Hacker News new | past | comments | ask | show | jobs | submit login
How a Man’s Bad Math Helped Ruin Decades of English Soccer (fivethirtyeight.com)
93 points by protomyth on Oct 30, 2016 | hide | past | favorite | 30 comments



If you're going to interpret stats, you need a lot of them. Reep obviously could not gather everything required to make a sensible conclusion by himself.

For one thing, Swindon Town has been a middling team since forever. No particular innovation in strategy would be happening there at the time when Reep was watching them. So he was probably not sampling nearly widely enough to notice anything.

If you look at Wilson's book called Inverting the Pyramid, there's a great history of football strategy.

There's no particular thing to do that will win you the game. There's always been a tug-of-war between pressing, passing, sitting back, and pressuring. Ukrainian coaches IIRC started to do high pressure, where you try to annoy the opposition into making mistakes high up the pitch. The riposte to that is that opponents who are comfortable on the ball will just... not lose the ball so much. And you'll get tired running after it, and you'll lose. This sort of balance thing comes and goes. Witness Barcelona passing it dozens and dozens of times to win a huge trophy haul a few years ago. And then they ran into a combined 7-0 beating by Bayern. Look at Mourinho, he somehow won the CL with Inter, parking the bus infamously against possibly the greatest Barcelona team of all time.

There's a big issue with using stats for adversarial games. It may be fine -though non trivial- to use stats to judge players by (tackles, red zone shots, etc), but you don't know how the opposition will react to your strategy. It's even more non trivial to work out how the other team will reshape itself given what you do, which depends on what they do, which depends...

It will be interesting, though. Stats is seeping into soccer steadily. We'll see better analysis and innovation, for sure.


I don't really like the article, the problem is that time of possession is a decent proxy for better team, or actually a proxy for team A is better at retaining the ball than team B is at getting the ball. So their counter claim is roughly, be the better team. Thanks, we already knew that one.

Actually that ties into nicely with a general problem of the presentation of statistics, anything involving numbers is usually presented as incredible reliable, while actually it is just an additional argument. A statistic may be trivial, like the one above, it may be bullshit or it may be conclusive, but the use of numbers by itself is not some kind of magic that reveals a hidden truth. Statistics is just a powerful method to construct arguments.


I posted in a sibling comment to this one about how this argument reminds me of one that happened in hockey.

Specifically I claimed that one of the big reasons that possession became more prized is because of Hockey Canada changing their developmental programs after missing gold at the Winter Olympics three times in a row. Specifically they started to do serious analysis on teams that were winning, both in the NHL and international level teams.

I was lucky enough to have parents well off enough that I got to play organized travel team hockey through my entire childhood (age 5 - 18). I'm 27 and I caught the beginning of the shift in thinking near the middle of my developmental "career". You could tell the difference immediately (and it wasn't incredibly kind to a physcally large forward like myself). Even more so, if you went and watched some of the high level younger teams though, the ones that got coaches from away and got taught the latest theory you could see the way the game was re-shaping itself even if at the top current level the stats were just showing that "good teams are performing well".

So that's why I think that dismissing theory and statistics as just saying that the other team is "better" in a roundabout way is missing the point. Quantifying which team is better and why allows changes to the developmental programs that produce players. If you look at the kids that came out of the developmental shift you can see the results of the statistical analysis, the 90s and early 00s was not kind to North American international hockey but as the kids that were taught with the new statistically informed program started moving up to the professional level you can see the level of play increasing and now it's Canada and the US that are the dominant teams on the world stage and all of that happened because the people running the programs started to look at the statistics of what actually worked and implementing them into the training and development of the next generation of athletes.


> the problem is that time of possession is a decent proxy for better team

Decent, but by no means perfect. In Atletico's La Liga winning season of 2013-2014, they enjoyed exactly 50% possession. The following year they actually had 45%, despite finishing 3rd. Leicester City won the league with 46% possession.

