Skip to content

Some new research could upset some long-held notions about hot streaks

Jul 17, 2014, 12:30 PM EDT


For as long as I’ve been following sports — at least following it closely as an adult — there has been big tension between what people say about players in hot streaks and what the data shows. Athletes and most people who cover them have accepted hot streaks — whether they be by shooters in basketball or hitters in baseball — as something that actually influence or predict future events. The stats, however — most notably studies of basketball players in the 1980s which led to the recognition of “the hot hand fallacy” — has strongly suggested that recent past performance in small sample sizes do not predict future events.

One new study, however, presents evidence that hot streaks have predictive power. From James Wagner at the Washington Post:

Green and Zwiebel studied two million MLB at-bats from 2000 to 2011. They neutralized for the abilities of the hitter and pitchers — such as lefty-on-lefty matchups and stadium sizes — and focused on 10 major statistical categories, such as batting averages, home run percentages and strikeout rates.

They found that a hitter’s past 25 at-bats were a significant predictor of his next at-bat. When a player is hot, they found his expected on-base percentage to be 25 to 30 points higher than it would if he were cold. Home run rates jumped 30 percent and strikeout rates dropped. For pitchers in hot streaks, future performance was improved, too.

They don’t reach any hard and fast conclusions here, though there are some which seem plausible. Mostly related, I believe, to a new understanding of what is and what is not “random,” as the hot hand fallacy is based on data related to random events. I’m certainly no statistician so I can’t judge either this or other studies in this vein on their merits with any degree of authority. Maybe this supersedes the last best statistical evidence on the matter. Maybe it’s flawed. I have no idea.

My personal takeaway, though, is that there is always something to learn about baseball. And that rather than try to understand it through opinions held based on personal beliefs, authority and predispositions, it’s better to understand it based on the data. Those who skew old school have always been a bit loathe to do this. Now, however, a favorite concept of the statistically-oriented is being questioned. I’ll be curious to see (a) if this new study holds up to scrutiny; and (b) if it does, how the stats folks take to having some long-held beliefs of their own challenged.

  1. [citation needed] fka COPO - Jul 17, 2014 at 12:36 PM

    Could be interesting, but we’ve heard this before. I’ll wait until the better math guys like Phil Birnbaum, MGL, Toirtap, or TomTango take a look at it. They’ve discredited these studies before.

    • moogro - Jul 17, 2014 at 8:39 PM

      The key thing to study is the “neutralized” filter used: what is in, what is out, and why.

    • tangotiger - Jul 17, 2014 at 10:03 PM

      We did.

      Here’s the thread we talk about it

      And here’s the response from the authors, and our discussion of it, including the authors responding to us:

      There’s over 200 comments in there. The big thing is to make a distinction between “streaks” and “recency”, which the authors do NOT differentiate.

    • hittfamily - Jul 17, 2014 at 11:22 PM

      If you go into the lab trying to “disprove” something based on personal bias, you probably can. I’d rather just live with the knowledge until another bias-less study comes out supporting the findings, or denying the findings.

      After all, 2% of scientists think global warming is a hoax. They looked at the data, and determined “LIES”. We don’t know how many of those 2% are currently being funded by the oil and gas industry, but I’d guess it’s pretty much all of them.

      • [citation needed] fka COPO - Jul 18, 2014 at 10:18 AM

        If you go into the lab trying to “disprove” something based on personal bias, you probably can. I’d rather just live with the knowledge until another bias-less study comes out supporting the findings, or denying the findings.

        I should have expanded my comment a little more. What I was referring to is people doing studies and drawing bad conclusions. For instance, posted a study about “gambling hot streaks”, and the accompanying paper concluded:

        …It’s about a study by academic researchers in London who examined the win/loss patterns of online sports bettors. The more wagers in a row a client won, the more likely he was to also win his next bet. That is: gamblers appear to exhibit the proverbial “hot hand.”

