Writers of Pro Football Prospectus 2008

21 Nov 2013

On Momentum

Brian Burke takes an analytical approach to the question of momentum, comparing the usual performance level of teams to their performance in the drive after a supposedly momentum-swinging play.

Posted by: Rivers McCown on 21 Nov 2013

16 comments, Last at 23 Nov 2013, 3:25pm by Theo


by Revenge of the NURBS (not verified) :: Thu, 11/21/2013 - 1:20pm

As the article suggests, momentum swings are a tricky thing to quantify. Much like a Nostradamus "prophecy", it can only be seen after the fact, and even then it's mostly a product of shoehorning things to fit a narrative.

Take, for example, the Colts-Titans game in the graph. I watched that game. My subjective observation was that the attitude (again, an unquantifiable thing) of the Colts changed during that personal-foul-fest drive in the 2nd quarter. They proceeded to score 17 straight points afterward, and I was sure that drive was when the momentum had swung. The next morning, I hear that Andrew Luck gave some big halftime pep talk, and that that's when the momentum swung. If we're talking about a real effect, it should be more clear cut. (FWIW, the upward slope in the graph begins when I said, so I'm right and you're wrong ESPN!)

by Bobman :: Fri, 11/22/2013 - 1:18am

I was thinking of the same game!

I had assumed the big mo event was the fumble recovery on the KO return (two plays later, game leading TD). Perhaps the penalty-fest or halftime talk was the key, because they started the 3rd Q with a nice solid TD drive. (Actually, on the penalty-fest 60 yard drive the Colts gave away 41 of those yards and still held TENN to a FG, which may have been the REAL key. "Hell, we can give them 40 yards and still keep them from scoring a TD? Damn, we're better than they are!"

In my mind, the most likely situation for actual momentum is when a team crumbles--psychological mistakes or injuries and substitutions lead to other mistakes and it snowballs (snowball: the very definition of momentum). I have no doubt that can and does happen. Look at the Colts vs Rams--they dug a hole so deep the refs were calling out penalties in Chinese. Not sure there was one event (though the FFTD or the 98 yard PRTD could each qualify)--I think it was more a case of Rams being opportunistic, a little lucky, and the Colts playing poorly and being unlucky. If a tsunami can be said to have momentum, the Rams had it in that game and it may have faded over time, but never flipped.

Mostly, I assume what is perceived as momentum is a good team or player being what he/it is. The Colts are better than Titans and even though they stumbled early, they asserted their identity late vs the Titans. Vs the Texans two weeks earlier, the coach collapsing mentally messed up the Texans--it was not some mysterious force lifting the Colts--they took advantage of an opponent that was distracted and had been playing over its head.

Three recent games where a lot of people could cite momentum, but I think it's just a mystical way of saying Team X is playing better or worse.

by BaronFoobarstein :: Thu, 11/21/2013 - 2:55pm

He seems like he's arguing against himself. He's trying to conclude that momentum doesn't exist, yet his data show that it does but is unpredictable. I especially like the comparison he makes to coin flips and how they are streakier than people tend to think they are. Obviously that means that the streaks do exist; you just don't know when you're in one how long it will go. The same is true of outcomes of football plays. Look at the play by play. Do you come across sequences where one team performed better than a strict comparison on their relevant abilities would predict? Of course you do. You just may not be able to predict how much longer such a sequence will continue nor what its outcome will be.

by Bobman :: Fri, 11/22/2013 - 1:20am

But that doesn't make the streaks "momentum" does it? Doesn't that show they're just part of natural variation noise?

Momentum implies a significant, irresistible force, like a tsunami or tidal wave. Those mini-streaks in the coin-flipping were anything but irresistible and significant--they can and do break any time randomly, or continue, or reverse.... Just noise, not some larger force.

by RickD :: Thu, 11/21/2013 - 3:30pm

Burke is falling into the trap that many sports statisticians fall into: namely, thinking that randomness is an explanation.

Failure to disprove a null hypothesis (the typical null hypothesis is 'this happened by chance') does not logically imply that the null hypothesis is true.

What he's done is postulated a trigger for "momentum switch" and then shown his postulate cannot be supported. As statistical tests go, this is a "so what?"

I could postulate that people driving in Manhattan follow random paths. I could then devise a test of this hypothesis: "when a driver sees a bus, he's more likely to turn left". And then, assuming that I have access to enough traffic data, I could even show that my alternate hypothesis isn't supported by the data. But that doesn't prove that my initial assumption is valid.

I really wish sports statisticians would stop writing "I've failed to disprove the null hypothesis" articles.

by Scott Crowder (not verified) :: Thu, 11/21/2013 - 4:11pm

The other trap he's falling into is thinking that the physics definition of momentum is the same as the psychological definition of momentum. Momentum, as applied to football, is a psychological phenomena. His comparison to a boulder rolling down a hill displays his misunderstanding of the use of the word and so he tries to disprove it using physical definitions.