The problem is not just the presentation of statistics, it's the assumption that if you do the same things title winning teams do, you will enjoy the same success. It may well have been that hoofing down the field worked for some teams (it did for Leicester), but that is no guarantee of success if you don't happen to have a fast center forward and a holding midfielder who runs like crazy. On the other hand, Spain won Euro 2012 without a proper center forward (Torres won the golden boot with only 3 goals, but don't let that fool you, Spain's goals were widely spread out and Torres scored one in the final long after the game was won) by playing possession football.


Going into a discussion related to statistics with a couple of anecdotes - tut.

Reep only seems to have followed one team - albeit playing against many others. If he'd got other like minded people to do the same for their teams (ideally all teams in all leagues!) as well then that would help to control one or two of the huge number of variables.

Then as the article alludes, if he'd looked at the sequences that lead to a score, rather than working back from a score to the sequence that lead to it, something really useful might have emerged.

One thing is certain however: whatever the English national (soccer) team decides is a good idea - it isn't. It's a bit embarrassing to be honest.


So that is why the English game was 'known for long balls'.

I wonder if Arsenal ( which used to play possession football some time back ) was successful in England on account of this.

Also, teams like Barcelona which play the game with long periods of possession and are extremely successful.


Arsenal still play possession football. They have not won anything for a while, to the point "playing like Arsenal" (i.e. hogging the ball without actually winning) has become a joke. Their early-'00s success was due as much to an exceptional generation of talents (Vieira, Bergkamp, Pires, Henry...) as to any other factor.

The same for Barcelona - Pep Guardiola, known for being the most successful manager to insist on "possession first" football, was not particularly successful at Bayern (he inherited a team winning all domestic and international trophies in 2013, and only managed to win domestically for the following two years), and is somewhat struggling at Man City.

Famously, Leicester City won the Premier League last year playing the exact opposite of possession football. Someone like Jamie Vardy, capable of quick bursts of speed but with relatively poor skills on the ball, would be embarrassing to watch at Arsenal or Barcelona, but was absolutely lethal for the Foxes last year.

Possession football, like "moneyball" baseball, looks nice on paper but IMHO is a red herring. Technically gifted players will like to hold the ball, and gifted players naturally tend to win more because they are better, so people link this and that and say you have to hold the ball to win. What you really need to be successful is to play a style that suits your players. If you have Messi, Xavi and Iniesta, you play possession; if you have Vardy, Mahrez and Kante, you play counter-attack and long balls in space.


Good counter-attacking football is very effective against teams that play possession football. That includes Barcelona.

Teams that have beaten Barcelona often just let them have the ball, then wait for them to make a mistake. And teams that are set up for long periods of possession have a hard time quickly switching to defence.


> They have not won anything for a while

Back to back FA Cups in 2013-14 and 2014-15? I think the rest of your analysis is pretty valid however.


Eh, I tend to forget the FA Cup and the FL Cup - after all, major clubs will happily sack managers who can "just" win one of those in a season (like Van Gaal last year)...


I think that the real reason is the weather, until the advent of the premier league and the huge cash injection that came with it pitches deteriorated to mud patches by november. Passing balls around in the pouring rain on mud with zero drainage is a mugs game.

Two things; ok there appears to have been an obvious fallacy in the reasoning, but the data was gathered at the wrong time - the game was changing, TV was becoming fundamental to it, chopper Harris was disappearing and strange things like fitness were becoming important. Secondly and most importantly, the analysis revealed something that was quite widely believed already. Ok it was wrong, but it was believed, and immediately evidence of the success of the strategy was assembled. I see this all the time; confirm someones ideas with data and you will be loved. It's rare that you will be found out, and even when you are it might be 20 years before it happens.


Arsenal's successes at the turn of the decade were based around a more counter-attacking style than they've employed since the stadium move.

Looks like this season they have taken a leaf out of Leicester's book and gone for the pacy trident of Iwobi, Sanchez and Walcott to launch counters with.


That was from the 1950's when Charles Reep (an RAF officer) figured out that most goals were scored with three or less passes. His idea was to play the long balls from the back over the defense, and let the forwards run on to those balls and have a scoring opportunity.


Um, yes, that's exactly what this article is about.


