Why Alex Ferguson Doesn't Care About Your ExpG Models (PART TWO)

In part one of this piece, I said that Alex Ferguson’s 2012/13 Man United team drastically overachieved their expected goals total and I showed where almost all of the extra goals came from (in and around the edge of the six-yard box). In part two, I’m going to explain how United posted such an underwhelming shots profile despite being a dominant team.

Strangely enough, it all starts with Theo Walcott…

The ‘Walcott Effect’

At some point after the 2011/12 Premier League season, Paul Riley posted a piece in his blog about Theo Walcott’s crossing. In it, he described how rarely Walcott’s crosses found their target.

Here’s Riley,

“It takes [Walcott] nearly two games to put in an accurate cross. He crossed the ball 134 times last year. Only 18 (EIGHTEEN) of them found a teammate.”

But (more Riley),

“When you look closer at Walcott’s accurate crosses the only way to describe them is ‘devastating’. 8[1] of the 18 crosses have been assists.”

It’s counterintuitive: less shots, more goals.

Now you could chalk that down to fluke, or to some otherworldly finishing performance by the other Arsenal players[2], but Riley doesn’t buy it. He thinks that when Walcott’s good at ‘putting it on a plate’ for his teammates. He shows how 6 of the 8 Walcott cross-assists come deep inside the opposition box, and makes is easy to imagine how Walcott’s crosses are easier to finish than the deliveries of the ‘into the mixer’ crowd.

He’s convincing because it jibes with what you see when you watch football. Getting into the opposition box with the ball at feet is the scariest thing a wide player can do. That’s what makes the Theo Walcotts, the Angel di Marias, and of course, the Cristiano Ronaldos of this world so good. When they do that, everything goes to shit; goalies stay pinned to their lines, defenders abandon their assignments, and the whole time, they’re all too terrified to make a tackle.

Buuutttttt… But it makes it harder to cross the ball. Put yourself in Theo’s shoes. You’ve left your man for dead, you’re in the box with the ball at your feet. It’s nice that you’ve pulled all these players out of position, away from the men they’re supposed to be marking, but it means they’re coming for you! They’re blocking your shot, they’re clogging up your passing lanes, not to mention that unless you place your delivery within inches of your striker’s foot or noggin, he’s not going to be able to react fast enough to direct the ball towards goal.

That said, if you can manage to thread the eye of the needle, if that cross finds its way to a teammate, almost any shot on target is going to be a goal.

When you’re reading the stats, that looks like less shots, more goals. 

Maybe you can see where I’m going with this…

Heating Up

Look at the wide players on that 2012/13 United team; Patrice Evra, Rafael, Nani, Antonio Valencia. There’s no messing about with ‘central wingers’ or ‘inverted fullbacks’, those are true-footed players with dribbling skills and real speed.

You like overlapping runs? Eat your heart out.

Those Walcott-esque crosses are the type of deliveries this 2012/13 Man United attack was designed to create.

Don’t believe me? Take a look.

fergie crossesmoyescrosses

That’s a heat map of all the crosses United took in the 2012/13 season (actually I’m missing data from a couple of games, but you get the idea). The second, just for comparison’s sake, is a heat map of the crosses United took when David Moyes was in charge. The difference is stark, Ned Stark.

Both 2012/13 and 2013/14 United based their attacking philosophy around crossing. Under Fergie, United actually crossed the ball  (~25 per game under Fergie vs. ~24 per game under Moyes).

It just didn’t feel like that.

‘Moyes’ men’ lobbed in a lot of crosses from deeper positions, and even when they got closer in, they rarely penetrated the opposition box. On the other hand, ‘Fergie’s buoys’ passed up on the lottery balls and frequently crossed from position deep inside enemy territory. I’m not putting that all on Moyes; Evra was on the wrong side of the aging curve; Rafael was in and out of the lineup with injury. But some it’s his fault; he tinkered constantly; he played guys like Phil Jones and Chris Smalling at fullback… But that’s a different discussion.

Why Fergie’s 2012/13 United Team Fucks Up Your ExpG Models

Phantoms. That’s what I’m calling them anyway. Goal-scoring opportunities that never have the chance to be registered in shot-based ExpG models, because they never turn into shots. These crosses are difficult to complete. It’s hard to pass through a thicket of defenders, and on the finishing side, it’s hard to react fast enough to get shot off. But make no mistake, even the ones that don’t make it to feet are (non-negligible) chances to score.

If you fail to take them into account, if you don’t include them in your expected goals models, you will be underestimating teams that use these tactics. Team like Alex Ferguson’s 2012/13 Man United.

Other Factors

The ‘Walcott Effect’ is not the only reason why 2012/13 United deceive ExpG models, but this article is already too long, and those other factors have been intelligently covered elsewhere. But briefly…

Robin Van Persie is Very Good at Scoring Goals

Robin Van Persie is an underrated passer, an intelligent runner, and has perfect balance and technique. He’s very good in any type of offense, but he was the perfect finisher for this ‘drive and kick’ style offense.

Robin Van Persie is Great at Taking Set Pieces

Possibly the one thing Van Persie does better than finishing is taking set-pieces. In case watching Arsenal struggle flounder on set-pieces over the last few seasons wasn’t enough, this is Mike L. Goodman writing for Grantland:

“Last season (2012/13), in the attacking third when van Persie was on the field, he was responsible for 49.8 percent of all free kicks taken by United.”

And that for a side which overachieved its ExpG values from set-pieces by a pretty ludicrous margin.