You'd have to actually watch the game, SEE when and what seemed to spark the change in momentum - as the above poster said, it seemed the penalties sparked the attitude change in the team - and then track the stats to see if the momentum change you detect was actually real.

Simply "They had a turnover and I've picked turnovers as being a momentum changer and nothing was proven" is not a good statistical analysis in these situations. You have to deduce the psychological state of each team. Does it "feel" like one team has responded and now has momentum? If it does, then from that point on do the stats bear out this hypothesis?

He's dismissing this entire approach as being 20-20 hindsight by discarding the fact that the momentum shift was felt AT THE TIME. He'd have to watch the game and personally feel the momentum shift at the time, as the researcher, to figure out if it were true. Or have others tell him.

Only way to really study if it's real is to enlist a team of football enthusiasts and have them independently watch the game and signal when they think there's been a momentum shift and what triggered it. See if the data bears out the shift and if enough of the observers identified the shift during the game.

Sadly, what he's doing is a waste.

by Bobman :: Fri, 11/22/2013 - 1:46am

I agree with Burke AND I agree with you. Damn, call me Janus.

In addition to the psych momentum you suggest (we know this is an emotional game played by emotional men, going all-out in ten second bursts, stopping, and then starting again), there is a kind of mo I cannot describe in a word. But mental mistakes snowball, and if guys get frustrated, or try to overcompensate, that can backfire as well. Say you're a CB and you've been burned on 3 consecutive short curls. On the 4th you charge in to break it up but it's a double move and you're burned for a 40 yd TD. Do those plays equal momentum? Are the other guys just better than you? I think the existence of that pattern is real and frequent and takes great play and discipline to break.

Another kind of mo is injury-related. Your pro-bowl center is hurt, so your OL will surely play less well after that. But how much less well? Say the RG shifts over and is 80% as good as the C. And the replacement RG is 80% as good as the starter. That leaves the middle of your OL playing at 64% of it's initial strength 100% x 100% x 100% when healthy, 100% x 80% x 80% with replacements. (should that be additive or multiplicative? If they never interacted, I'd say add, but since they work together a lot, double teaming guys, pulling, trapping, etc, I'll stick with multiplying.) After they start getting beat a few times, maybe they start holding and get flagged--more negative plays. Frustration builds and they make mental mistakes. Other OL guys, trying to bail them out, start neglecting the blitzers they should have been picking up, etc. The QB suffers, the passing game, the RBs. All because one guy hyperextended his elbow. That looks like a clear case of a snowball effect (which defines momentum) with the mistakes getting bigger and more frequent as the result of one initial event. But as an observer of the game, would you really call that momentum within a game, or merely the line playing poorly as the result of an injury?

Oh, and I love your suggestion of having a room full of fans, say 20, watch the same game and independently record mo shifts, deaths, and increases. See if they match and compare to data. Repeat a hundred times or more to see if it makes sense. I bet it will have a lot of consistency. Still might say more about the viewers than the players. Not sure.

by DEW (not verified) :: Thu, 11/21/2013 - 4:43pm

The problem is that the people postulating that momentum exists (typically meaning "sports media") themselves don't define a momentum trigger, so in attempting to disprove the existence of momentum in sports, those attempting to apply statistics are forced to define trigger events themselves.

The real flaw here is that Simms, Dierdorf, and their Amazing Friends are postulating the *existence* of a real, tangible effect: that at times in sporting events, a single play or event will in turn create a psychological effect on the players that causes one team to play better, one team to play worse, or both, so that before and after this event the results on the field are different than they had been earlier in the game. It should therefore be up to those claiming momentum *exists* to define the concept and prove its existence with empirical evidence.

But of course, they don't, largely because they're not *IN* the statistics business, but in the narrative-of-sport business. And therefore, statisticians are left to try and put the theory to the test as best they can, because the objective is to get the fans, the players, the coaches, and the media to talk about sports as they actually are, not based upon narratives that are applied subjectively and in hindsight due to humanity's bias towards pattern recognition. So we get articles like this one, which can never really satisfy unless the statistics surprise us and show that momentum does exist.

by Led :: Thu, 11/21/2013 - 5:58pm

"And therefore, statisticians are left to try and put the theory to the test as best they can, because the objective is to get the fans, the players, the coaches, and the media to talk about sports as they actually are, not based upon narratives that are applied subjectively and in hindsight due to humanity's bias towards pattern recognition."

You realize that is question begging, right? I have no particular dog in this fight. I can see both sides. On the one hand, I suspect commentators often say dumb stuff and make up false stories to rationalize results. On the other hand, I have had the subjective experience of a momentum change as a player and a spectator, which may have been an illusion but also may not have. So I can't rule out the possibility that it has a real effect in certain circumstances, albeit difficult to prove. But if you begin with the premise that every narrative is false unless and until it can be proven statistically, you're just assuming the null hypothesis without proving it. You may think you're avoiding this by shifting the burden of proof but that's just a debating trick and not a logical or scientific argument. "Momentum doesn't exist" and "momentum does exist" are both statements of fact, and the assertion of either as true requires evidence. In the absence of conclusive evidence, one ought not give a dogmatic answer either way. "I'm dubious that momentum exists, but I can't rule it out" or "momentum has not been demonstrated statistically" is about the most one can say.

by Mussell Shoals (not verified) :: Fri, 11/22/2013 - 7:38pm

The person making the initial assertion always bears the burden of proof. The initial assertion is "momentum does exist." "Momentum doesn't exist" is a refutation. It's conditional. Without the initial assertion, the refutation is unnecessary, and we return to the default.

