Was Uriah Hall’s spin kick in the co-main event of UFC Japan lucky, or even the dreaded “fluke”?
Luck is a pejorative in MMA, and in sport in general. We tend to tie it to the word “deserve”, and it often ends up loaded with disrespect. It’s reinforced by the way that luck is often used as the crutch of the bitter: fighters claiming that, although they lost, they’d beat their opponent 9/10 times. It’s a popular tactic used by salty fans.
Any conversation looking back at fights using probabilities can quickly dissolve into a sludgy mess of angry hypotheticals which normally deserves the same level of serious attention as P4P discussions. The natural response is to close the door on this kind of nonsense: a fight’s winner is who won it; there’s no such thing as a lucky punch, and someone throws a strike and connects with it is wholly responsible for the outcome.
Closing the door does block out a lot of potentially interesting and even “true” things about MMA. In trying not to disrespect fighters and in protecting ourselves from unnecessary arguments with idiots we can sometimes miss how filled with variance and fundamentally messy MMA actually is.
Dead dinosaurs and incomplete data
I once studied paleontology briefly, and it gave me a bit of an appreciation of the difficulties of working with incomplete data. The vision of paleontology which I started out with is one where animals die, go into the ground, and then people dig them up. Then, you put the bones together, try and figure out which one ate which from the pointy teeth, and a vision of an ecosystem is assembled.
It doesn’t work like that unfortunately. Only a tiny percentage of organisms die in a way which mean that they’re likely to be preserved. It normally requires getting quickly buried in some kind of sediment. Apart from anything else, this dramatically biases the fossil record towards shallow aquatic environments. Then, the fossils have to avoid getting crushed by geological processes, and then they actually have to be exposed and dug out of the ground. This means that, among other things, you’re hugely unlikely to get large complete animals. That sweet T-Rex in your local museum? It’s a frankenstein of casts, put together from a collection of animals which didn’t live in remotely the same time or the same place.
You’re probably asking: what the hell does this have to do with MMA? Because for me, the fossil record was something where I thought what everyone was looking at was representative… and it wasn’t. Instead, it represented a tiny pinhole look at a single moment in a vast world; incomplete data which had to be used to infer complex conclusions.
This is what happens in this sport. A fight is a very short and very isolated period in the participants’ lives, where they’re under powerful stressors, with a number of influencing factors we can only guess at including injuries, mental states, home life, and gym preparation, in a bogglingly complicated sport with far too many potential factors to keep a handle on. In predicting them we use profoundly limited information to try to guess at what is going to happen in the fight. Afterwards, there’s an attempt to ascertain what happened and why, but much of what happens and the reasons why are hidden. Fighters can get much better, or much worse, almost invisibly. Our own technical ignorance or bias are huge factors.
Probability and tennis; or why MMA is random as all hell
Most MMA fighters fight a single opponent once, or twice, or very infrequently three times in a career. At the extreme end of the spectrum Travis Fulton fought Rory Prazak 8 times in MMA and 3 times in boxing. When fighters win or lose we can state that the man who won was the better man. The narratives of the sport are constructed wholesale around the idea that the person who won -apart from in the worst of robberies- is the better fighter, and that the fight would go the same way if it was run again. Because rematches occur so infrequently, there’s nothing really to indicate that this isn’t the way that things work.
Is it really how it works? Let’s take a brief look outside of MMA and even combat sports, and take a look at our own Karim’s other sport: tennis. Like MMA, tennis is one-on-one, but what it’s useful for demonstrating to us is the sheer volume of rematches. As opposed to the one or two fights someone in MMA might have, people play each other 20, 30 times over the course of a career.
Here are some of the head-to-head matchups of, for example, Federer:
Federer 14W-11L Murray
Federer 10W-23L Nadal
Federer 21W-21L Djokovic
The supposedly better player (if there is one here) does not win all the time, or even 90% of the time. There is a huge amount of variance. Say Federer and Murray only played each other once throughout their careers: probabilistically, there’s a significant chance (44%) that Federer loses that fight, and that we’d just assume that Murray was better than him for the rest of their careers (God forbid they played and he went 2-0).
Of course, MMA and tennis aren’t the same, but of the two, MMA is inarguably far more packed with variance. It’s not even close. A tennis match takes place over at a longer period of time, giving more chances for a superior player to win out over a greater number of exchanges. Beyond this, MMA fights can be finished- a single exchange can end the fight, where all a favoured fighter’s advantages in speed, power or technique etc. are nullified. Beyond this, there are more nebulous and subjective factors, like the way that MMA is still technically a very immature but complex sport, and so defense lags far behind offense across the board. People get hit by stuff a lot and sometimes unpredictable stuff happens because of it.
If you wanted to give tennis outcomes something close to the sheer variance in potential outcomes of MMA, you’d have to do something ridiculous like sprinkling the court with a number of pressure pads which instantly won the match if you hit them hard enough with the ball. It’d make the matches more exciting!
