“Guts and Stomps”, NBA Style
Posted by Neil Paine on November 16, 2010
The Miami Heat have been under fire recently for dominating weak foes and losing close games against good teams. The conventional wisdom is that this reveals a major gap in Miami's armor -- they just can't close the deal against stronger opponents.
Is this really true, though? And does it even matter?
In 2005, our friend Aaron Schatz wrote a piece for Football Outsiders called "Guts and Stomps", wherein he tested the same phenomenon in NFL football. Guts are close wins against good teams; stomps are blowouts of bad teams. As it turned out, the team with more regular-season "stomps" (that's big wins vs. bad teams) tended to win the Super Bowl and the conference title game more often than the team with more "guts", the team with more "skates" (close wins over bad teams), and even the team with more dominations (big wins vs. good teams). In other words, being like the Heat -- winning big over bad teams and not being able to close the deal against good opponents -- was actually predictive of Super Bowl success!
Does this hold for the NBA? 3 of the Heat's 4 losses have been by 5 or fewer points against a good team, so I considered a "gut" to be a win by 5 or less against a >.500 team. Likewise, all of Miami's wins have been by 9 or more (and all but one has been against losing teams), so a "stomp" would be a win by 9 or more against a <.500 team. Finally, "skates" are wins by 5 or less against a <.500 team, and "dominations" are wins by 9 or more against >.500 teams. Which category is more predictive in the conference finals or NBA Finals?
Since the merger, here are the Finals & Finals + Conf. Finals series records of the teams superior in each category:
| Category | Finals | Finals + CF |
|---|---|---|
| More Stomps | 19-13 | 50-40 |
| More Dominations | 22-8 | 59-32 |
| More Guts | 13-14 | 47-43 |
| More Skates | 9-20 | 35-52 |
In the NBA, dominating good teams is clearly the best indicator of postseason success. Teams that had more regular-season dominations (big wins over good teams) won 64.8% of their "final four" series, including 73.3% of their Finals matchups. But the second-most predictive attribute of "final four" success was having more stomps -- that is, destroying the league's weaker teams. And having more stomps was actually a better indicator of success than having more guts (close wins against good teams), just like Schatz found in football.
In other words, it looks like this criticism of Miami has no basis in reality. And in fact, their inability to close the deal against good opponents actually appears to say less about their chances of a deep playoff run than their ability to manhandle poor teams. As Schatz writes in the intro of every Football Outsiders Almanac: "Championship teams are generally defined by their ability to dominate inferior opponents, not their ability to win close games."
November 16th, 2010 at 2:17 pm
I would have been interested to see how teams matched up when the "guts" or "dominations" came against them. For instance, the Heat have been "gutted out" three times, the Celtics have been gutted out once, and neither have been "dominated." Based on reading this article, I would expect there to be no correlation, but prior to reading it, I would have guessed it mattered.
November 16th, 2010 at 2:29 pm
I wouldn't say this criticism has NO BASIS IN REALITY after this analysis. All this reinforces to me is that there is no statistically discernible patterns associated with CLOSE WINS in the NBA. WINNING BIG is a better indicator than anything (winning or losing) close games. This leads many to the conclusion that it's luck, not execution, that determines close wins. The criticism of the Heat, however, is much like that of last year's Orlando Magic - being "front-runners". Beating up demoralized teams and scoring tons of garbage points but being shaky in tight games. It's a valid criticism after watching the Heat's poor execution down the stretch of the Jazz game.
The problem with saying "there is no such thing as clutch play because over time the numbers state that close games are won by luck" is that you can't categorize "clutch-ness differential". All things being equal, yes, I believe close games are, over time, Gaussian - the data support that. But that's because over time, you will have more CLUTCH teams playing CLUTCH TEAMS and UNCLUTCH TEAMS playing UNCLUTCH teams than you will a demonstrably CLUTCH team playing a DEMONSTRABLY UNCLUTCH team. If the 2010 Jets played the 2010 Lions every Sunday for the rest of eternity, I don't believe the data would show that winning close games is a function of "luck". The more clutch, more veteran, more savvy team would win more close games.
November 16th, 2010 at 2:51 pm
96 Bulls, only lost ten, but looking at it, you get the sense they were vulnerable to a good opponent, and were beating up on bad teams. Unlike the 86 Celtics, who were coasting and losing to bad teams.
