StatsBombIQ StatsBomb Data
October 17, 2016

xCommentary

By Ted Knutson

I caused a bit of a kerfuffle on Twitter yesterday with my comments about football commentary, so much so that I thought it worth the time to explain it in longer format here.

For those who are blissfully unaware, this is what I said:

comms_1

comms_2

comms_3

comms_4

That last part was the real impetus for the complaint. My son watches Match of the Day religiously. Every Sunday morning, he sits captive in his pew, listening to the wisdom of the commentator pastors preach about football.

Sitting there together Sunday morning, watching the Arsenal highlights, we heard this from Phil Neville:

“Walcott should have scored 4, or at least 3…”

MotD then walk through the first two goals, the first made via graft from Theo, the second mostly by being in the right place at the right time (though it’s a beautiful knockdown and turn). Then we get highlights of the next chance…

“He’s GOT to score. It’s a simple chance.”

walcott_chance_50res

The shot is 7-8 yards out, off a headed pass. Walcott hits it cleanly, on the volley… WITH HIS LEFT FOOT. Now granted, it went straight at the keeper, but the problem here is this is anything but a simple chance.

Let’s break it down by layers.

heads_vs_feet_final

First, it’s a shot with feet, 7 yards out. That’s between a 40-60% chance of scoring.

However, it’s actually taken out of the air on the volley/half-volley from the side, which makes it closer to a cross. Now we’re in the 20-40% probability range, and probably at the lower bound because it’s a volley.

And finally, it’s with Walcott’s weak foot.

I’m an Arsenal fan and have been for about two decades now. There was definitely a period of time where I wasn’t sure Walcott HAD a left foot.

Now this is a guess because the sample size is too small to actually calculate it, but let’s say Walcott is probably half as likely to score from there with his left foot than his right.

We have now gone from Neville’s “simple chance” (which was never actually simple) to at most a 20% likelihood of it being a goal.

And this is my problem with commentary like this: It’s wrong.

Not just off-by-a-little-bit wrong, or yeah, mostly-in-the-same-range-but-imprecise wrong. This is completely-misanalysed-what-happened-on-that-chance wrong. It would be more fair to say Walcott did well to keep the ball down and get it on target with some power than it is to say he missed an easy chance.

The bigger problem here is that this happens all the time. We have commentary on football that doesn’t understand how the game actually works. I have heard commentators say goals should be scored off headers 12 yards away from goal. They must watch most of their football on some other planet, because that’s not how football here on Earth actually works.

Neville told the audience a very difficult chance was simple and that Walcott should have scored, neither of which is actually true. This is an opinion that is then easily refuted by data.

Lest you think I am cherry picking, let’s fast forward to the analysis of the WBA v. Spurs game in the same show.

Talking about Dele Alli, Shearer says, “I love the way he gets into the positions. He’s not afraid to miss them” (referring to shots).

*nodding along* Cool. Me too. Dele is such a clever midfielder in and around the box and I really love his sense of spa…

Shearer: “He’s got to work on his finishing.”

Wait, what? Why? Why would you say that? Aside from the fact that Alli is young and all football players presumably work on their finishing in some way, why would Shearer explicitly say this about Alli?

Unfortunately, we never find out. The show then cuts to show Alli’s goal, which just happens to feature an insanely good finish.

alli_goal_lowerred

Nine yards away, outside of the boot, across his body, in the corner and away from both the keeper and a defender on the line. The level of difficulty on this is incredibly high.

Funnily enough, if you look at both Alli’s and Walcott’s goal scoring stats, you see that they actually score more than we expect them to versus expected goal models. This is a pretty reasonable indication that both of these guys are at least above average when it comes to finishing chances.

But the Match of the Day analysts are telling us Theo missed a simple chance and Dele Alli needs to work on his finishing.

We Can Do Better
I don’t think Match of the Day or general football commentary needs to become a bastion of precise statistical analysis. It’s possible you could post the expected goal of a shot as a pop-up graphic on the screen, but it’s probably not necessary, and I certainly don’t want Alan Shearer saying a shot from a certain location is normally a 43% chance of being a goal.

