Playoffs Pessimism Part II: Are Stretch Runs Sticky?

October 19th, 2015 | by Matt Pacenza
Photo by Melissa Majchrzak/NBAE via Getty Images

Photo by Melissa Majchrzak/NBAE via Getty Images

In my previous post, I evaluated a claim that optimistic Jazz fans make: That a team that improved significantly during last season will get even better this year. I found, however, that such back-to-back jumps in wins are rare — alarmingly so.

However, as many fans point out, the Jazz don’t really need to improve. All they need to do is to play as well as they did after last year’s All-Star Break. Those Jazzmen, of course, went 19-10 down the stretch, significantly better than their prior record of 19-34.

What most Jazz fans think — or at least hope — is that the team’s late season run is more predictive of how they will be do this coming season.

That, my friends, is a proposition we can test!

Let’s turn to the data and ask a simple question: Are the 2015-16 Utah Jazz likely to win as much as they did after the All-Star Break (their winning percentage of .655 translates to a 54-28 record)? Or will their mark be closer to their full season total of 38 wins?

In other words, do stretch runs stick?

The Data

Just as with my previous post, I looked at the last 11 NBA seasons, back to 2004-05. I looked for teams that either improved or declined significantly after the All Star Break.

The research project was simple (but time consuming!): Find all the teams with big late season jumps or declines, then see how those teams did the following year.

A reasonable cut off for a “big” post All Star Break change was a change in winning percentage of .200. The Jazz easily fit in: They improved from .358 team to a .655 team, a difference of .297.

How many teams improved by at least .200? In the past 11 years, 17 teams have. Of those (and we’ll list them at the bottom of the post) the Jazz actually had the fourth-biggest improvement behind only the 2004-05 Nuggets, the 2004-05 Warriors and the 2006-07 Sixers.

Here’s another way to look at how much the team improved: The Jazz jump during the 2014-15 season was the equivalent of going from the West’s 13th-place team to its seventh-best. Before the All Star Break, they played like the Kings. After, they were nearly as good as the Spurs.

That’s improvement. Setting more cialis patent deutschland granular with this test as the procedure has developed from its canadian viagra inception, the brady. Weaken and therefore it is important buying generic cialis online to keep control of your bladder. Right time to get back to free online viagra samples a life filled with gratitude.

Now, if we’re going to study how much changes “stick,” it’s worth our time to also look at those teams who get much worse after the All Star Break. So how many teams have gotten at least .200 worse after the break? 19. Only one, the 2006-07 Pacers, saw their record plummet by as much as the Jazz skyrocketed last year.

And More Data!

OK, so how do those big-change teams fare the following season?

Let’s start with the improving clubs. We now only have 14 teams to work with over a 10 year period. (The other three — the 2014-15 Jazz, Celtics and Pacers — all had their jumps last year. Their fates are undecided!)

Those 14 teams, post All-Star Break, had an average record of 22-8, for a winning percentage of .735., which was .261 higher than their pre-break marks of .475. And the next season?

Those teams averaged 40 wins and 42 losses, for a winning percentage of .494.

Hmmmm. In the year after their big post All Star jumps, those teams won at a rate .242 lower than their late season hot streaks. In other words, those teams in that following year gave back, on average, 93 percent of their gains.

Whoa. Let’s look that data visually:


How “Big Jump” Teams Fare the Next Season

Now, let’s look at the declining teams. Here we have 18 to study (the 2014-15 season saw just one team whose record plummeted, the Raptors.) Those 18 teams, post All Star Break, had an average record of 10-20, for a winning percentage of .345, which was .236 worse than their pre- break marks of .581. And the next season?

Those teams averaged 42 wins and 40 losses, for a winning percentage of .513.

So, during this decade of study, late-season collapses were a bit “stickier” — but still not that much: Their winning percentages in the year after the big post All-Star decline were .168 better than during their late season collapses. The decliners thus gained back about two thirds of their success the following season.

The Discussion

I went into this study feeling fairly confident that late season runs (and collapses) would turn out to be not as predictive as most fans might think. I’ll admit I’m genuinely shocked how little those final 30 games tell us about how a team will do the next year — especially when a team has a hot stretch run.

It turns out the best predictor for how a team will do next season? By far, it’s how they did for the entire season, not how they did down the stretch — and it’s not even close. Let me share a bit of fancier math and run some simple statistical analyses.

