Ignoring conventional wisdom

Red Sox vs. Orioles
Since the introduction of Sabermetrics many concepts and terms have started to enter the main stream. OBP is now a regular on the NESN broadcast and OPS is often a common term in any presentation. That has not lead to a change in much of the conventional wisdom that we have started to be seen as not true. Just watch a few minutes of the ESPN Wednesday night game and by the third time you hear Joe Morgan call for a bunt you'll know what I'm talking about. One common misconception to me that would make player analysis much better is the understanding of streaks and slumps. To put it quite bluntly there is no such thing as a streak or a slump. There may be peaks and valleys in any given season, but that doesn't mean your doing anything different. Bill James wrote about this in one of his earlier papers Understanding the Fog. He said that there was no hot or cold streaks, but simply a cluster of events that made it appear to be on or the other. The point was the placement of them was completely random.
Red Sox vs. Orioles

Since the introduction of Sabermetrics many concepts and terms have started to enter the main stream. OBP is now a regular on the NESN broadcast and OPS is often a common term in any presentation. That has not led to a change in much of the conventional wisdom that we have started to be seen as not true. Just watch a few minutes of the ESPN Wednesday night game and by the third time you hear Joe Morgan call for a bunt you’ll know what I’m talking about.

One common misconception to me that would make player analysis much better is the understanding of streaks and slumps.  To put it quite bluntly there is no such thing as a streak or a slump.  There may be peaks and valleys in any given season, but that doesn’t mean you’re doing anything different.

Bill James wrote about this in one of his earlier papers Understanding the Fog.  He said that there was no hot or cold streaks, but simply a cluster of events that made it appear to be on or the other.  The point was the placement of them was completely random.

Another way to think about this of course with all statistics is gambling.  No matter the game you play; poker, blackjack, roulette, etc, you have a certain chance of winning and losing.  If you win or lose a larger number of chances on a “run” it’s inevitable that you will return to normal averages if you continue to play.  You could flip a coin 1000 times and while you may have several periods when you get the same result in the end you should be very close to 500:500.

There are many reasons to keep this in mind.  The first is to never assume a player is one end or the other when a streak happens.  Jason Bay was neither the player who had an OPS of .689 in July or a 1.123 in April.  He was the player who finished with a .921 OPS in 2009.  This is very important for talent evaluation and fan expectations.

The next reason is making managerial player decisions based on things like slumps and matchups.  Over the past few years fans have had access to more numbers than ever and some of the worst numbers are things like last 7 games or matchups.  If a player has hit over .500 with an OPS over .600 in his last 7 games it will have no effect on the next game or AB.  The same goes for someone who has gone 5 for 5 against the current pitcher.

As a scientist I regularly frequent the Science Blogs and I recently found this post from last year on The Frontal Cortex blog there.  The central argument is actually looking at basketball and the hot hand, but also takes a look at Joe DiMaggio and The Streak.  While many think this was so far outside the normal variance of results there is evidence that this length of a streak should have occurred in the history of the sport at some point.

The streaks ranged from 39 games at the shortest, to a freakish baseball universe where the record was a remarkable (and remarkably rare) 109 games.

That doesn’t mean that the streak was a sure thing as most of the streaks occurred before the 40′s and an amazing 300 of the streaks were accomplished by a fictional Ty Cobb.  This is just showing that it was something we should have expected and wasn’t a sign of increased talent during the streak just an extreme “run” of positive outcomes.

Perhaps I should mention that while I don’t believe in the traditional slump or streak there are always outside forces.  Bay also was in a slump in 2007 that lasted nearly his whole season, but that was attributed to injury.  This is usually the case as a slump could have evidence to explain it, but a “hot” streak cannot.

So, what can we learn from this?  That no player should be labeled as “streaky” as though it’s a bad thing.  There might be players who are a bit more variable, but no one is really streaky.  A player who relies on the three true outcomes (homer, walk or strikeout) might appear more “streaky” as his perceived value results mostly from homers.  If these homers are bunched then his value to the team appears more “streaky”.