What does this mean? It means that our small sample sizes in MMA are probably obscuring that other factors beyond individuals being “better” or “worse” play a far bigger role than we like to think. Even in a sport as relatively consistent as tennis, there’s a clear and obvious way that that just being better on the day can change outcomes significantly.
Probabilistically, sometimes shit happens.
Mousasi vs Hall
So what happens when we make a pick, and what happens in an upset? Often we construct a kind of simplified mental model of potential outcomes (I do anyway). It might look a bit like this if we turned it into a chart:
Don’t pay attention to the specific proportions, but more to the fact that Mousasi’s outcomes are way more likely than Hall’s. So, we pick him to win, and in my model it’s probably most likely to be a submission. As it turns out, we’re wrong and Hall knocks Mousasi out. Then we have a fundamental question to ask: why?
- Was it the model being wrong; i.e. our incomplete data and ignorance gave us an image of the fight which wasn’t accurate? Fundamentally, was one fighter better or the other worse than expected? As the most extreme example of this, look at Renan Barao-TJ Dillashaw 1.
- Or are we talking probability– was the model mostly functioning as planned, but threw up one of one of the more unlikely outcomes because that’s how probability do.
Now, our models are always wrong to a greater or lesser extent (models which simulate real life always are, by definition)… but this was one of most textbook examples of an upset matching up with how people thought the potential outcomes might line up, of the model appearing to be accurate but of probability (or luck) being the main effect. Hall’s main path to victory was considered to be something like this:
“Hall has to land the unorthodox strike of his life. Mousasi has an ridiculous chin, so Hall is going to have to land the followup of his life as well.”
Most people agreed that this was very unlikely, but then it happened. Mousasi steamrolled him for the first round, but then spinning jump kick, flying knee, and Hall’s excellent GNP to close the show. Nothing that happened indicated that our prior information about the two fighters was lacking.
Increasing the odds vs just doing it
Regarding the strike itself, Connor and Pat have gone over it in very insightful ways, both in their podcast, and in Connor’s article. The key to me is in the role which probability has in almost every strike.
A jab, for example, can be executed well, at the right time, but even then there are a number of things which can’t be controlled- a slight shift of the opponent’s head in one direction, and the jab can break their orbital, nose or jaw. A slight shift in another one and you might break your own hand on their skull. The classic “fluke” MMA jab is the one which Nick Diaz used to knock out Robbie Lawler- it was a fantastically executed jab (or hook, it’s a bit between the two), but now that we’ve seen these fighters over the course of many years, we can say with some confidence that Diaz landing a jab which starches Robbie Lawler is not a very likely thing to happen.
The Hall spinning kick needed a lot of things which Hall couldn’t control to happen, primarily involving Mousasi moving his head into exactly the right place. Even if Hall read the general direction of movement, landing the kick directly to the jaw was far outside of his or anyone else’s control, no matter how well the kick was executed. A high-percentage move does not need to be “carried” by variance in the same way.
More, as Connor and Pat also mentioned, Hall did little to increase his odds. He didn’t push his opponent into the spin kick by establishing a threat which forced Mousasi into it. An example of setting up a low percentage strike is the way that McGregor or Holloway mix the rear power straight with the spinning back kick to the body- avoiding these involves the opponent moving in opposite directions, so opponents trying to avoid the punch walk into the kick.
Hall just kind of did it and hoped it was going to work. To me this is one of the problems with his approach overall; that each of his clean, well-chambered strikes is an isolated knot of probability which doesn’t set up the next.
MMA has a lot of flukes
A lot of people have mentioned this win in comparison to other high-octane KO wins and asked if they’re considered to be flukes, like Renan Barao’s spinning back kick over Wineland. To me, they generally are. They come on a sliding scale, at which Hall’s strike is at the absolute extreme end, but low percentage strikes remain low percentage, no matter who throws them and how well they do it. Given the nature of probability, some of them will inevitably land, some of the time. Most of them won’t. In another reality, Ben Henderson took Frankie Edgar’s head off with a hurricane kick. One or more of Anthony Pettis’ incessant cartwheel kicks has knocked out one of his opponents. Low percentage strikes which miss tend to get forgotten, however.
Flashy finishes are traditionally dramatically overrated ex-post when it comes to accurately predicting fight outcomes in the future. Paddy Holohan’s running uppercut of Josh Sampo and Niklas Backstrom’s flying knee over Tom Niinimaki led to classic examples of over-rating fighters due to highlight reel swag.
I do like Uriah Hall a lot. I have a lot of sympathy for gifted athletes who get stuck inside their own heads. I was very happy to see him pick up the win. I think that ranking him under Mousasi because it was a big upset is stupid. It was a beautiful moment, one of my favourite of the year so far. It wouldn’t happen again, and won’t for a long time.
About the author
