A different perspective

[Barbara 15,149]

I'm with Rapha here... I don't see how such stats could be really useful to a sport like football. But why not try?

Just for background information: many other types of stats had been used in the past (by media, FIFA, confederations, etc) and they were dropped because they simply didn't reflect what happened in the game or because the winning tactic changed and some stats became outdated.

I'm no expert in baseball so I can't say why it worked or not there, but possession in football has been overrated ever since Spain and Barcelona aced it. The thing is it isn't possession (alone) that makes both great, it's the players they have. I've seen teams - and even Spain more recently against Italy in the 90 minutes (extra time there was a change pro-Spain) - where they keep high possession and if you ask who had better, chances, deserved to win and in some cases even won, they will say the other team... and again with high possession % and completed passes. Spain has a ridiculously low % of uncompleted passes, including against Italy. That never reflected - or rarely reflected - on them dominating the match or being closer to score than Italy. First half especially the match had always been in Italy's hands. Their terrible strikers may be to blame.

Possession is all hot and happening (thanks, @Ankit) now, but it's only a matter of time until teams it figure out - like Italy did - on how to make it ineffective.

Football is such a dynamic sport. possession is a thing again almost 40 years after Netherlands made it a trend in 1974 WC although they had a different kind of possession (a lot more pressure defending on their attacking field, more objective possession, more fast-paced, but still the annoying excessive passing instead of shooting). Then we go to defensive football like Brazil and Italy in 94 or compact and involving like France 98 or fast paced or dependable on crosses on the box, or slow-paced but controlled, etc, etc, etc.

It's like working in cycles, today is possession, and if you ask me the next will be fast-paced/objective again. Next year in Brazil I still think it'll be all about possession, but the next Euro I'm positive possession won't be talked as much as it is now because it won't be as important then. It'll be a more vertical football compared to the horizontal tactic Spain exported to the world since their first Euro.

[Peace. 4,352]

While I find very interesting the stats that you brought up, TorontoChelsea, in your first post ; to my personal belief, (trying to find) a formula to analyse the impact of a player is kinda "irrelevant" and doesn't have its place in football.

First off, I gotta admit that I am not a big fan of statistics. Numbers don't lie — but they say what you want them to say. Just ask your government, they know how to make stats speak. Into my opinion, stats should not be used to prove a point, but to back it up, to illustrate it. You will never find the right formula — unless you make up a calcul as complex as the ones used in meteorology (and in physic in general). Statistics are depraved from all sort of judgment of value ; Mathematics cannot grasp the notion of good and evil.

I'll use a few examples to illustrate what I am trying to say :

• Some times ago, some people were using a (misleading) statistic to argument that, yes, contrary to expectation, Mikel was actually doing a lot of passes forward. I don't recall exactly the figures, but it was 60% of forward passes against 40% backward (or something along these lines). In truth, this stat means nothing (at least in the way it was used). Indeed, the computer which makes the stats / OPTA draws a circle around the player, and from 0° to 180° the pass is considered as pass forward ; from 180° to 360° the pass is considered as backward. Which means that despite having 60% of forward passes and 40% of backward passes, Mikel could as well have 100% of sideways passes.
• To continue upon what Manpe said : which player has the best positive impact and the less negative impact ? The players which makes 15 long balls but which fails to complete 10 of them ? Or the player who makes 10 long balls and fails to complete 5 of them ? Like this, it's impossible to say — and mathematics won't be able to give you an answer since they take these facts out of their context. Is an incomplete pass a negative impact ? Not necessarily. Because, you can fail to complete your pass, but in some case it can force the opponent's defensive line to go deeper in their half of the pitch ; it can allow you to drive away the pressure of your own half ; it can allow you to put pressure on their back-four ; et cetera... A contrario, a long pass which is completed can have a negative impact. For example, if the player you pass the ball to is in a bad position, then even if your pass is completed it can lead to a loss of the possession and/or kill an offesive. It also depends upon the style of football you play. If you play a possession football like Barcelona, or a counter-attacking team, 5 long passes failed out of 10 won't implies the sames meanings.
• I'll quote Choulo "For example, losing possession or missed pass in defense should have, let's say 5 times, the weight of losing the ball in the final third." It's not as simple. Some missed passes in the final third can be more deadly for you than some missed passes in defense. A missed pass in defense can lead to nothing because the ball went into a position where the other team could not make anything from it ; it's not a dangerous loss. A missed pass within the final third can lead to a deadly counter-attack because it was on a corner situation, and you were on of the last defenders. A missed pass in defense can be positive because it's a clearance and frees you from pressure (for a will, at least). A missed pass in the final third can kill your last chance to win the game. Et cetera...