In conclusion we can’t stop Boston.com from labeling every time someone goes 4 for 4 in back-to-back games as a “hot” streak, but we can realize what is actually going on.  The streak is nothing more than random variation and we need to look for the true talent level.

Categories: Jason Bay

After taking an interest in sabermetrics and statistical analysis Troy began trying to use it to an advantage in fantasy baseball. He started the website RotoSavants.com and also spent time at HardballTimes.com and FantasyPros911.com. After a few years the interest in the Red Sox drew him to start a Red Sox-oriented site (Yawkey Way Academy) with fellow writer Lee Perrault. A short time later he joined Fire Brand. Writer from: December 14, 2009 – July 24, 2010, March 3, 2011 – May 10, 2012.

38 Responses to “Ignoring conventional wisdom” Subscribe

  1. jim January 2, 2010 at 9:12 AM #

    I don't understand this at all, and the links you cite prove nothing about baseball hot streaks, and some of the basketball comments suggest newer research has cast doubt on the basketball claims. The reason coins do not have hot streaks is that they do not get tired or sick or injured or lose confidence or and do not have families or girlfriends or sick kids. Most important, flipping a coin or even shooting a foul shot is much simpler than a baseball swing, and many players, unlike coins, develop bad habits with their swing that need to be corrected, sometimes because of tiredness etc. This seems like very casual "science."

    • TroyPatterson January 2, 2010 at 11:04 AM #

      This is what I am talking about. You need to prove what you just said and not make claims based on what you think should happen.

      I can't give you the study from The Book since that is a copyrighted work, but here is a partial explanation from Dave Cameron at USS Mariner.

      Also here is a study by Sky at Baseball Analysts on players in the middle of extensive hitting streaks.

      I agree that a player could slump based on outside influences and this is something I try not to forget as a statistical analyst, but there is little proof that this change is significant.

      The coin flip was a simplified example to prove a point. You could think of a game of poker as a single at bat. Poker requires skill and talent to know what hand to play, what hand to fold and how to bluff. You only have a certain chance of getting each winning hand and have to play with what you are given. It's possible to get three aces for two hands in a row, but that doesn't mean you should expect another on the next hand.

      • jim January 2, 2010 at 1:47 PM #

        I don't have to prove anything; I didn't write a blogpost on firebrand. You made a series of claims and provided very poor arguments. The coin flip example was a terrible example and certainly does not "prove a point." The poker comparison is not as bad, but is still pretty lame. Of course a sane person would not think that being lucky in a couple of poker hands means you will be dealt good cards in the next hand. But if poker were all luck then I could beat the best players half the time. The parts of baseball that are luck can certainly vary and cause parts of slumps and hot streaks. But the claim that luck is the cause of all hot streaks is a claim that needs better support than lame analogies to gambling. I've only had time to look at the first of the links you supply, but the first one seems fairly modest in what it "proves" which is that once you notice a guy is on a hot streak there's a good chance it will end–which means that being hot or cold is not predictive. This is not the same as "there is no such thing as a streak or a slump." Can you address the claim in the comments section of your first cite–that there is some new research that contradicts the claims in the article?

        • Daern January 2, 2010 at 8:26 PM #

          Your claim about poker is incorrect–with all the variables in play, even if the game was all luck you could not conceivably expect to beat a pro 1/2 of the time. Same with baseball; a lot of luck involved in it, but the amount of variables mean that a 50/50 split is unrealistic.

        • Aaron January 2, 2010 at 9:34 PM #

          Jim, I think the predictivity of the streak is really the key here. If a hot streak does not mean that a player will do well in near future then what does a hot streak mean?

          • TroyPatterson January 2, 2010 at 10:16 PM #

            exactly. If a 20 game hit streak doesn't effectively predict what will happen in game 21 then what does it matter.

            You can label it a Hot Streak, but to automatically assume it is a sudden increase in skill, but that he then returned to his normal skill makes no sense.

