Apparently, according to ‘sources’ (probably his agent) Man City want to buy Theo Walcott this summer. Man City – which means Pep Guardiola wants to buy him! I consider this unlikely. For one thing they have Sterling already, who is of similar development as a player but far younger, and secondly Theo just isn’t good enough for Pep to want him.
Theo is a nice lad, polite and well-mannered, no doubt loves and respects all his family and is kind to animals. Unfortunately that doesn’t make him better at football, or worth a reputed £140k a week. Lots of Arsenal fans have noticed that he’s not been playing very well recently, and many, me included, have run out of patience with him. If he was going to develop into a world class player he’d have done it by now.
But as usual in football, different opinions abound. Some say that you need to look at his stats to see that actually he’s very good, because he’s completed a certain percentage of passes, or scored a certain number of goals per 90 minutes played, or created a certain number of chances, or any of a dozen other things that now get measured. This is a gross misuse of stats.
The problem with nearly all stats in football is that football is an almost infinitely dynamic game. No two situations are exactly alike and most situations have astronomically large numbers of variables. The closest you get to an exactly repeated situation is a retaken penalty. What are the variables here? It’s the same two players, the ball is in the same position, the keeper is still on his line; the attacker runs up and shoots in the direction of the goal. Leaving aside the Cruyff/Olsen variation of passing to a teammate who runs into the box, if you repeat the action there is not much different in either the set up or potential execution. So you can look at instances of this situation and see from the results of them whether the shooting player or keeper should have done better. It’s the football equivalent of pitching in baseball or bowling a single ball in cricket. The difference is that those actions ARE baseball and cricket, while penalties are not even one per cent of football. Stats from penalties tell you how good your penalty taker is at taking penalties and to a lesser extent how good your keeper is at saving them (lesser because a well-taken penalty can almost never be saved other than by anticipation – a measurable skill in itself given enough data, but if a keeper anticipates too much then a good taker will adjust – so there is far more useful data about the taker than the keeper).
There is another variable with penalties (as England players and fans will be aware): pressure. The Times have researched penalty stats from the Premier League from its inception in 1992-93 right up to 2013-14. In this time there were 1,925 penalties awarded. Of these, 218 were given to teams who were either winning by at least two goals already, or losing by at least three. In these situations, whether the penalty is scored or not is unlikely to make a difference to the match result, so there is little pressure on the taker. This is reflected in the outcome, with just over 86 per cent being scored.
In the other 1,707 penalties the outcome could either make the teams level or bring the score back to within a goal. Clearly this situation has a greater chance of affecting the result, and the success rate dropped to just under 75 per cent. So it appears that on average players score more often from penalties when they are under less pressure. On the other hand it may be that goalkeepers try harder when it makes more of a difference. Or both. Whichever way it is, it demonstrates a pattern – one that is useless to managers and coaches, but potentially useful for punters involved in spot-betting during matches: bet more on a goal when the penalty matters less.
Despite the uselessness of this information to clubs, for me this is a ‘good’ stat. It explains by using a large data sample from a long period what is (slightly) more likely to happen in a given situation. Moreover, the stat can’t be (mis)used to ‘prove’ anything else.
Back to Theo. What stats can be used to ‘prove’ whether he’s any good? Let’s just take a couple of examples: passing accuracy and goals scored per 90 minutes on the pitch. Now I accept that passing accuracy is not normally the first thing people point to when judging Theo, but I’ll do it anyway. Bear in mind I’m not really judging Theo so much as judging the use of stats to judge him, if you follow that.
Tactics play a part in passing stats: Arsenal largely play a short-passing game where possession is key. Thus ALL their players should have very good passing stats because more often than not it’s ten yards maximum. Back and forth it goes: you have it; no you have it; no you have it back. Some statisticians now count how often the ball is passed forwards; in Arsenal’s case just marginally more than it goes backwards. Whether Theo passes forward more than his teammates I don’t know. Frankly it’s unimportant, because there’s another variable that is never measured, and it’s the problem with any stat to do with an individual player: did he make the right decision? Did he do the best thing with the ball at that precise moment, given the state of all the other variables – where all the other 21 players are, what they’re doing, which way he’s facing, how fast he’s moving, how he received the ball, what height, speed, spin was on it… the list goes on.