“Correlation” is a statistical concept which tells us how much one set of data is linked to another. In this case, it’s a simple tool for observely how closely a following year’s winning percentage is tied to previous records.

I took the data for my 32 “big-change” teams and ran correlations to see how well their winning percentages in the subsequent sedason correlated with three variables: The team’s record post-All-Star Break, pre-All-Star Break and for the whole season. The result ranges from 0 to 1. 1 would mean the two sets of data march in lockstep with one another; 0 means there is no connection.

A correlation of .3 is considered a “weak” relationship, .5 is considered “moderate” and anything above .7 “strong.”

Pre ASB 0.634
Post ASB 0.294
Full Season 0.629
Next Year 1.000

Pretty clear, no? Most surprising to me is not only do full season records correlate better with the following season’s record than the post-All-Star marks, but the strongest correlation is actually the pre-All-Star record.

Not only are these teams worse than they were during their hot streaks, they’re actually a bit closer to what they were at their lowest.

Why? Clearly, there is no one answer. But let’s speculate about a few likely causes. During their “hot streaks,” teams enjoy excellent health, which often doesn’t persist the next year. Other teams notice that hot streak, begin to take that team more seriously and scout and plan for them better. Some of the “hot streak” is just dumb luck: Shots fall that typically wouldn’t. The teams win a disproportionate number of close games. A few calls go their way. They run into tired teams, or teams resting a key player, or teams battling injuries. The next year, those trends tend to return to the average.

Did the 19-10 Jazz benefit from a few of those trends? Probably.

Now, let’s state the obvious: Data is not destiny. And, as they say in fine print on those ads for financial products, Past Performance is Not Necessarily Indicative of Future Results. Just because late season hot streaks haven’t “stuck” for nearly all teams, that doesn’t mean it won’t for the 2015-16 Jazz.

However, just as with the last post, optimistic Jazz fans anticipating smooth sailing to a playoff spot need to realize that history — and math — aren’t on your side.

One final caveat: Some fans have suggested in comments that the Jazzmen are more immune to these trends because of the nature of the team’s big jump, largely due to a massive improvement in team defense — and the insertion of one long-armed French center into the starting lineup. Defense, the argument goes, is less fluky. Once you make a big jump on D, you stay there, assuming coaching and personnel broadly stay the same.

In other words, maybe the Jazz are better poised to buck these trends, as a D-first club. Folks, I sense another proposition to test!

Stay tuned.

P.S. Here’s that full table for all you gluttons.

First, the late season hot teams:

Pre ASG Post ASG Next Year
2004-05 Denver 24 29 45.3 25 4 86.2 44 38 53.7
2004-05 Golden State 15 38 28.3 19 10 65.5 34 48 41.5
2004-05 New Jersey 23 30 43.4 19 10 65.5 49 33 59.8
2005-06 Sacramento 24 29 45.3 20 9 69.0 33 49 40.2
2006-07 Philadelphia 17 36 32.1 18 11 62.1 40 42 48.8
2009-10 Milwaukee 24 27 47.1 22 9 71.0 35 47 42.7
2009-10 Miami 26 27 49.1 21 8 72.4 58 24 70.7
2009-10 Phoenix 31 22 58.5 23 6 79.3 40 42 48.8
2010-11 Houston 26 31 45.6 17 8 68.0 42 40 51.5
2011-12 Boston 15 17 46.9 24 10 70.6 41 40 50.6
2012-13 L.A. Lakers 25 29 46.3 20 8 71.4 27 55 32.9
2012-13 Denver 33 21 61.1 24 4 85.7 36 46 43.9
2012-13 Miami 36 14 72.0 30 2 93.8 54 28 65.9
2013-14 Charlotte 23 30 43.4 20 9 69.0 33 49 40.2
And now the late season slumpers:
Pre ASG Post ASG Next Year
2004-05 Cleveland 30 21 58.8 12 19 38.7 50 32 61.0
2004-05 Portland 21 30 41.2 6 25 19.4 21 61 25.6
2004-05 L.A. Lakers 26 24 52.0 8 24 25.0 45 37 54.9
2004-05 Orlando 28 24 53.8 8 22 26.7 36 46 43.9
2005-06 Portland 18 33 35.3 3 28 9.7 32 50 39.0
2005-06 New Orleans 29 23 55.8 9 21 30.0 39 43 47.6
2006-07 Washington 29 21 58.0 12 20 37.5 43 39 52.4
2006-07 Minnesota 25 27 48.1 7 23 23.3 22 60 26.8
2006-07 Indiana 28 24 53.8 7 23 23.3 36 46 43.9
2009-10 New Orleans 28 25 52.8 9 20 31.0 46 36 56.1
2010-11 Utah 31 26 54.4 8 17 32.0 47 35 54.5
2010-11 San Antonio 46 10 82.1 15 11 57.7 62 20 75.8
2010-11 Atlanta 34 21 61.8 10 17 37.0 50 32 60.6
2011-12 Miami 27 7 79.4 19 13 59.4 66 16 80.5
2011-12 Portland 18 16 52.9 10 22 31.3 33 49 40.2
2011-12 Minnesota 17 17 50.0 9 23 28.1 31 51 37.8
2012-13 San Antonio 42 12 77.8 16 12 57.1 62 20 75.6
2013-14 Indiana 40 12 76.9 16 14 53.3 38 44 46.3
Matt Pacenza