However, I do think our commentators need to be less wrong.

Telling kids that the Walcott chance was simple and “should” have been finished is incorrect. This happens constantly, and I think the reason it happens is because so many of our ex-footballers who are now commentators were poorly trained in what should and should not be a goal.

The only way to overcome this is through education, and the best way is teaching via example. Get the data from their own finishing, or from that of their teammates, or current favourite players, or whatever, and walk them through the actual probabilities. I have done this myself with both footballers and coaches, both of whom have found the concepts very easy to grasp.

Additionally, telling the audience that Alli needs to work on his finishing might be fine, but you need to justify it with actual logic and example, and not just use it as a throwaway line.

The great part about having guys like Shearer on Match of the Day is that he was an amazing goalscorer. Leverage his knowledge to talk about what he thinks players could do better in front of goal, but in specifics not clichés.

Why do you think Alli’s finishing needs to be better beyond the fact that every shot he takes does not turn into a goal?
How does Alli make his finishing better?
What part of a technique does someone need to improve to allow them to put away these allegedly simple chances? Or did the goalkeeper simply close off all possibility of a chance becoming a goal by good anticipation and positioning?

So many times there are coaching badges on the couches that go completely unused as part of the commentary, but these little elements are the actual expertise the analysts bring to the show. Segue into past video analysis of Alli’s poor technique and suggest how Pochettino could improve that via drills going forward.

And while on the topic, maybe producers need to stop assuming that these ex-footballers actually know everything there is to know about football. It’s now clear to the general audience that they don’t, but somehow that never stops them from being both overcritical and constantly speaking in cliches that lack insight.

Story Time
The models for expected goals aren’t perfect. They never will be. That is an up-front admission that doesn’t remotely limit the impact of using them.

Right now the models are so much better than the pundit estimations it’s farcical.

It doesn’t have to be that way, and with some experts it’s definitely not.

Last summer I sat in a room with Bob Bradley – you may have heard of him – and I talked to him about how our models work in evaluating chances. His reply to me was that there are plenty of instances where a chance is either better or worse than the model thinks it is, because someone is blocking a player’s shot, or because “two of my guys have a body on a player trying to make a header” close to goal, or any number of other reasons.

And he was totally, completely correct. Without tracking data, we lacked the appropriate info to evaluate the different factors he throws into evaluating his own team, and even with tracking data, we couldn’t see everything with his expert eyes.

On the other hand, Bob’s eyes can’t evaluate every touch in every game across 27 different leagues. And no one in the world could sell us tracking data to go along with it. We could neither clone/scale Bob’s expertise, nor buy data that would let us make the dramatic improvements he wanted, so we simply did best with what we had.

Which, at that time, was probably better than 95% of other clubs in the world in using and evaluating data.

Expert eyes plus the data is far better than the data by itself. But right now, data is dramatically better than the “experts”, or at least the ones we see regularly on TV.

At some time in the future, it will be commonplace for experts to leverage data models and tell us why something was a better or worse chance than we might think based on the specific game situation. Unfortunately right now, using data at all is still a scary thing.

No One Will Watch/Read That
Back when I started writing about football stats in 2013, penalties were lumped in to all goalscoring stats, and no one in media ever used rate stats like per90. Nowadays we see some level of sophistication in both of these topics, even in mainstream outlets.

Basically, we are three years down the road, and a small proportion of the media talks about football in the way we started to back in 2013. Should we be satisfied with that? Hardly. The advances made in the last three years in public football research and knowledge about the game dwarf anything we knew back then.

The two things that haven’t changed at all are the older generation of newspaper writers and the entirety of television punditry. Columnists like Martin Samuel and most of his generation are a lost cause, which is fine because I’m not sure having them as champions for new things is a comfortable fit anyway. Television though…

Here’s the thing – the audience for football on TV is changing rapidly. Consumers have more football knowledge at their fingertips than ever before, but somehow almost none of that makes it through to broadcasts. Meanwhile, nearly every other sport on the planet is rapidly pushing forward with ways to add insight to their product.