This is the dynamic nature of football: any player can be anywhere on the pitch moving in any direction at any speed – as can the ball. It’s little use if Theo or anyone else passes to a team mate 100 per cent of the time if he never chooses the best option. But the best option is only measured by watching the game and seeing what everyone else is doing. This is how coaches have judged players since organised football began, and if you watch enough football and have the right mind for it then you get to be very good at it. I should stress that this alone does not make you a great manager – there is an awful lot more to that, including being able to man-manage players as individuals, which few people can do really well.
Think about another stat for Theo: goals scored per 90 minutes played. This is often used by the terminally lazy or stupid to compare how good one striker is against another. It’s a very rough indicator of talent (or the application of it) but nothing more. Suppose Walcott scores 0.6 goals per 90 minutes and Giroud scores 0.5. Is Walcott better? Maybe yes, maybe no. You also need to look at, among other things, how many chances they both had, how good the chances were (loads of variables there), what tactics the team adopted, what exact role each of them were playing (was it to their strengths?), what tactics the other team adopted (perhaps man-marking), the strength of the other team, tiredness and injuries of both teams, and who their teammates were on any given day and how well they were all playing. And all that is before we think about overall contribution to the team effort and whether each player made the best decision when shooting, or whether they should have been setting up a teammate for an easier chance. You can get some of this info from the stats, but far from all. I’m not saying Theo is a bad player based on stats. My point is just that they don’t prove he’s a good player either. For truly exceptional players such as Messi, stats may confirm greatness, as they’re so far from the average, but do they prove he’s better than Ronaldo, who plays at the same time in the same league? Never mind proving anything against Pele, Maradona, Cruyff or Di Stefano.
Mesut Özil is another Arsenal player who divides opinion, even among Arsenal supporters. I think all Arsenal fans recognise that when he’s got the ball he has great vision, is a superb passer and can create chances seemingly at will. There was talk of him breaking the Premier League assists record of 20 held by Thierry Henry – I’m told assists have been recorded since the start of the PL, though of course as an isolated number the information is of highly debatable use. (He’s also allegedly the world record holder for most assists in a career, which is another gross misuse of stats by ignoramuses.) His assist rate has fallen off in the last couple of months, but even so I don’t think we need to worry too much about whether Özil is making the right pass most of the time, as it’s clear by watching him that he is one of the rare players whose vision is well above the norm. He’s not going to get everything right every time, but he instinctively knows what to do with the ball and where the good options are. There are lots of stats that will back up Özil’s effectiveness in possession, and his most ardent fans will gleefully quote them.
In fact one of my Arsenal-supporting friends who has a popular twitter account wondered if Özil was as good or better than Dennis Bergkamp, as their stats for goals and assists are comparable. This again is misusing stats, in this case by taking two isolated numbers and proposing that a simple comparison is enough proof to compare the effectiveness of two players. Fortunately sanity largely prevailed, and when I pointed out that Bergkamp was a better player irrespective of these particular stats, the vast majority immediately concurred. One summed it up by saying Özil is a great playmaker, but Bergkamp was a great footballer, which is about as good an explanation as you’ll get in one tweet. In any case there are no doubt players at other clubs and in other leagues who have comparable goalscoring and assist stats, but that doesn’t make them fit to lace Dennis Bergkamp’s boots.