    • Aaron January 2, 2010 at 12:00 PM #

      The chapter on streaks in "The Book" starts on page 53, definitely a good read on the subject. The question to me is whether a player's "hot/cold hand" is predictive and it seems that from the studies that have been done streaks are not statistically significant when making predictions about a player's future performance. That doesn't mean that there aren't some streaks that are predictive (injuries, other stresses, changes in approach) but those streaks are a small percentage of all streaks and need to be identified on a case by case basis. In other words, it's not the streak itself in those cases that is predictive but rather the injury or change in approach or whatever the other factors are in play that are the predictive factors.

  2. Ryan Hoffman January 2, 2010 at 12:18 PM #

    I would say that a hot streak is possible for a hitter based on something they fix in their swing or stance. They can actually improve based on this the same way a pitcher improves when the pitching coach notices something in their delivery. If you think of it this way, a hot streak could be something that accidently or intentionally gets fixed in a batter's swing, and a cold streak is when something goes amiss. Streaks can happen, but these extremes runs mentioned in the article also can.

  3. Trevor January 2, 2010 at 3:22 PM #

    What about Adam LaRoche, who always plays worlds better in the second half?

    • Aaron January 2, 2010 at 3:51 PM #

      A. LaRoche has a career BABIP of .289 in the first half compared to a second half BABIP of .342, which can account for much, if not all, of his splits. I know that your question was more general so anyone who has seen a study about first half/second half performances of all players can perhaps answer you better.

      • Trevor January 2, 2010 at 5:10 PM #

        Do you know of any studies on BABIP? It's an interesting stat to me, and I can generally agree that it is a good indicator of luck, but for some players there seems to be exceptions. In this case, LaRoche consistently seems to post these weird splits. Is it just an anomaly? Some pitchers, too, thrive on keeping their BABIP low consistently, like Mariano Rivera. There must be more to BABIP in some cases than pure luck, right?

        • Aaron January 2, 2010 at 5:56 PM #

          I'm not an expert but when I just searched for "BABIP studies" I hit some pretty interesting looking stuff – "… David Gassko's study detected a predictive relationship between strikeouts per game, home runs per game and BABIP. An explanation for these …," "… This has always been the difficulty with BABIP. For pitchers, this is the reason some early studies found no control over BABIP. …," " … To control for these factors in Davenport's study is to say: Let's see if good pitchers have lower BABIP than weak pitchers, but first let's …"

  4. Cliff January 2, 2010 at 5:10 PM #

    This article is the kind of thing that causes people to reject sabermetrics. Just because an activity mimics probability does not mean it is solely governed by probability. To compare hitting a baseball to flipping a coin or games of chance is a total misrepresentation. Players are not genetically branded with a set average as coins and cards have set probabilities. While I agree with the concept that most "slumps" can be caused by numbers evening out, you devalue your whole argument by bringing in improper metaphors.

  5. Patrick MacKenzie January 2, 2010 at 5:54 PM #

    It appears to me that the theorists who trivialize the play on the field are people who didn't play the game at all, or if they did not at the highest level. Those who have played the game and understand the way it works on the field KNOW for a fact that they fall into ruts, slumps, whatever you want to call it.

    • Daern January 2, 2010 at 8:28 PM #

      So you played at the highest level? How was your baseball career?

      Without backing up your argument, how do you expect anyone to buy it? The statistical laws of chance are well-documented; the superstition of streak is not.

      • Aaron January 2, 2010 at 9:10 PM #

        No "Patrick MacKenzie" that I could find in the Baseballreference.com or fangraphs.com databases.

    • Aaron January 2, 2010 at 8:47 PM #

      I've never heard of a theorist who "trivializes the play on the field." We all love baseball here.

      Also, you must agree that those who haven't studied advanced statistics shouldn't really be commenting, right? Since they don't really KNOW what it's like to understand a statistical theory.

  6. Joe Veno January 2, 2010 at 6:44 PM #

    When I see JD Drew hit 12 homers in a month, or Manny go through phases where he is a better hitter because he shoots for the opposite field more, it makes me less inclined to accept these "theories."

