Recently 7AMKickoff did a great blog post
about how important total spend (ie wages plus transfers) is in football success, rather than just wages spend being important, as recent theory has had it. It was a very detailed and informative blog post, so first may I say how much I admire anyone who puts as much work into a blog as that. I always stand in awe of people prepared to make that much effort and do that much research. Personally I’ve always thought and indeed said to anyone who’d listen that total spend is important, I just never had any research to prove it. So it was nice to feel vindicated.
I’m not a trained statistician, and as I’ve blogged before I’m firmly of the view that the recent overuse of stats in football is largely ridiculous and can be used to prove or disprove almost anything. Which player makes most passes? Who cares? You can’t tell how well someone has played just by counting passes, or even ‘successful’ passes, because you don’t know whether the pass made was the best option at the time (even if it was deemed ’successful’ by going to a teammate), whether the player who received the ball immediately lost it, whether a goal and a different match result would have occurred by doing something else, like shooting instead of passing. You’ll only know all that by recording hundreds of other stats or, far more simply, by watching the game.
I’m also a bit wary of people who are too trained up in stats because they can over complicate things and rely on stats to give them all the answers, and in real life that doesn’t work unless you have a closed system with limited variables. In football – and I’ve said this before as well – there are too many variables to predict winners and losers with certainty, which is of course the beauty of the game. In football, or weather forecasting or the share market, stats can only really tell you what’s happened in the past. You can make generalisations about the future, but you can do that with lottery numbers as well and you’re not going to make a fortune from that.
Having said that, I like the kind of stats that 7AMKickoff presented, because they’re ‘big’ stats. Who spends the most on wages and transfers is one figure that is made up of dozens of smaller figures and therefore it’s useful because collecting all the data averages things out. Similarly the stat of where a team finishes in the league is made up of so many other stats all put together over nine months that it’s a good measure of what league football is trying to find out – namely, which team is best. That may sound obvious, but we tend to take these things for granted.
A chart that’s nothing to do with this blog, just to break up the amount of text
So the research quoted by 7AMK matched up the ‘Who won the league?’ stat (and its infinite variables) with the ‘Who spent the most?’ stat (some variables) and found that the clubs that spent more finished highest, and the clubs that spent most finished at the top – in general. The more variables are included and then averaged out, the more likely you are to find reliable patterns. Man City winning the title last season was great for lovers of stats, because hey, they spent the most! So they were bound to win, right? But don’t forget that they only won in the last minute by the slenderest of margins – well, the slenderest since 1989, anyway. Any number of things could have prevented that last goal, last win, last point: the ref could have added up injury time differently and blown early, one stray pass or shot in the last move, the keeper standing somewhere different, a defender out of place – hell, a balloon thrown on or a seagull dive bombing the game – literally almost anything could have meant that United would have ended up Champions instead.
But that would be kind of okay for the stat lovers, because United are big spenders too and City would still have been second, so that still ‘proves’ that big spenders succeed. However, what the fine margins show is that it’s all very well explaining events afterwards with the statistics, but what you can’t do is predict future events with them. They don’t tell us what will happen next season (even if we know in advance every penny that every club is going to spend), except in a fairly general way. They just tell us what is likely to happen. But to be honest you don’t need to be a stats expert for that. Where will Man Utd be at the end of next season? I predict top three. It didn’t take me long to work that out. In a sense of course I am using stats for that prediction, because I’m aware of how well Man Utd have done for the last 20 years, I know what players they’ve got and how good they are. I don’t really need to know the details of their wages spend relative to their rivals and I can’t predict their transfer spend very easily anyway.
What I don’t know is anything else that might happen. I expect their manager to be the same one they’ve had for years; I expect their level of injuries to be roughly equal to other clubs'; I expect they’ll get roughly the same amount of luck with decisions as they normally get; I don’t expect half their team to die of legionnaires disease or the FA to rule that they must play all matches away and start with a three-goal deficit as punishment for diving.
But even if I’m pretty sure that nothing outlandish is going to happen, I can still only predict roughly their league position one year in advance, not exactly. And that’s just the same if you use stats. The stats will only give a probability, and the exact probability is up for debate depending on how you interpret the stats, and the weight you give to different ones.
The 7AMKickoff post pointed out (at least I think it did, but I couldn’t see it when I just re-skimmed – it might be in the comments) that no club outside the top three spenders had won the league in over 20 years. Is that a surprise? Should we expect the league to be more open than that? These days all PL teams have very close to the same knowledge of players, tactics, medicine and other variables that could disrupt the chances, so spending money on buying and paying the best players is the biggest variable left. On that basis money becomes more likely to be the determining factor.
In my opinion there are two main wild cards: the manager and luck. (I’m defining ’luck’ as everything that can happen at random during a season: referees decisions, the weather, injuries, anything outside the game that affects key players (family bereavement, divorce, etc) – the list is long, but is equally likely to affect any team.) If the season was only ten matches then luck would play a much bigger part, but the reason that all the top leagues have 18 or 20 teams is that a 34 game season is the minimum needed to average luck out to a reasonable degree. The team top after ten matches often doesn’t win the league; the team top after 20 usually does, but not always – and of course the other positions change as well, usually much more than the lead. You have no chance of predicting the exact final table after 10 games, and very little chance after 20. Luck’s not eliminated entirely after 34 or 38 matches, but given the physical demands of the game you’d need a season three years long to play enough matches to remove luck (almost) entirely, and of course by that time players would come and go, develop or retire, and you’re not even measuring the same squads you started with. A year is long enough for one competition, and 34 or 38 games means the cream has enough time to rise to the top without making everything too predictable.