Johan Cruyff pointed out that in every match each player has the ball for an average of about three minutes, which means that he doesn’t have the ball for an awful lot longer, even allowing for the fact that it’s usually only in play for a total of around 65 minutes of the 90. Cruyff’s point was that what a player does when he doesn’t have the ball should be a bigger contribution to the team effort than his actions with the ball. So what does Özil do when he doesn’t have the ball? There are two situations to think about: either his team has the ball, or the opposition does. Without referring to reams of stats, I’d say that when Özil’s team is in possession he fully contributes, as his positional sense is excellent by any standards and he has the skill to receive the ball and use it for the benefit of the team even when fairly closely marked.
When the opposition have the ball he is a bit less productive. Specifically when the opposition win the ball with Arsenal fairly high up the pitch and Özil ahead of the ball, he doesn’t appear to feel the need to get back behind it very quickly. (Admittedly some other Arsenal players these days don’t set a great example!) Perhaps the best way to illustrate this is to say, what if you were playing a 2-a-side match: one outfield player and a keeper – would you want Özil as your outfield player? Well not if he wasn’t going to bust a gut to get behind the ball. In a team of 11 the effect is lessened, but it’s still there.
Players like Özil used to be called ‘luxury players,’ and some would argue that it’s worth having luxury players because a single moment of brilliance they produce can win a match. There’s no definite right or wrong here, because of course the nub of football, part of its beauty, is that if your defence stops the other team scoring then a single moment of brilliance to create a goal for your team can win matches and even trophies. And even if it doesn’t, as Cruyff again pointed out, people remember the beauty, the style, the philosophy and adventure of great play as much and sometimes more than the result. His Holland team of the mid-1970s didn’t win the World Cup, but their football is rightly revered.
Obviously there is more to defending as a team and making a contribution to the team than just getting behind the ball. There is marking, pressing, tackling, intercepting, and supporting your teammates by covering them when they’re doing all these activities. How much of this does Özil do and how good is he at it? I don’t know what his stats are, but even if they’re all available, are they necessarily measuring the right thing?
Suppose Özil makes four out of five attempted tackles in a match (it’s just an example; bear with me). Does this mean he’s played better than Koscielny, who made four tackles out of eight? Not necessarily. He might have been in the wrong place another five times and consequently not been able to make tackles that he should have at least attempted. He might not have got back behind the ball, meaning that a teammate had to attempt a tackle instead, possibly when he was less well-positioned to do it, which gives a higher chance of failure.
Another Özil stat often quoted is that he seems to run further than anyone else on the pitch. What does this prove? Does it prove he’s trying hard? Probably. Does it prove he’s effective? No. He might be, but if he’s running at the wrong time or in the wrong direction it’s not going to be helping the team. Conversely he might be making huge efforts to run into devastatingly good positions, only for his less visionary teammates to fail to get the ball to him. The fact he’s run 11km in a game isn’t useful to know without a lot more information.
The point of all this is that stats alone are not enough to judge a player or compare one with another, especially – ESPECIALLY – if you’re cherry-picking one or two stats and presenting them without context.
All stats have to be put into context, and some need more context than others – sometimes including what everyone else on the pitch was doing at the time and was the best decision made. You can – so far – only get that by watching the game, and to be honest re-watching it several times.
However, some analysts think they’re getting to the stage where their stats can really influence things on the pitch. Arsène Wenger is a convert. He works closely with Jaeson Rosenfeld, the American founder of StatDNA, the company Arsenal bought. Gabriel was Arsenal’s first signing who had been initially flagged up by analysts’ stats rather than traditional scouting, and Arsenal are widely held to be at the cutting edge of stats usage. My question in that case would be, why did Leicester win the Premier League while Arsenal faded after leading in January? Clearly the whole subject has a long way to go.
So when will football stats get to the point of being able to measure things like the best course of action at any point in a game? Perhaps never. Each player makes literally thousands of decisions in every match, many of which are dependent on what other players are doing, and most decisions are made without conscious thought. There would be so much information that it would have to be compacted and summarised in order for any coach, pundit or fan to assimilate it. In which case you may as well, you know, watch the game…
In part 2: Were Leicester lucky? What are football stats good for anyway?