In the previous few years Liverpool had been very successful by concentrating on possession to the point of working the ball up from the backs rather than hoofing the ball up with a big goalkeeper kick. They were a clear counter-example to this admonition that the way to win was hoof it and chase.

Sounds like "horses for courses" to me [the correct advice varies according to the characteristics of the situation].


Not really up on Football (soccer) but this reminds me of the puck possession vs dump and chase zone entry debate from hockey.

For a lot of reasons (up to and including global geo-politics) North American teams for a long, long time over-emphasized dump and chase The basic tactic is that you carry or pass the puck up until it will no longer cause an icing call then throw it into the other teams defensive corner and send your forwards in to generate a turn over.

Until reasonable statistical analysis this was thought of as the "right" way to play the game. Actually it was long AFTER the statistics showed it because politics was playing a role in what the statisticians wanted to find. They wanted to prove that the NA way of playing was "right" so they went and found stats to prove it. It wasn't until a combination of the NHL rising ascendant, the fall of the Soviet Union and I would argue the fact that Canada missed Gold in the '92, '94, and '98 Winter Olympics that forced Hockey Canada to look at their developmental programs and do some honest analysis.

Today though, much like in this article short quick passes and maintaining possession through zone entry is preferred and the statistics back that up. I'd imagine that if what you say Liverpool has been doing it true it's for similar reasons, maintaining possession allows you to generate more shots, shots are important if you want to score, ergo keeping possession is important for scoring.


Err, ahh, English soccer... that's American football, right?


However the mighty Wimbledon, (now AFC Wimbledon) play a passing game of football now, time moves on.


On the other hand, Leicester, last year, played to get the ball towards the opponent's goal ASAP, not for possession (http://www.bodhiresourcing.com/rise-data-analytics-football-..., http://www.skysports.com/football/news/30385/10267972/leices...)

That didn't end up badly for them.


If you think Leicester played anything like the Sheffield united, Wimbledon, PNE et al of old you need to check your history!

The main issue I have with Reep's analysis is he seems to ignore the defensive benefits of possession.


I not think that, but they did show that "Not more than three passes" doesn't imply kick-and-rush. My comment was triggered by your statement "time moves on".

So, the math still may be right.

For an extreme example, scoring without possession is as good as impossible. That could lead to a rule "keep the ball"

However, that doesn't imply one should optimize for possession alone, as the way to maximize possession is to play backwards when even pressed a tiny bit, scoring own goals if needed to keep possession.

Also, I don't see huge defensive benefits of possession if you want to win. In the end, it is about two teams both getting equal number of possessions (+/- 1), so to score (which is needed to win), teams will want to maximize their chance of scoring on possession and to minimize that of their opponent.

Only if your chance of scoring is lower than that of your opponent or if you are ahead 'enough' (where 'enough' depends on time to play and speed of you opponent), you will want to minimize the number of opportunities both teams get, even if it means your probability of scoring on possession goes down.


The Wimbledon that won the FA Cup moved to Milton Keynes. Their fans stayed in South London and turned into fans of AFCW.

Very unusual in European soccer for a club to move towns. US sports franchises do it more often.


Leicester City adopted this guy's style and won the league though :D


Was it a case of confusing correlation & causation?


Soft of, but it was more precisely the Bayesian fallacy.


I'm not sure what the Bayesian fallacy is (I find several papers with the term in the title, but couldn't find a definition) but it seems like this could be the base-rate fallacy (most goals come after short sequences of possession, neglecting the base rate: most sequences of possession are short).


P(A|B) does not equal P(B|A). The Bayesian fallacy is inappropriately conflating the two. In this case, The commentator confused P(scoring | long possession) and P(long possession | scoring). It is very similar to the base rate fallacy to the extent that if someone commits the Bayesian fallacy they almost always also commit the base rate fallacy.


Damn it. I misread read the title. I was just having am argument about B.O. I guess that's why I read "breath" instead of "math." I read the article twice looking for how his breath changed things. So disappointing. Good article, though.


I did the same, odd.


Niggle: The title should be "How _One_ Man's …", not "How _a_ Man's …". This reflects the actual title of the article, and avoids an ambiguous reading (was this bad math attributed to _men_, or a singular person?)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: