Behind the Scoreboard May 16, 2009
Do Hitters Change Their Approach During a Hitting Streak?

As I am sure everybody has heard, Ryan Zimmerman recently completed a 30-game hitting streak, putting him alongside 52 other players in the history of baseball who have hit in 30 consecutive games or more. The streak was a nice bright spot in another otherwise dismal season for the Nationals, but unfortunately, as so often happens, the streak ended right as he began to gain national recognition for hitting in 30 straight games.

Thirty games is a kind of marker of when a streak really becomes serious. The national media start paying attention, the fans begin to invest in it, and the pressure really starts to build for the player. Nobody takes notice of a 10-game streak and few recognize a 20-gamer, but when player he approaches and reaches 30 games, it becomes serious. My question is how having a serious streak changes a player's performance. This will be the last in a three-post “streak” of posts about streaks.

As I said earlier, there have been 53 such streaks in the history of baseball. First, let's take a look at how they break down.

As you can see, a lot of the streaks were broken up after either the 30th or the 31st consecutive game. Is this some evidence that players buckle under the pressure of a high profile streak? Let's try to fit a theoretical model to the data and see if theory fits reality, or if something else may be going on. One would think the data would take shape of the following: y = 53 * (a ^ (x-30)), where x is the length of the streak, y is the number of players completing an x length streak, and a is some coefficient to be fit by the model, basically representing the probability of continuing the streak an additional game. When the model is fit, we find a=.755, meaning that all things being equal, the probability of continuing a 30+ game streak an additional day is about 75.5 percent. Let's look at the graph to see how well it fits.

As you can see, the model slightly over estimates the chances of extending the streak to 32 games, but underestimates the chances of extending the streak to around 40 games. It also predicts an exceedingly low probability of a 30-game streak reaching 56 games (1 in 1500), when in fact one did occur. Is this just an artifact of chance, or is there something going on? Largely due to DiMaggio, a chi-square test strongly rejects the theoretical model. However, if we put DiMaggio in a "46 games and over" category, we still find that the model fits only marginally well, with a p-value of .12.

The difference between the real data and the model is basically that instead of having the same probability of continuing the streak in each subsequent game, the real data seems to suggest that probability of continuing the streak gets higher as the streak goes on. However, this is largely due to the fact that a player who has hit in 40 consecutive games is likely to be better than one who has hit in 30 consecutive games. Indeed, the three-season average of hits/plate appearance was .287 for those who had streaks between 30-34 games, but the H/PA was .301 for those who had streaks of 35 games and higher. So while an unknown player with a 40-game hit streak is more likely to extend his streak than an unknown player with a 30-game hit streak, it's unclear whether the chances differ for a specific player.

Perhaps more interesting is to compare players' overall statistics to their performance while they have a 30+ hit streak brewing. Do players change their approach to try to extend the streak? If so, we might expect to see less walks and less homers during the pressure filled portion of the streak. Also, what happens to their overall batting average? Obviously, these players are "hot," but they are also under a lot of pressure and pitchers may especially bear down to try to get them out. The comparison in BAV, HR/PA, unintentional BB/PA, and IBB/PA are listed below (the expected values are based on three-season averages weighted for the number of games over 30 that each player had his streak going).

As you'd expect, players with streaks have a good batting average at .303. They also don't homer much, clocking in at .022 HR/PA, which is good for 13 HR's over the course of a 600 PA season. As you'd expect, they also don't walk much, as they're OBP is only .358 - not great considering a .303 BAV. Below, we can take a look at the comparison of total stats vs. the stats during games in which a player had a 30+ game streak (including the game where the streak was broken).

Perhaps the most surprising thing is the fact that the batting average during games where a player has a streak going is 20 points lower than his usual batting average. This statistic has a high standard error (.026), so due to small sample size, it's hard to prove anything with statistical significance, but it's still interesting. It certainly blows a hole in the "hot hand theory" (which of course has been done many, many times before), but does give some credibility to the theory that pitchers are bearing down and/or pitching carefully to hitters with a streak.

The batters, for their part, seem to be taking 25% less walks with a streak on the line (again, not enough power to prove statistical significance however), which is good for the streak, but bad for the team. However, if it comes down to the late innings and the player is without a hit, it makes sense that he would hack away when a walk would likely end the streak.

While it might be sporting to avoid intentionally walking a player with a long hit streak, pitchers don't seem to care, and actually intentionally walked hitters more than expected (where we expected 2 IBB, there were actually 6). So, while Nats fans may have been angry when Zito intentionally walked Zimmerman in the 7th inning with the streak on the line, it wasn't unprecedented. The increase in intentional passes is probably due to a perception that the player with the streak is "hot" and more dangerous than usual, even though a look at these statistics will show that pitchers have no reason to worry.

While more data (and thus more streaks) are needed to draw hard conclusions, preliminary evidence shows that players may have a tough time getting hits with the streak on the line, and while their power remains the same, they do tend to walk less. So while it may be exciting to watch your favorite player with a long hit-streak, there is some evidence that the effect may not be as positive for your favorite team.

I heard Tommie Agee once had a 22-game streak with the Mets where he had just 23 hits. If that's true, Agee could have had an OBP of .300 or less during the streak if he also had his usual low number of walks.

Very nice analysis. I'd like to see actual versus "expected" wOBA or OPS to see how much if any players are hurting their teams by apparently trying to continue their streak. I'd also like to see how these numbers change as the streak goes on - i.e., at what point do players and pitchers appear to change their approach to continue the streak or not, or do those changes in approach gradually increase, etc.

It is of course "sad" (for the opposing team) that these players are IBB'd so often when in fact they are likely worse hitters as they appear to alter their approach for the worse.

Sky I like how you point out differences and then merely state that they do not necessarily rise to the level of (arbitrary) statistical significance, rather than reject any differences that are not "statistically significant." That is a bugaboo (among many) of mine - rejecting numbers that do not meet some arbitrary level of statistical significance. Stating the difference along with the standard error, as you did, should be all that a researcher needs to and should do. I don't know where this convention of "accepting" or "rejecting" an hypothesis based on some arbitrary level of statistical significance came from, but it is a bad idea. The fact that some use 2 and some use 2.5 is a clue that something is wrong with the concept. I guess the concept comes from social science experiments where the researcher is "supposed to" do one of two things: Either accept or reject an hypothesis. As if there is not other alternative. In reality of course, the correct alternative is to simply assign a level of certainty to one hypothesis or another, or to recognize that the hypothesis or conclusion is often not an "either/or" but an effect along a continuum (from no effect to a large effect)...

I enjoyed this article. I found myself this evening just looking over some links I hadn't visited for awhile and happened on this site which I had bookmarked several years ago (I'm not sure but I think it was just after I completed "Money Ball" which made me more interested in this type of research.

I am not sure of the source of your data on streaks or if there is reliable data for somewhat shorter streaks (lets say from 25). My first take is that you are analyzing the tail of the distribution of all streaks by only looking at the tail. I don't know how many many hundreds of thousands of streaks there have been in the history of baseball. We do know that the range for the length of a streak is between one and fifty six. It would be very interesting to me to see a graph of all of the streaks (I would understand that this is unlikely and maybe the data simply isn't available). If a curve were fit to that data, I wonder what the portion from 30 to 56 would look like.

Thanks for a very interesting analysis.

Thanks to all for the nice comments. I think it really would be useful to get data for the 25-game streaks and above since that is probably closer to the point where players start thinking about them. However, I could only find lists of 30+ streaks - I suppose it would be possible to use retrosheet and identify shorter streaks. Doing so would give a lot more power due to the increased number of PA's in the sample.