Matt Pacenza

When he isn't writing about the Jazz, Matt Pacenza is an environmental activist, Arsenal fan and world-class blowhard about many matters. A native of upstate New York, with a background in journalism and nonprofits, Matt lives near Liberty Park with his wife and two sons.
Matt Pacenza


  1. IDJazzman says:

    Interesting food for thought, Matt. I am an Engineer by education, so I have dealt with a lot of numbers. Not an easy thing to do when there are so many variables. I know when the problem can be simplified, then it was a lot easier to determine an accurate outcome. As you pointed out, the one thing that will cause the Jazz to follow the trend of past teams that have made a jump, is injury. Have the Jazz had their injury with Dante or will they experience an abnormal amount of injuries for the upcoming season and they actually end up with a worse record than last year’s? It is possible that the Jazz are nearing the peak of their capability with the existing talent they now have and the improvement left, is just not there. I have personally never believed that the record after the All Star break from last year was any indication of the upcoming success for this year. There were a number of teams that the Jazz played that I think wanted to loose, because of tanking and other reasons. No one can argue that they are not a much better team from 2 years ago. For me, to make the problem simple, I go through and look at the improvement for each player that is possible for this coming year. I don’t see a lot from anyone player, but I do see a little from several players, mostly Hood and Burks. I think the losing of Dante and the gaining of Burks could be a wash. Due to the amount of improvement among so many players added together and the coaching taking hold, I think there will be a small improvement in total wins, assuming no major injuries, low to mid 40’s possible. IMO, I think that how they improve into the 50 win category will depend on if the team was lucky enough to have drafted a superstar in Dante Exum and Lyles becomes a fringe All Star or better, or they trade for an All Star. Jazz are still 2-3 years away from being a top 10 team in the NBA, if all goes right. Otherwise, maybe they end up where they started when the rebuild began. An average NBA team. I have faith they are doing it right and they are headed for bigger and better things than when Paul Millsap and Al Jefferson was here. Time will reveal the answer. Good write and thanks for all the work.

    • Matt Pacenza says:

      Thanks! I agree low to mid 40s is the most likely bet. In Vegas, they’re at 41ish, which sounds about right to me. But I do think a lot of fans are looking at that finish and thinking playoffs are a given, unless injuries hit. And I don’t quite see that. I see a young team getting better whose road to contention is likely to be bumpy — because such roads nearly always are. That’s the point of the “research.”

  2. LKA says:

    Analize, graph, and whatever all you want. How about let’s play the season.

  3. Andrew says:

    Great article. I was with you, I expected the data to not support stickiness, but I didn’t expect it to be so decisive a result. What’s the p-value for these results?

  4. Aged fan says:

    This is interesting, and potentially useful information. But I’d also say that say that 14 cases is hardly enough to establish a statistical trend. It might point to one that could hold up over time, but that trend may not solidify either.

    But in some ways, 14 cases is ideal in the sense that it provides great opportunity for comparison through case studies. Since there’s only 14 teams, someone could (if they were willing to invest a bit of time), try to figure out what was going on in each of these 14 cases, and see if the Jazz’s case looks anything like those other cases?

    For example, were there injuries, trades, new coaches, very young teams, a playoff push, an unsustainable period or two of individual performances? Where did the improvement come in the other cases — was it more due to eking out close victories that they hadn’t been getting earlier during their hot streak or to greatly improved point differentials? Does it make a difference whether the improvement was on offense or defense? What is the role of team continuity from the first to second year? Etc., etc.