        The problem with the study, and I’m going to paraphrase significantly here, is that the researchers just looked at betting won/lose, and not how much the gamblers were winning. The reason? The gamblers were more likely to place a safer bet the more they won (and note, the reverse was true that losers were taking longer and longer odds). So if the just keep taking the safer bets, does that show a “hot hand?”

        A more detailed explanation (whom I paraphrased) is done by Phil Birnbaum here:

  2. stex52 - Jul 17, 2014 at 12:46 PM

    There is a lot of what would appear to be common sense in this. If a “hot streak” relates to be physically well, alert, feeling better than one has at other times, then you would expect a physiological improvement that would likely correspond to an improved result.

    I understand the immense power of statistics in relating to what is and is not important in predicting results. But sometimes a “small sample” – a single player batting – say 50 times over ten games- might relate very clearly to his health and physiological responses at that period of time.

    • natstowngreg - Jul 17, 2014 at 1:15 PM

      You’re right, statistical analysis can be very useful.

      In the end, it’s about a human being whose performance is affected often by what’s going on in his head. A hitter seeing a pitch as a beach ball or a golf ball. A pitcher gaining or losing confidence in his slider. A player distracted, worrying about a sick kid.

      Sabermetric extremists disdain such things because they can’t quantify them, and/or they have trouble dealing with people. Reactionary baseball types focus on them because they don’t understand the value of analysis. The best reporting leverages both. That’s something Wagner and Adam Kilgore do well in their coverage of the Nats for the Post.

    • paperlions - Jul 17, 2014 at 1:17 PM

      I agree with this. The small effects found in this study likely are related to the non-independence of player physical condition among plate appearances or pitching performances through time (i.e. a physically and mechanically sound player perform better than he does with nagging injuries or with inconsistent mechanics)….not because of some mythical hotness or coldness that is transient.

      Heck the effect could even be related to home vs. away games. On average players perform better at home than on the road, maybe because of travel related fatigue, the comfort of sleeping in your own bed, and the more familiar schedule that comes with home environs than when travelling. The article notes that they accounted for stadium effects, but doesn’t say they accounted for home-road splits, which are very real and well understood.

  3. ck101 - Jul 17, 2014 at 1:08 PM

    I can certainly agree with the idea that statistical work has brought into doubt the idea of the hot hand, and more particularly, how common it might be. But if you don’t think there’s something to the idea that a player can for any number of reasons, fall into some sort of groove that for a short period of time makes him better than he usually is, you ought to come up with a more satsifying explanation than randomness for things like Kirby Puckett’s weekend in Milwaukee in 1987 (over two games, he was 10 for 11 with four homers and two doubles). Every amateur, weekend athlete who’s golfed, or played basketball, or done anything athletically for any period of time has experienced short periods where everything was just easier than it usually is, and it’s hard to believe all of it is just random.

    The anti-hot hand case that’s stronger is that you don’t know when they’ll start or how long they’ll last, so they really have no predictive value. But to say that every discrete event is completely independent, and that a hitter who has made solid contact in 7 of his last 9 at bats is likelier to do so in his next at-bat than if he were mired in a 0-for-22 slump, seems ignorant of reality.

    • danaking - Jul 17, 2014 at 2:07 PM

      ck, you beat me to it. Any physical endeavor has hot and cold moments, no matter what it is. I used to be a musician, and there were days I just knew I could do pretty much whatever I wanted and it would be good. Easy, even. I was never able to induce that state, and it could leave at any instant, but there’s no doubt that, while I was in the zone, I was a better musician.

      I’m a follower of sabermetrics, read the Bill James Abstracts since the first Ballantine edition in 1982, and subscribe to Tom Tango’s web site. I’m coming to think the reason sabermetricians don;t like to acknowledge hot streaks is because the sample sizes are, almost by definition, too small to be statistically significant. This is one of those situations where, when asked why he went 4 for 5 with 12 total bases and a hitter says, “I was seeing the ball well today,” the wise option may be to defer to his judgment. He was there, he knows what he’s talking about, and the sample size is too small for statistical analysis one way or the other. (“Hot streak” can obviously be defined as more than one game.)