November 16th, 2010 at 3:05 pm
I usually love what you guys do, but 50-40 vs. 47-43 is NOT a statistically significant difference, and it's not close. It's interesting to debunk the common myth that winning close games is more important, but don't get sucked into believing you've actually found evidence for the opposite.
November 16th, 2010 at 3:29 pm
Well, one thing is for sure: it seems that the so-called "stat geeks" were wrong about the Heat and they won't be winning 68-75 games this season.
They've still been pretty good, though, and will most likely make the playoffs this season. Once they get there, it's anyone's game...
Still like Boston in the East but I think teams like Orlando, Chicago, and Miami will also be in contention for a place in the NBA Finals.
P.S. There's only so much I can say about the Heat after only 10 games. But I think they're a 50+ win team and have a legit shot of winning the East.
November 16th, 2010 at 4:14 pm
Doesn't this analysis avoid half of the question?
The wins over bad teams is great and all, but what about the losses? Can you complete the analysis using like "Stomped upon" and "Gutted open" ("Skated on" and "Dominated" wouldn't apply to Finals teams I guess), to see if the losses to the good teams mean as much or more than the wins?
November 16th, 2010 at 4:17 pm
Oops, that's "Dominated" and "Gutted Open" in terms of Finals/Conference teams.
November 16th, 2010 at 4:38 pm
You may get that sense, but the fact is that the 72 win Bulls lost half their games to teams that were 42-40 or worse. Only 5 losses were to good opponents. They weren't vulnerable to those teams at all. They beat them with as much regularity as they did the mediocre to poor teams.
November 16th, 2010 at 4:52 pm
This would be great news for Miami if the Heat was competing for the Super Bowl.
November 16th, 2010 at 6:48 pm
Convenient that the criteria happen to be set based on a very small sample size for one team. There is no statistical basis for the classifications.
Also, this article is statistical garbage because it is based on the records each year of the one team that lead the category. That in itself could be the result of some noise in a given year.
This article is basically a hand wave and saying everything is fine in Miami. You can't just make up meaningless crap to justify your point. Let's see a better analysis please if you want to try to make a point. Better yet let's just see what happens before you try to justify only 10 games.
November 16th, 2010 at 7:32 pm
Tsunami, this does not make sense. If you're looking only on regular season, "clutch" teams should meet other "clutch" teams as frequently as "unclutch" teams meeting "clutch" teams (in fact, a little less frequently, as they can't meet themselves). If "clutch" teams get to clutch situations against other "clutch" teams, it only means "clutch" is highly correlated with quality, which means there is no clutch... a good team is good in clutch as well. There is no way around it, if the data shows "clutch" is a myth, then it is. It's similar with the "hot hand"...
November 16th, 2010 at 7:35 pm
Tsunami, this does not make sense. If you're looking only on regular season, "clutch" teams should meet other "clutch" teams as frequently as "unclutch" teams meeting "clutch" teams (in fact, a little less frequently, as they can't meet themselves). If "clutch" teams get to clutch situations against other "clutch" teams, it only means "clutch" is highly correlated with quality, which means there is no clutch... a good team is good in clutch as well. There is no way around it, if the data shows "clutch" is a myth, then it is. It's similar with the "hot hand", there are great psychological explanations, but the phenomenon simply doesn't appear...
November 16th, 2010 at 7:41 pm
Adam - the article is not all statistical gargabe, nor is it made up meaningless crap. It does not show convoncingly that "Stomps" are a very good predictor. But while it is pretty hard to disprove something, these data are pretty useful to show that the common sentiment ("teams need to be able to close out close games") is not true.
This has nothing to do eith the Heat per se, and I doubt Neil would have mentioned them if not for the criticism about them losing to good teams and "only" beating bad ones.
Agree with AHL though, analysis of the loss statistics would be neat. Although it would be even more difficult to interpret those data - the sample size will be a lot smaller, since we're talking about teams in the conference finals, and those generally don't lose much...
November 16th, 2010 at 9:08 pm
If this were an article about disproving the myth of "teams needing to be able to close out close games" then wouldn't the data be based on margin of victory alone not record? Something such as records of NBA champions in close games during the season or playoffs would be a good start.