If I were to say gerbils live in the center of the sun, a counterargument of gerbils do not live in the center of the sun is unnecessary unless I can prove the former.

The only difference between my gerbils example and the concept of momentum in sports is convention. By force of repetition, the idea of "momentum" has gained a footing in the minds of athletes and spectators. In this way, it is similar to ghosts, Sasquatch, dragons, etc. And likewise I do not need to be able to prove dragons do not exist to have rational reason to not be scared of them, not make decisions accounting for their dangers, etc. It is true momentum in sports has not been falsified. It is true too that until it has been substantiated it does not need to be falsified.

by Led :: Fri, 11/22/2013 - 10:58pm

"The person making the initial assertion always bears the burden of proof."

Why? Any person making any assertion should have evidence for the assertion. The absence of sufficient proof of x is not proof of not x. What you are expressing above is simply prejudice. Not prejudice in the sense of bigotry, but prejudice in the original, more literal sense of "pre-judging." It's question begging in high heels.

"The initial assertion is 'momentum does exist.' 'Momentum doesn't exist' is a refutation. It's conditional. Without the initial assertion, the refutation is unnecessary, and we return to the default."

If you mean "conditional" in the logical sense, you're not using it correctly. If you don't mean it in the logical sense, then I have no idea what you are trying to say. If by the "default" you mean "either x or not x" then we agree. But I don't think that's what you mean.

by Theo :: Sat, 11/23/2013 - 3:25pm

If someone comes with the hypothesis that 'momentum exists' then the burden of proof is on that person to proof it. As opposed to someone else to disproof it.

Trying to proof that 'momentum does not exist' is just bad science. It assumes that the default is that it exists; yet it has never been proven.

Without proof, momentum is a rumor.

This brings me to my opinion that momentum does exist.
Good plays bring back an energy boost to the team. Endorphines and testosterone start building, pain is less, plays go easier, tackles are harder, scores go easier.

The real question should be, do good plays happen more often after other good plays - taking randomness into consideration and then, do good plays happen even more often after multiple good plays - which is the definition of momentum.

by BaronFoobarstein :: Thu, 11/21/2013 - 6:16pm

If you're going to argue where the onus lies consider that it would be a stunning find if one were to show it doesn't exist. There are just too many mechanisms at play that vary over time. Players get tired and so perform comparatively worse. Coaches spot a schematic advantage in game that they exploit through play selection. A quarterback gets sacked too many times and becomes overly distracted that it may happen again. A referee notices that a tackle who has been effective all day is actually subtly holding the linebacker.

by LionInAZ :: Thu, 11/21/2013 - 9:41pm

You've just made the argument against the idea of momentum. Talking about 'momentum' sweeps all those little (and sometimes big) adjustments that go on throughout the games by both sides. It just dumbs down the discussion. If Aaron Rodgers goes out with an injury and the Packers proceed to lose a 3-score lead because they can't move the ball, it seems pretty silly to me to talk about a 'momentum shift', when a n obvious explanation is smacking you upside the head.

by Bobman :: Fri, 11/22/2013 - 1:53am


Some guys can call that momentum, some guys can call that an injury that leaves a team inferior to what it was a minute ago. I think one view is trying to "make it seem bigger" and more appealing to our emotional human brains while the other side is just being a little more clinical. But both are describing the same phenomenon. It surely exists: Team A is playing well. Duh. The reasons they give (momentum versus an injury or bad play calling, or sheer luck) say as much about the people talking as they do about the events on the field. More maybe.

by Mussell Shoals (not verified) :: Fri, 11/22/2013 - 7:55pm

... none of which presupposes momentum. If I lose my queen to my opponent's bishop and can not recapture, I have not lost momentum. I have lost material. Momentum assumes a non-tangible advantage in excess of the tangible advantage. If a referee flags a tackle for holding, that flag and its impact are recorded. Momentum assumes something in excess of the penalty will change the game, through one player or team becoming discouraged, or fired up, etc.

It would not be stunning to discover momentum doesn't exist. It would be the simplest explanation possible. "Momentum" as a sports (etc.) cliche is figurative, borrowed from physics. Competitions can swing between outcomes wildly, no momentum required. Momentum assumes, like a rolling boulder, that one play's success or failure contributes to future successes or failures additional to the recorded, obvious impacts of that play. That, in the case of a sack, the sack contributes beyond the yardage and down lost. That the "energy" or whatever of the sack feeds into subsequent plays, and makes one team play better and one team play worse.