  7. tom January 2, 2010 at 7:08 PM #

    the thing is, that when i have been (on a streak) things slow down for me and i have (more time to read the pitch). obviously i dont have more time to read the pitch but it feels that way and im more confident and the same goes the other way, when im cold i have less time and less confidence. now to put a string of hits together is a different beast and one that supports your claim, i can be cold and have great contact with the ball and i can be hot and have horrible contact. so to say that there is no slumps isnt very accurate because there are many influences that go into all of this. perhaps a batter hits a certain pitcher well is because of the way the pitches presents his pitches. i always fond that the delivery of a pitcher made it easier or harder for me to hit them.

    • Aaron January 2, 2010 at 9:01 PM #

      This is an interesting thing to look at, Tom, because athletes all know what it's like to be in the zone. But when you look at the numbers "the zone" doesn't appear to exist, or rather, doesn't exist as a predictor of near future events. To me that raises a great question, which is always a good sign. Right now it seems like the experientialists are saying the math is wrong because of their experiences and the mathematicians are saying the experientialists are wrong because of the math but that seems like a pretty silly argument. Until I can find a mathematician who says the math is wrong I'm not going to discount the math and I'd love to hear a psychologist comment on the human ability to accurately judge one's own experiences which would shed a lot more light on the arguments of those who are discussing their own experiences of "the zone."

      Any psychologists out there with a background in human self perception?

      • tom January 3, 2010 at 12:11 PM #

        the other thing to take into consideration is career years, if math says that it will all equal out then how can you get a career year, with 500+ at-bats the math will equal things out. also, if albert pujols didnt get pitched at like it was ok to walk him all the time would his batting average go up, down, or stay the same. if you are ok to walk a batter you will always paint the corners and make it harder to hit with good contact. this has to change the equation because it brings in variables that are not in the batters control. i guess my point is that i believe in math as answering questions but this if very short sided in this article. an era doesnt tell the story for a pitcher and in just saying that there is no streaks is limiting this discussion. i do believe this article is onto something but does need further investigating.

  8. Cliff January 2, 2010 at 10:44 PM #

    My other question is why should we not characterize certain players as streaky? Is there another word to describe them better? You would have to say that certain players like JD Drew and Jason Bay tend to accumulate their statistics in shorter periods versus say, Youkilis or Pedroia, who have relatively even contributions throughout the year. This is neither a positive nor a negative but is still a characteristic nonetheless.

    • Aaron January 2, 2010 at 11:57 PM #

      You're probably talking about counting stats too but I took a quick look at the four guys you mentioned and calculated the standard deviation for their monthly (and again, you might be talking about smaller periods of time but doing it monthly was much easier) career OPS. Youk – .083, Bay – .070, Pedroia – .064, Drew – .054.

      So Youk seems to see the biggest fluctuations in his monthly performance and he's the only one with a standard deviation more than ten percent of his career OPS. That seems like a pretty small standard deviation to me.

      • evanbrunell January 3, 2010 at 12:13 AM #

        I do think that there's something to be said for hot and cold streaks, and I speak for experience. I will, however, caution that the "effect" of said streaks is, for the large part, small. Most streaks can be ascribed to luck and random variation — assuming no outside influences (sickness, etc.).

        • TroyPatterson January 3, 2010 at 1:52 AM #

          I think Evan is right here and that I'm not trying to say you can't label a good group of games as a "streak", but that this information should not be used to set lineups or make trades. There is no predictive nature from a slump or streak in and of itself.

          Aaron was that standard deviation for career or 2009?

          • Aaron January 3, 2010 at 1:07 PM #


          • Aaron January 3, 2010 at 2:45 PM #

            Also, just a caveat, those numbers aren't weighted. So some of these guys have had more at-bats in certain months which would effect the numbers a little bit. I just wanted to do a real quick look at those four players.

      • Cliff January 3, 2010 at 2:52 AM #

        Yea I was mostly speaking about the old man stats. OPS seems to even out a little better, but things like average, doubles and home runs seems to come in flurries for JD and Bay where as you can't point to one great month for Youk or Pedroia.

        Once again, it's neither a positive or a negative as I don't think this trait should be considered more or less valuable. I just think it's another adjective to describe a player.