Winning Penalties and Scoring Own Goals Can Be a Skill

There’s a good deal of luck that goes into getting these freebies, but you have to think that a team which specialises in putting the ball at the feet of quick-footed dribblers who attack the box (and whip in balls from the byline) has a good chance of getting a few of both each season.

The 2012/13 Man United vintage was not lucky, it was a dominant team that had the title race sewn up by April. Fergie’s  ‘buoys’ were well trained at set pieces and they had an elite finisher in Robin Van Persie. But that extra ‘oomph’ in conversion rate? That was the result of phantoms: goal-scoring opportunities that don’t show up as shots. It’s not sorcery, it’s not special sauce, it’s not some kind of cosmic imbalance. It’s just what a ‘drive and kick’ offense looks like on a football pitch.

[1] Riley actually says 6 of 18, but then goes on to talk about 8 cross-assists, so I think it’s 8 of 18. [2]

After all, he was playing with Robin Van Persie that year.


Alex Ferguson Doesn't Care About Your Expected Goals Models

The Past is a Foreign Country; That’s Why It’s So Easy to Critique

When the facts have seeped out, when you’ve had time to think things through, it's all so obvious: Germany had the best, deepest squad in the tournament; Luis Suarez can make any manager look like a genius; and "c'mon, you never realised he was on 'roids?".

Haha! Totally joking about that last one... Totally. #GTL.

But some things defy explanation, they just sit there as you examine them, laughing back. Some things are like Manchester United’s final season under Alex Ferguson.

Okay, maybe I’m overstating it a little, but that 2012/13 season is a puzzle.

United cruised to the title, closing things out at Villa Park with a couple games to spare. RVP, Shrek and the Buoys clubbed together to score 82 goals in 38 games, gifting Alex Ferguson his 13th and final Premier League title.

So far, so good.

What's odd is that United posted the stats profile of a good, but certainly not great, team. Total Shot Ratio (TSR), usually a pretty good indicator of quality, ranked United in 8th place for the 12/13 season, sandwiched between Southampton and Newcastle. On its own that's not a game-changer, after all, TSR would have put Spurs atop the league that same season. But even Expected Goals (ExpG) models, which take into account more complex shot data like distance from goal and where the pass prior to the shot was played from were are flummoxed by how this version of United so drastically outperformed expectations.

Whatever Remains, However Improbable, Must Be the Truth?

That, apparently, is the opinion of Dan Altman.

"I computed the probabilities for the 2012-13 season... For Manchester United, the difference between expected goals and actual goals was sixteen. Without those sixteen goals, Manchester United probably would have finished fifth instead of first."

That's a quote from Altman's latest piece in the New Yorker on the same topic, in which he uses his bespoke shots model to break down exactly how United overachieved offensively.

His conclusion? Multi-dimensional luck (more or less)[1].

Luck on penalties and own goals. Luck on free-kicks and corners. And then some more luck converting chances in open play. He says that if you ran the season back in his shots model, United would be extraordinarily unlikely to score such a high number of goals again. In which case, the team probably wouldn’t have won the title, in which case, Alex Ferguson might not have been quite so quick to retire, in which case the world would be short a great wordsmith, (and Davey Moyes wouldn’t be aging in dog years).

Altman’s points are pretty convincing, especially in the light of how much United have struggled since then, but even he admits that the odds of this level of over-performance are pretty fracking long.

Which is why there might be a different answer. I’m not saying he’s wrong, but I think there’s an explanation that demands shorter odds.

Just In Case, Here’s What We’re Looking At

Usually, when I'm looking for an answer, it helps to take things down from the abstract clouds of ExpG models and put them into a form that’s more easy to understand.

Here’s Man United’s shot chart from the 2012/13 season.

*The hexagons represent the number of shots taken, and the color indicates how efficiently United converted their chances into goals. Red is good. Blue is bad.
*The hexagons represent the number of shots taken, and the color indicates how efficiently United converted their chances into goals. Red is good. Blue is bad.

One thing is probably going to stick out to you - that big red clump with the YOWZA!!! annotation (or maybe you're a data hipster, in which case you're probably into the five light-blue hexs). When Altman says Man United overachieved their expected goal output, this is why. In Fergie’s last season, Man United took the juiciest shots in the game (close to and directly in front of goal) and squeezed, and squeezed, and squeezed some more. The shots in that bright red area were converted at roughly 13% above what you would expect the average Joe (Allen) to do with the same opportunities. By my calculations, based on the number of shots, that's worth somewhere between 17-18 goals over the course of the season.

(I swear I'm not making this up...)

Now, Marco Van Basten’s one hell of goal-scorer, but a) He retired in 1995, and b) He never played for United. Robin Van Persie did, but he's just one (non-superhuman) dude, and he didn’t take ALL of those shots…

At any rate, other than that one scorching zone, United are surprisingly average finishers – a couple points stronger on the wings, but nothing to write home about[2].


You have to say, it doesn’t look completely random. It’s not like United are lights out from a bunch of unconnected areas. This is a team that is PHENOMENAL at scoring in one area.

Why? Well if you want answers this morning, you’re shit out of luck. I’m tired and I’m looking forward to the latest episode of Serial.

Tune in for part two tomorrow.

[1] I’m shortchanging Altman a little bit here. He also talks about things like RVP's finishing quality, but you can read it yourself, ‘cause, you know, the Internet.

[2] Interestingly, this version of United doesn’t shoot a lot from distance, which probably goes some way to explaining the low TSR numbers, if not the ExpG anomolies.