I won't go any further in my examples and won't detail each and every possible situation. I presume (and hope) that the few examples that I gave are enough to grasp my opinion.

Stats are a good thing, but when used with parsimony. Stats have their limits ; and the more you add statistics to others, the more this limite will get tighter — and then you will debouch to absolute nonsenses. I take the rating from whoscored.com as the perfect illustration. In their overall rating, the ladyboy has a higer grade than Oscar, Petr Cech, Azpilicueta and Ashley Cole. The vast majority of Chelsea fan (in this forum..) are agreed to say that Ivanovic ; but whoscpred.com rate him more than Kompany and Nastasic. That's wherein lies the danger of adding one stat to another one.

That because of these reasons that to my mind, judicious to use formula to analyse the impact of a player. If you want to analyse the effectiveness and impact of a player in regard of the goalscoring departmen (for example)t, then you'd better, in the first time, to analyse what your own judgment and perception say, and then analyse his stats (his goalscoring and assists statistics, his goal to shot ratio, etc...), one after the other — separately —, and then comparing them.

We cannot rationalize football and its analysis, because it isn't something rational. There are too many special cases ; too many truths and ways to reach them.

[Muzchap 8,966]

@Peace great stuff

But!!! And I cannot believe I'm saying this - I do think there could be merit in this type of analytical approach...

However - I believe that the weighting needs to be applied

A) In possession

not in possession

For (A) the weighting or blended contribution needs to be judged on how decisive the team are and how players contribute to the general attacking play - so many times teams make it to the attacking third and then pass back into defence - now is that a positive or negative attacking manoeuvre? To me it is negative! It is a triumph for the defending team..

The problem I have and especially with the England team - is that we get to the edge of the box and then end up in the 'Dither Zone' - the Dither Zone is dangerous, it's a knife edge between attacking reward (goal or chance of goal) vs a misplaced pass (turnover) leading to a dangerous counter attack...

I think you would also need to add weighting around the formation used too - some formations are more attacking/defensive than others...

All easily do-able though - its just statistical modelling with weighted input parameters - everything can be calculated on Excel A DB would be more fun though - and I'd also include a self learning (adaptive) probability engine in there too see if it can within a tolerance start accurately predicting results based upon players, weather, previous form etc if we pull that off - either richer than our imagination (fleecing the bookies) or employed by every club globally

I never thought I'd ever be interested in statistics again - but I'm quite liking this thread

[TorontoChelsea 4,064]

We cannot rationalize football and its analysis, because it isn't something rational. There are too many special cases ; too many truths and ways to reach them.