  9. Len January 3, 2010 at 1:33 AM #

    Could there be, to prove a point, a massive undertaking to see how a player performed considering his last at bats? How often he got a hit having gotten hits in his last x games, something to that affect, I'm having trouble putting this idea into words. Would this prove a point?

    • TroyPatterson January 3, 2010 at 1:50 AM #

      I think what your looking for was in "The Book". It looked at player who had previously entered what they defined as a streak or a slump and found that a "streak" had no predictive ability of future outcomes.

    • Aaron January 3, 2010 at 1:11 PM #

      Len, this is precisely the question and Troy is right, there's a great chapter in "the book" by Tom Tango starting on page 54. The bottom line is that Tango found almost no predictive power in streaks for hitters. He did find that if a starter had four good starts in a row his next start was likely to be good as well (so that's interesting,) but bad streaks for starters were not predictive.

  10. Sam January 3, 2010 at 10:33 AM #

    Another factor to consider is a player's physical conditioning, which can hurt or help the success rate of his swing.

    But Troy accounts for this: third to last paragraph. There are plenty of reasons to explain a slump, but none to explain a hot streak, he says.

    If a player is in top shape, with minimal distractions off the field, and going about as well as he can go, baseball is largely reduced to a game of probabilities.

    • tom January 3, 2010 at 12:13 PM #

      as my dad always says "i would rather be lucky then talented any day" i dont know if i believe in this statement but it is interesting anyway

  11. Aaron January 3, 2010 at 1:12 PM #

    sorry, I was talking about pitchers only when I said "starters."

    • Len January 5, 2010 at 3:49 PM #

      Well that's pretty fascinating

  12. Sam January 3, 2010 at 6:36 PM #

    If I have my choice, I'd take lucky & talented, please.

  13. athos16 January 5, 2010 at 7:07 PM #

    Simple statistics, even advanced statistics are performed in a vacuum. The occur after the fact. They are cold mathematical models that will only be good for building video and board games. What they don't take into account is a player's overall talent level, his physical and emotional fitness, and the talent and state of the opposition. Anyone who played ball at any level knows that there were days when you just dragged, had nagging little injuries that did effect the way you played. On certain days your visual acuity and concentration were so good and such powerful match you would just see the ball better, and react to a pitch better. When your physical skills match…. watch out. Ted Williams in his book, pointed out that there were days when he could actually make out the individual stitches on the ball. Joe Morgan used to talk about the ball looking like a grapefruit. I was never at that level but I can tell you there were days that the ball was just a blur flashing by me, on other days I could see the rotation and movement and the ball seemed to float more. Needless to say I hit significantly better on those days. They weren't always safe hits, but the were solid. Defense determined the rest. The same could be said for leg speed, throwing speed, glove work, etc. I am not even going to touch on confidence, and attention. So many factors come into play, statistics are the least important factors.

  14. Aaron January 5, 2010 at 8:19 PM #

    I'm not really sure what you mean when you say "statistics are the least important factors," athos. The least important for a player on the field in the heat of the action? I think most people would agree with that. If a 3B is thinking about his range factor as he's charging a bunt that's probably not a good thing. The least important thing for a player who is training to get better? Maybe, most players should probably focus on strengthening their physical skills when training although many players could probably gain from a deep statistical analysis of their strengths and weaknesses. The least important thing for managers and general managers? There are a lot of very well paid and very intelligent people actually in those positions who would disagree with you. Every GM uses numbers and statistics to evaluate the player pool and even if you're talking specifically about more modern statistical tools many of the most successful GM's in the game rely heavily on them. The least important factor in how many people enjoy the game of baseball? If that's what you think then you are factually wrong. There is a large and growing community of fans who find that statistical analysis deepens their appreciation and understanding of the best game in the world. The least important factor in your enjoyment of the game? I obviously can't say whether that is true or not but if it is then more power to you. In fact, anyone who tells you that you must understand the modern baseball statistics in order to enjoy the game is wrong, period. I would be willing to argue, however, that there are certain parts of the game that one probably can't understand without an understanding of some of the newer numbers. Just like there are certain aspects of the game that can't be well understood when viewed through the lens of statistics, like some of the more psychological aspects of the game or why managers still wear uniforms.