November 16th, 2010 at 10:03 pm
Why would it matter by how much they win those series? The margin of victory is used to predict future results, but playoff wins aren't predictors, they are what is being predicted. No need for additional info.
And again, many people say close games vs good teams matter more than runaway victories vs bad teams. This is not the case. The data have holes, like not accounting for losses, but margin of victory is certainly not what's missing.
November 16th, 2010 at 11:07 pm
I should have been more clear. If you wanted to prove that a team didn't need to win close games to win an NBA title then you need to show multiple NBA champions who have had poor winning percentages in games that were, let's say plus or minus six points.
I don't know how you could say this data proves anything about having to win close games. It doesn't help because it doesn't show the winning percentage and it doesn't show how each champion actually performed in close games.
November 17th, 2010 at 12:19 am
it seems to me beating up on bad teams means you are maintaining focus ALL THE TIME rather than turning it on when you need it. But maintaining focus is a skill you achieve through practice, and generally teams who think they can turn it on when they need it find out that they don't have what it takes to maintain the pressure needed to get through a seven game series
November 17th, 2010 at 12:30 am
Maintaining focus by blowing out bad teams doesn't really prove anything at all to me. Take last year for example. Both the Lakers and Celtics were relatively average compared to Orlando and Cleveland, two teams that routinely blew out weaker teams. The Lakers and Celtics would frequently get in close games or lose to some of those teams.
It the long run, it really didn't matter. The Lakers and Celtics more than proved themselves in the playoffs while the Cavs and Magic did the exact opposite.
November 17th, 2010 at 2:23 am
I think that beating up on bad teams shows the superior talent that winning teams generally possess. In the NBA there are generally bad teams, good teams, and real contenders. The good teams generally skate by the bad ones, and the real contenders generally stomp the bad teams. The dominations by this theory are real contenders beating up on good teams. The NBA champ is almost always one of those real contenders, rather than a good team that makes a killer playoff run. When real contenders play good teams, I'd bet they get a lot more wins (by 6-8 points) + dominations than guts. Getting to the meat, there just aren't that many times two real contenders play one another, and those are the only real games that are independent of the difference in talent when looking from the perspective of the league champ. Without making the distinction between good teams and real contenders, the data isn't really dispositive of anything. I think that if you extrapolated this study to all teams, you would find stomps correlates with being a real contender more than just champions, and that the guts number is miscontextualized because it includes real contenders that let good teams stay in the game, but ultimately pull out Ws. Guts games between real contenders probably correlates with champions. Isn't that what most of the finals/conf finals games are anyway?
November 17th, 2010 at 11:08 am
Shouldn't you run a regression to see how the factors work independently from each other? Teams that dominate will tend to stomp, but will also tend to gut less. Also, skating is considered bad, but is also negatively correlated to stomps. We need a way to separate these correlated factors apart and see which factors are more important, rather than a simplistic win loss analysis.
November 19th, 2010 at 12:05 am
I think what you are seeing here is the well-known result (see e.g. Wages of Wins) that efficiency differential is the best predictor of team strength. Having many blowouts indicates a large efficiency differential, while playing close games indicates a smaller efficiency differential.
One possible explanation for the effect you're seeing is that winning close games is probably largely determined by luck (when the difference is only 2 or 3 shots then every miss matters), so a team winning unusually many close games is likely to be a statistical fluctuation.
November 19th, 2010 at 11:25 am
I am with some of the other commenters. This is fascinating, and I am certainly willing to believe that conventional wisdom is wrong, but I'd need to see a larger data set to be completely convinced. Something along the lines of the Win and Loss profile in terms of these categories of all playoff teams. If there were no strong correlation in that set between guts wins and playoff wins, then I would be more convinced.
Also, it seems likely that if this were being done without the context of the Heat, the cutoffs might be different...in particular, if one cutoff is set at 5 pts, then the other would more likely be 10 pts than 9, which would alter the Heat profile. Which makes me wonder if there is maybe some better way to deal with this statistically than having what are necessarily arbitrary cutoffs.
A lot of analysis, such as Hollingers Daily Ranking, rely heavily on Point Differential, but I wonder if some sort of weighted victory margin might tell us something - maybe even something as simple as Victory Margin x Opp Winning Pct...
November 19th, 2010 at 11:49 am
Here are my conclusions from watching years of basketball:
(1) Ability to win close games is somewhat overrated. It's ALWAYS better to blow teams out than to win the tight games.