The issue with any statistical analysis is sample size. For one particular game, 60% passing might actually be good if they do lead to goals. However, over the course of a season, these sorts of things even out. When you're talking about a number of passes in the range of 1,500, then the special cases become irrelevant because the sample size is big enough to render them statistical anomalies. If you're 60% passing over the year (as say, a midfielder) you're doing a horrible job. As I've said, judgement and watching the game is crucial and will always be crucial, but statistics are being used by everyone for a reason. Yes, they can be misconstrued and some of them are difficult to understand, but this "you can't understand football through statistics" is nonsense because it's already being done. Yes, you have to balance the statistics with what you know already, where a player fits into a team, what his role is, and so on, but the statistics are already here and they are not going to go away, they are going to become even more widely used. We're not talking about the notion of good and evil or anything abstract and unanswerable, we're talking about something finite and measurable. In this case, how many times a game does a player concede possession? Possession is not everything but conceding possession is undoubtedly a negative just as winning back possession is a positive. You can measure someone's success in long balls, someone's crossing success. Someone's total distance run. Someone's peak speed, etc...Everything that can be measure is being measured and is being analyzed.

Here is an interesting article on statistical analysis in football. http://sportsillustrated.cnn.com/2012/writers/jen_chang/04/29/performance.analytics/index.html Of course, the analysis they use is already so far ahead of what we have access to. I'll also quote from a conference on football analytics done last year (rest of article is bleh). (This is not a defence of my own statistic which is more fun than anything, but of statistics in general).

Importance of situational context

A majority of analytics done today seems to be of the ‘aggregate’ nature like pass completion, statistics around the shots vs goals etc., However this information has limits on its utility without the context. For example, pass completion % under pressure from two defenders is a more telling stat than the overall pass completion. StatDNA seems to be one of the few companies working in this area. Companies like Opta seem to have more real-time data which suits the media much better than companies’s who cannot collect and analyze data in real-time.

Role of traditional talent scouts

The traditional talent scouts still play a big role. Analytics is not a substitute to the old-school scouting methods. But analytics can play a big role in reinforcing the work of the talent scouts. As Steven Houston pointed out, if a team is looking to sign a right back and the scouts have a list of top-10 right backs, analytics can be used to justify or dispute the presence of a player in that list. Analytics adds an extra dimension to the data and helps teams make more informed decisions.

Challenges

The biggest challenge is getting good data to work with. Soccer analytics is still in its early stages of growth. A lot of work is going on in the area but what we are seeing is a very thin slice of the whole pie. Most of the work is happening under wraps. Teams are trying to get an advantage over the rest with their custom in-house data collecton and analysis, for obvious reasons. Almost all companies in this field are privately owned and all of them hold their data close to their heart.

A single repository of data even at a league-by-league available publicly could do wonders to drive research and get more enthusiasts on-board. It is not to say that it is impossible to get data today but it isnt straightforward and is definitely an entry barrier.

[Zinedine Iniesco 172]

There are so many dimensions to statistics in football that using them without other evidence with it is not recommendable imo.

But statistics are relevant, and they can help clear up or emphasize some things.

Now this statistic you showed in the OP, what is the main purpose you are trying to get out of that? Contribution? Because even then, if you are to compare defenders like Azpilicueta or Cole, surely the amount of turnover or passes wont tell you enough about their positioning, mistakes and losses when it comes to duels, keep integrity of their defensive lines...

Same with midfielders, what if you compare a player like Khedira to Martinez, if you go to whoscored.com and compare their defensive and passing numbers there isnt huge difference but when you see them on the pitch there is differences of tremendous proportions when it comes to runs, covering, marking, etc.

So i dont think you can create a formula that is too exact, we have to search further.

Im writing this quickly and being a little bleak with the explanation of my point perhaps but the point is that right now im not so sure there is a universal formula one can use to measure "contribution" properly, and i think stats are a good thing but should be used with other evidence in football.

There are way too many dimensions to cover, if anything the statistics shouldnt be so broad on all players.

We should probably define the rules to see what the importance and contributions of the players are, differentl and separately depending on what their roles are to get more exact tellings.

And maybe that can help us understand with numbers how efficiency of a scheme works.

I think i saw you mention something about ronaldo's shots seeming less impressive and also taking away from other's efficiency and you are correct.

Numbers can help emphasize things. But we need to create very good dimensions to use those numbers in, to get more value out of them.