  4. spursareold - Jul 17, 2014 at 1:13 PM

    Crash Davis: I told him that a player on a streak has to respect the streak.
    Annie Savoy: Oh fine.
    Crash Davis: You know why? Because they don’t – -they don’t happen very often.
    Annie Savoy: Right.
    Crash Davis: If you believe you’re playing well because you’re getting laid, or because you’re not getting laid, or because you wear women’s underwear, then you *are*! And you should know that!

  5. paperlions - Jul 17, 2014 at 1:30 PM

    There are some interesting considerations, despite the authors contentions, these effects are actually tiny and much smaller than anyone that “believes” in hot streaks would have thought.

    Setting aside the inability to account for the lack of independence of the events in their study (i.e. player health and mechanics vary through time and plate/pitching appearances are not interdependent of them), statistical significance for a sample size of two million can represent a tiny effect…with that much data, almost any difference can be detected, even differences that are too small to worry about within the context of baseball.

    An OBP difference of 25-30 pts in the next plate appearance sounds big, but that is a 2.5-3.0% difference (one event every 33-40 PAs). None of the players quoted in the article would think that during a hot streak, they were expected to observe one more one positive outcome per 33-40 PAs….they would think it would be an effect about 10 times greater than that.

    The Home run effect is interesting. Home runs are hit about once ever 40 PAs for an average HR rate of 0.025. A 30% increase in HR rate means that during a “hot streak” a players expected HR rate is 0.0325, which over a season, would be a big deal (about 20 extra HRs)…but as presented it is difficult to tell if this is related to player health, mechanics, being at home versus the road (or some combination of those), or the hot streak itself (assuming a hot streak represents something other that the other factors).

  6. apkyletexas - Jul 17, 2014 at 2:15 PM

    Nerds. Watch the freaking game and put away your slide ruler and D&D dice.

    • clemente2 - Jul 17, 2014 at 2:57 PM

      Idiot. Put away your computer and drool.

    • stex52 - Jul 17, 2014 at 3:28 PM

      Thinking not allowed, huh?

      • paperlions - Jul 17, 2014 at 5:04 PM

        You can think. You just can’t support your argument with information. It let’s everyone be right and think that everyone that disagrees with them is an idiot.

    • Nick Danger - Jul 17, 2014 at 5:45 PM

      What the heck is a slide ruler?

  7. yahmule - Jul 17, 2014 at 2:36 PM

    Eh, small sample size…

    • paperlions - Jul 17, 2014 at 3:08 PM

      I don’t know who you have been talking to…but it is not small, it is average sized, maybe a little above average.

    • stex52 - Jul 17, 2014 at 4:57 PM

      It is still okay to observe and try to construct mechanisms for results. My general experience in many years of engineering is that statistics is a tool for when you do not have a defined mechanism for a process. If you can define a physical process, then go for that and use the statistics as a back up approach.

      In baseball you need stats because there are so many variables but the data base is available.

  8. hittfamily - Jul 17, 2014 at 11:26 PM

    I’d buy it. When I played, there were times when things were going so bad, you just hoped to put the ball in play. Then there were other times that you are furious because the SS snared a liner. Results change your mindset at the plate. On a hot streak, it’s alright to get to 2 strikes, because you know you can still rip one. On a cold streak, you better try to put every strike in play, because you don’t want to get to 2 strikes.

Leave Comment

You must be logged in to leave a comment. Not a member? Register now!

Top 10 MLB Player Searches
  1. H. Ramirez (2448)
  2. G. Stanton (2410)
  3. G. Springer (2389)
  4. C. Correa (2336)
  5. S. Strasburg (2329)
  1. J. Baez (2325)
  2. B. Crawford (2239)
  3. H. Pence (2235)
  4. M. Teixeira (2159)
  5. E. Encarnacion (1994)