(2) The very best teams can blow out good teams. The Lakers were able to do that several times in the playoffs last seasons and it paved the way for a second straight title.
(3) "Clutch" factors are somewhat overrated. The best predictor of winning the close games is to be the team with the (small) lead. I would take the LA Clippers over the LA Lakers every day of the week and twice on Sunday if you gave them a 5-point lead with 5 minutes to go.
With that said, there's really no need to justify Miami Heat's performance up to this point. They've been a pretty good team so far, but they have many visible flaws and will most likely fall well short of their ridiculous preseason expectations. They look like the third best team in the East and will need to improve their play considerably if they are to win it all.
November 19th, 2010 at 1:00 pm
Actually they look like the best team so far according to this site. Everything Neil has posted so far supports that notion, and they're unlucky if anything to not be 9-2. In a 7 game series they would be favored by this website over other teams (with homecourt advantage).
They need to make minor tweaks, because they've lost by the slimmest of margins to good teams, when Wade choked at an all-time level. This can be easily corrected, Wade can easily improve, whereas Boston will probably get worse as the season gets along if anything.
And even if they don't win 70 games, they might do it some other season they're that good and young.
Who are these three superior teams in the East? New Orleans and Utah play in the West dude, get it straight.
November 23rd, 2010 at 12:47 pm
just a random opinion:
teams that regularly blow out their opponents are more likely to rest their key players for a significant amount of time, thus minimizing the risk of injury or over-exhaustion in the long run.
also, teams that pull out a lot of wins in close games tend to have the sense of entitlement, i.e. that they can turn it on whenever they want to. this could really hurt them in the long run.
December 3rd, 2010 at 3:55 am
This and Aaron Schatz's breakdown is a nice, almost "user-friendly" way of articulating a point that could be much more succinctly and accurately stated by analyzing how badly a team beat or got beaten by their opponent based on that opponent's average margin of victory or loss (and then plotting the overall scores of each team against, say, postseason success). This is largely the gist of my rating system (and I'm sure plenty of others) and it communicates the point because teams that lose to good opponents by a small margin get penalized less than they would losing to average teams by a small margin, while teams that beat bad opponents by a large margin get awarded perhaps slightly better -- depending on how badly they beat those opponents -- than teams that beat average or opponents by a small a small margin.
In fact, it's all relative to the exact points margin, so you don't have to declare any arbitrary thresholds from the beginning (i.e. you don't have to define new categories like "stomp" and "crunch" and whatever).
January 10th, 2011 at 12:02 am
My take on why LeBron James teams tend to under-perform their point differential in the playoffs and why the Dallas Mavericks have an astounding 103-49 record since 2004/2005 in regular season games decided by 5 points or less.
http://getbuckets.fantake.com/2011/01/09/dirk-nowitzki-and-lebron-james-black-swans/
Basically even the best statistical models will only work 99% of the time; there's no way for them to account for "black swans" -- things we have never seen before. And there have never been two basketball players quite like LeBron and Dirk before.
February 13th, 2011 at 7:11 pm
Thought it was time to think more on the Heat this season with regards to this analysis, so I've compared how the elite (top 8) teams have done against other elites this year both in decisive games, and games overall.
Quick takeaways:
-Whether you look at not-so-close games or all games, there's not a huge difference. I'm fine with the conclusions from your article, but I doubt you'll find many teams who look bad only because of the luck of close games.
-Heat still not looking that great.
-Spurs, Mavs, and Celtics on the other hand, look quite good.
http://asubstituteforwar.wordpress.com/2011/02/13/guts-domination-revisited-nba-10-11-elite-vs-elite-records
March 8th, 2011 at 1:01 pm
I have a question about this statistical study. Doesn't the population of "stomps" include many (if not all) of the teams with high numbers of "dominations?" Since dominations are the most accurate predictor of future success in the NBA, doesn't the inclusion of teams with a high number of dominations skew the data for teams in the stomp category who have very few dominations? In other words, the Heat, with very few dominations but plenty of stomps, are benefited by the dominations of other teams included in the stomp data.
Shouldn't there be a separate category used to predict the success of the Heat, named something like "Stomps without dominations." And shouldn't that data set be the one weighed against close wins over good teams?