The combination of these two elements makes football look like the kid wearing a dunce cap in the corner of the classroom.

It’s really not hard to talk more intelligently about the game. A lot of the work done in analytics is easily explainable to footballers, which means it’s both useful and digestible for a television audience.

Arsene Wenger mentions expected goals in a press conference. Do we think that’s by accident, or because he uses that as a tool to determine how well his team and individual players are executing on the pitch? And IF the concept is completely embedded at Arsenal and other teams in the Premier League, maybe it deserves to be broadcast on television?

Instead we get “distance run” stats, which to my knowledge have never been proven as relevant to anything in football other than telling us who happened to run further in a match. Which must be interesting for reasons that are completely unknown to me, because I have always been the type of person who wanted to figure out how to get the best results from the least amount of work, not the other way around.

I had a throwaway comment yesterday that we should take the Sunday Supplement panel and privately ask them how likely they thought a number of different shots were to turn into goals, then discuss the results. This sparked a lot of ire from certain members of the journalistic community. To me it’s an interesting experiment because again, I don’t think the actual difficulty of chances is very well understood, even among people who cover the game on a daily basis. To some journalists, I had clearly profaned their own Sunday morning church of aspirations with the prospect of making their elders potentially look silly.

My point was not to make people look bad out of ignorance. My point was that I want our press to speak more intelligently about the game, at least in this respect, and the only way you can do that is via education.

I am a huge, avid consumer of sports media. I subscribe to both Sky and BTSport. I am a Guardian member. I have been paying for ESPNInsider nearly constantly since 2002. I have also written for the Daily Mirror and appeared (briefly) on Sky, TalkSport, and BBCRadio 5Live. When I say we can and should do better, my criticism is both as a paying customer and as a producer of media content.

Back when Monday Night Football first started, there was a clear common belief that the average football fan probably wouldn’t be interested in smarter television about tactics and analysis. MNF blew that idea out of the water, but strangely we haven’t seen a flotilla of clones try to copy the formula.

I’ve heard the argument that people won’t watch or read more smart statistical analysis with their football either. They might be right, but it’s not as if anyone has really tried.

Meanwhile, Fantasy Football is one of the fastest growing ways to interact with the Premier League. There are literally tens of millions of unique players who take part in fantasy football leagues every week – a game that is entirely driven by statistical output of players.

This is to say nothing of the legions of Football Manager and FIFA Ultimate Team fanatics who also deep dive into games that are also entirely driven by player stats.

Hmm…

Meanwhile, in an age of media cutbacks, the world’s largest sports media company ESPN just launched a stats and analytics vertical because they thought there was a rabid audience out there who want to read smarter analysis of all sports, so it made sense to collect that information into one easily accessible place.

Hmmmm…

Oh my god do people gamble every week on the Premier League. With odds. That are probabilities.

Hmmmmmm…

OptaJoe has 852K followers on their English account alone. WhoScored and Squawka nearly 600K a piece.

¯\_(ツ)_/¯

There is an audience out there. Leveraging their interest while making a workable media property is the tricky part, but some combination of fantasy sports, gambling, statistical analysis (especially visualizations) and tactical insight probably makes sense.

Conclusion
I want smarter football commentary.

I want it so that my children, who are rapidly becoming football consumers themselves, can understand more about the game they are learning to love without me having to fact check every step of the way. And I want it so that I don’t want to simply turn off the sound or fast forward the analysis section every single time I watch your program.

The bar here is pretty low. I have personal experience teaching a lot of these concepts inside of football, and they are honestly not that hard to grasp.

I also want to listen to experts who clearly have something unique to contribute, and see them show their expertise in a way that teaches me new things. And I want to hear far fewer cliches from guys who are paid significant sums of money to appear on my TV and contribute knowledge and insight while doing anything but.

I will take insightful, lesser-known commentators and analysts on my football every time versus well-known, cliché-spewing sound machines. The vast majority of people I talk to will too.

There is an audience here. If you build it, they will come.

Article by Ted Knutson