    • Matt Pacenza says:

      Totally agree! I was tempted to do some version of that, but ran out of steam and time. Be interesting, for example, to look at those few teams for whom the stretch run WAS sticky. 2004-05 New Jersey, for example.

  5. ccottle1990 says:

    How about Alec Burks in all of this? We can’t say that we weren’t dealing with injuries during that stretch because we were. Hayward wasn’t 100% for a part of it. We made a move to start Exum at PG which was like playing 4 on 5 on offense. Burks add’s something to the equation, the loss of Exum changes things. Hood finally being healthy factors into things as he wasn’t the first half of last year. Gobert wasn’t starting the first half of last year. What would be an interesting view is into why those teams made a jump in the second half or fell in the second half. See how the roster is compared between the two years. Did a player play lights out in the second half, only to sign somewhere else? Someone injured? etc… I believe the Jazz finish in the mid to high 40’s this year.

    Statistics are an interesting subject but you always have anomalies. In fact, they almost have to happen and we are predicting things based off of human’s which is also why things like economics fall under the social sciences. We can make projections with a certain amount certainty but there is always the human element. It’s the eye test if you will. Did those other teams get their rise from a gimmic? A scheme nobody was ready for? The aforementioned flash in the pan player who use the spot light to move to greener pastures? A player catching fire for an extended period?

    So another interesting thing to look at would be to look at each player on the Jazz and see if what they achieved is sustainable or repeatable or improvable. Without having done it myself, I feel pretty good saying the only player who was truly playing out of their mind was Gobert so we can expect more of a return to the mean this year with his play but that mean is without Kanter or Exum and with a healthy Burks and Hood. We will not be as dominant defensively but we will also be more balanced offensively and that is why I think we end up winning more games and not follow the trend here.

    Also, 14 teams is an extremely small sample size. The confidence level in such a small sample size is not high. You could flip a coin and get heads 14 times and tails on the 15th try and it not be a statistical anomaly. You still have a 50/50 chance of getting heads or tails. This all could mean jack squat because that is the nature of statistics.

  6. Gilbert says:

    Matt, I think that one of the key components you are leaving out of this analysis is the age factor. Ultimately, the question at the heart of all this is simply: “Was this stretch more of a hot streak or is it indicative of sustainable growth?”

    In trying to answer that question, age is a very important factor. At very young ages, players mature and grow quickly. Though you definitely see the two-steps-forward-one-step-back type of ups and downs, most of the time when a young player makes a leap in production it is a clear indication that that player has had a break through and has become a much better player than he was before the leap occurred.

    This phenomenon would, logically, apply at the team level as well.

    Older players also have ups and downs, but once a player is beyond the age of 25 or 26, usually that is just a matter of the types of variance you mention in the article. Health, chemistry, morale, momentum, and sometimes dumb luck all conspire to create those types of fluctuations. When a veteran is experiencing a period of markedly increased production, you can rest assured that at some point in the future, sooner than later in all probability, their level of play will regress to their typical level.

    This too can be applied at the team level.

    When you look at it through this lens, the data takes on a much different complexion. Using BBreference, I’ve run through each of the 14 teams in the study and determined the following:

    — Average age of the 14 teams was 27.5 years of age, nearly a full year above league average
    — Teams ranged from 25.5 yrs (13-14 Bobcats) to 30.7 yrs (12-13 Lakers)
    — No other team on the list had an average age of <26

    The Jazz, of course, sported an average age of 23.4, more than 2 years younger than the youngest team on the list and a full 4 years younger than the average of the team in the study.

    This is an important point, because of the growth patterns exhibited by young players as opposed to the concept of natural (i.e., random) variance.

    It is quite possible that the Jazz will give back much of the gains that they showed at the end of last year. Many things have changed and the team must start from scratch. However, I think the analysis you've done here is based upon a faulty premise because it fails to make an apples to apples comparison and ignores the essential hope upon which most Jazz fans (consciously or not) pin their hopes for the upcoming season: youth.

    It's also a statement in and of itself that there isn't a single team in the last decade plus who had the type of stretch run the Jazz enjoyed at a similar point in their growth curve.

    Anyway, food for thought.

    P.S. — thanks for all the time and hard work! This was very informative.