naturally expect them to play worse.

One way to determine how likely it is that a player is going through a cold steak is to examine how “lucky” he’s been. If a batter is cold but seems to be hitting the ball well, for example, we can be reasonably confident that he’s just been unlucky.

I like to analyze BABIP—Batting Average on Balls In Play—for both hitters and pitchers to help determine their recent luck. For most players, BABIP will regress toward .300 over the long run. If a hitter is on a cold streak but has a low BABIP, it’s a sign that he’s just been unlucky and his future production should improve from better luck alone.

The situations to potentially avoid are when players have been performing poorly, but also haven’t been getting unlucky. When a batter is hitting .100 over his past 10 and also has a BABIP at or above .300, that’s a bad sign. It means that his lack of success is probably due to his own play and not the result of bad luck—and thus probably a bit more likely to continue in the future.

Buying Low

The final thing to consider as it relates to streaky play is that daily fantasy baseball exists as a marketplace, so we’re not concerned only with anticipated production, but also with cost. DraftKings often raises the price of players who are on “hot streaks” and lowers the salary of those who are in the midst of “cold steaks.” That means that, even if hot streaks were predictive, we might still want to target players who aren’t playing well of late; it all depends on pricing and how much DraftKings compensates for recent play.

Due to the amount of randomness that appears to exist in baseball outcomes—specifically for hitters—it makes sense to be wary of “hot” players and be more bullish on “cold” ones. If streaky play isn’t at all predictive—which is close to being the case—then a player’s hot recent play won’t have any effect on his future production. That means that, if you’re paying top-dollar for a player whose salary has risen because he’s crushed the ball in recent games, you’re likely overpaying. And vice versa, cold players can often offer value because their production is likely to experience positive regression.

I’m more likely to “buy low” on cold hitters than cold pitchers, however. The reason is that batter performance is much more volatile from game to game. A hitter has just a handful of swings during a game, so he can be hitting the ball quite well and still see poor results over a pretty large sample of games.

On the flip side, a pitcher who is throwing the ball well is very unlikely to string together, say, five straight poor starts. With 100-plus pitches thrown in many games, pitcher performances are more reflective of actual performance than hitter performances. That means that, although it’s still challenging to identify true streaky play in pitchers, it’s easier than with hitters. Though I’ll still sometimes buy low on pitching, the reduced volatility at the position means that it just makes more sense to play it safe with your arms and really buy low on struggling hitters.