So we know how to deal with luck: we set the season to a particular length to cater for it.
The influence of managers is harder to determine or cater for. In a sense managers can be lumped in with other fairly random variables such as injuries or referees’ decisions, except that those things (conspiracy theories aside) are the same for everyone. And of course each team only has one manager, so there’s no chance of averaging effects out, as with injuries or refs’ decisions. So how much influence do managers have? Because if everything can be decided by spending, why bother with a manager at all? Is football management (and the high wages it commands at top level) a giant con trick? If the successful managers are just the ones who spend the most, then where is the skill? I honestly don’t think we’ll ever know.
There is an interesting case study from last season with Chelsea. They started off with AVB, did poorly, sacked him and went on to win two trophies. Unfortunately for several reasons this doesn’t really tell us much. We don’t know what would have happened if AVB had remained in charge; we don’t know whether the new manager would have had a better start than AVB; and in any case in the competition that best measures how good a team is – the Premier League – they didn’t get any better when they changed managers.
The President of the USA congratulates the last small club to win the league
What would be really interesting would be if the 20 PL managers were allocated randomly to teams on the first day of the season. This would have to be a surprise to them of course, otherwise they’d have no incentive to prepare properly. But if the FA introduced a rule that on average one year out of five all managers would be randomly given new clubs on day one, we’d have some really interesting data on whether managers do make a difference. Would Fergie still challenge for the league with Wigan’s resources? Probably not, but would he at least get them mid-table when their spend says relegation battlers? I suspect he would, but I’m not so sure about some of the others.
Another way to test managers would be to give every club exactly the same wages/transfer budget, but I can’t see this one being voted in. Really the top clubs should have little to fear from this because they already pay their managers the most, so they must believe that they’re the best. Surely they’d still be the best if everyone was spending the same? No doubt the clubs would use the argument that they must compete in Europe too, so we missed our chance to try this one in England back in the 1980s, when we were banned from European competition.
Anyway, leaving my dreams of showing up some managers aside, I believe that they do have some effect and are one of the main factors in why statistics can’t predict completely the finishing league positions. Both my intuition and the spending stats predict that Man U will finish top three next season, but I would give Fergie some credit if they win it.
I guess my point is that stats are interesting and can be useful, but I still kind if shrug my shoulders a bit at their use in sport. You can look back and see patterns and explain why things happened, but can you really explain a last minute Man City goal or a terrible goalkeeping performance at West Brom by looking at spending? Stats are for the big picture, not the detail.
Token picture of Mrs Sagna
But while we’re here, what of Arsenal? Unlike most of their rivals in the top six or seven clubs, they don’t spend a lot on transfers – in fact since 2008 there’s a £20m profit. Not only that but Wenger likes to follow a different wages model to other clubs, with a more equal split. You could argue that the Arsenal model works at least as well as the standard model because Arsenal finish third or fourth after spending the third or fourth most on wages, but less than those around them on transfers. On the other hand you could argue it doesn’t work because all our best players keep leaving. Also, the youngsters are paid more than necessary by Wenger’s egalitarian policy, so he could be more efficient and pay them less, while still ending up in the same league position.
Personally I don’t give up hope that Arsenal can win the league just because the spending stats are against us. I think the manager has got stuck in a rut and needs to be shaken out of it; perhaps Steve Bould is the man for that job.
The 7AMKickoff post that inspired this one sought to explain Arsenal’s lack of recent titles in monetary terms, while recognising that you could not of course just go out and blindly spend extra money and expect success. Nonetheless it concluded that a wisely spent £120m would do the trick. For me there’s still a problem here: both wages and transfer valuations are dependent on the whims of individuals. Wages to a lesser extent, admittedly, as there is a wide enough market for players, clubs and agents to make comparisons. But a few years ago Thierry Henry was transferred to Barcelona for a smaller sum than Darren Bent was transferred the same summer. Cesc Fabregas went to Barca for less than Andy Carroll went to Liverpool. No one – NO ONE – is ever going to claim Darren Bent is, was or ever will be fit to lick Thierry Henry’s boots never mind be better than him, and only the most myopic Scouser is going to think Carroll is even close to Fabregas. So where does that leave using transfers to predict anything else? Up a tree, I’d say. Similarly, some players rarely or never get transferred, or never so far – how much is Jack Wilshere worth? I hope we never find out, but the fact is until someone buys him we’re basically guessing. Robin van Persie’s value on a well known website was listed as £42m last time I looked, but we’ll be lucky if Arsenal receive half that for him. And this valuation was supposedly done by experts. Those kind of subjective judgements rather ruin the case of people who rely on stats too much.
So my conclusion is this: stats are great fun and can be hugely entertaining, but don’t try and prove anything to me in football with them.