Statistics in Football Manager What are they all about?

[tts0]

Hi, as a baseball fan I'm really into statistics and analytics, so I would like to know what all those stats in FM are about, because something just doesn't add up to me. E.g. "chances created" supposed to be a sum of assists and key passes(which are passes/crosses leading to scoring opportunity that weren't converted), yet in full details simulation FM also awards key passes to players whose passes weren't conrverted into any kind of shot. This implicates the fact that number of FM-like chances created must be lower that FM-like key passes. Yet stats created via quick sim shows many players with few key passes and assists but with lots of chances created.

What the heck is this all about?

[Able.Ryder]

Chances can be created without a prior key pass (run with ball, mistakes from opposition player, etc) and a key pass is not always followed by a shot or by a scoring chance.

So, to me, these stats are not linked.

[MBarbaric]

also, chance can be created even if there was no shot on goal. in essence, football is way more complicated than baseball regarding statistics. You can have a player that has certain role, i.e. holding midfielder making 8 tackles per match without ever making an assist. you'd think he was poor at creating chances, but if you give him another role he might have 0 tackles and 1 assist per match on average. 

not to mention that pure statistics (quantitative analysis) reveals very little in football (compared to baseball or basketball) while real analysis lies in qualitative analysis - actually watching the game and seeing why and how something happens. in general, quantitative analysisi in football is used to pick-up trends and confirm what is observed in qualitative analysis making the statistics practically useless on their own. this, i believe, is very different to baseball where you can compare, i.e. two pitchers 1 to 1 only based on stats. this is largely due to relative lack of diversity in position compared to football where each player is part of larger collective.

[tts0]

Guys, everything you say is probably truth, but the thing is:

a) IRL "chances created" literally mean assists + key passes, so why it might be something else in the game?

b) In my whole FM "adventure" I have never seen a player awarded with "chances created" after a run or something. Its either pass or cross thats converted into a shot, which is the way it suppose to be. BUT "key passes" are essentially not a component of "chances created" stats. Its almost all the other way round(In full details simulation) or completely unrelated(in quick sim). So, again, how am I suppose to interpret those stats? 

Funny thing is, I actually prefer it the way its calculated in the full sim rather than real life("key passes" could be interpret as a potential secondary assist), but when it comes to the quick sim it just doesnt make any sence at all.

[Able.Ryder]

1 hour ago, tts0 said:

Guys, everything you say is probably truth, but the thing is:

a) IRL "chances created" literally mean assists + key passes, so why it might be something else in the game?

b) In my whole FM "adventure" I have never seen a player awarded with "chances created" after a run or something. Its either pass or cross thats converted into a shot, which is the way it suppose to be. BUT "key passes" are essentially not a component of "chances created" stats. Its almost all the other way round(In full details simulation) or completely unrelated(in quick sim). So, again, how am I suppose to interpret those stats? 

Funny thing is, I actually prefer it the way its calculated in the full sim rather than real life("key passes" could be interpret as a potential secondary assist), but when it comes to the quick sim it just doesnt make any sence at all.

I'm not really sure why you are so confident about what "chances created" means IRL; at least IRL in the game of football.

Again, a chance is literally a chance. If a player receives a pass from his GK in his own half and then runs with the ball until he hits the opponent's post, would that mean a chance has been created by the GK?
And in the (almost) same case, would you award the keeper an assist after dropping the ball to his full back that goes and dribbles all the players until it reaches the opposition penalty area and scores a goal (in off the post)? I know it's technically an assist, but was it a key pass received from the GK? The same can be said about assist given to players that are fouled and a goal is scored from the resulted free/penalty kick.

Chances can appear in a lot of circumstances: key passes, Messi-esque runs, interceptions, missed interceptions, crosses, poor clearance by the opposition defenders and so on.

And please don't get my wrong (I'm just trying to understand what you want to understand ) but in football, as MBarbaric said, statistics in football can never be the subject of a movie (or at least not one for Brad to star in).
Although, given how assist are awarded now, I would not be surprised to see pre-assists stats (like in hockey, if I'm not mistaken).

[Snorks]

So many stats in football are contextual, which means they can;t really be statistical doesn't it?

So my number 9, receives the ball on his chest, turns and beats a defender creating a bit of space and shoots.

Now, he can do that in the center circle or on the edge of the box.

The stats would be recorded the same, but clearly, one is 'more valuable' to the team. So without some kind of weighting system built-in, football manager stats can only really be ever be a 'guide' to what your team are doing.

In baseball, the stats are less fluid and contextual. The pitcher can only pitch from the mound, the batter can only hit from one spot on the field. 

Football stats can only give an overall impression of trends, whereas baseball stats can be absolute. 

[tts0]

59 minutes ago, Able.Ryder said:

I'm not really sure why you are so confident about what "chances created" means IRL; at least IRL in the game of football.

Because its actually an "academical" term right now, that's used by every scouting/analytics websites/apps/databases. You can find the definition in the links below:

http://www.squawka.com/en/football-stats-definitions

https://www.optasports.com/news/optas-event-definitions/

And what Im trying to understand is how the game actually generate those stats, because its certainly not the way it suppose to be. Plus it differs depending on whether you quick sim games or sim in full details. And thats what confuses me the most, because I can very well understand what a "key pass" might be in full sim but utterly cant in quick sim; and obvioulsy I can't full sim every single league when many are unplayable by default. 

[tts0]

All right, boys. I've just run a simple test. TOP 5 leagues full sim over a half of the season. And then used this in-game analytics tool for random matches to try to define what those stats actually are. And OMFG its all over the place! Sometimes a key pass is also a chance created, sometimes it is not; sometimes its also a cross, sometimes not; sometimes its an assist, sometimes its not! And sometimes its either or neither. And the worst part is two almost identical situations can be represented in stats either way.

Im not gonna say that this makes those stats completely useless(yet obviously badly implemented), but this way it makes it so subjective that you can even treat those stats like a stats anymore. 

I think that this matter should be revisited by the devs at some point. 

[Able.Ryder]

1 hour ago, tts0 said:

Because its actually an "academical" term right now, that's used by every scouting/analytics websites/apps/databases. You can find the definition in the links below:

http://www.squawka.com/en/football-stats-definitions

https://www.optasports.com/news/optas-event-definitions/

And what Im trying to understand is how the game actually generate those stats, because its certainly not the way it suppose to be. Plus it differs depending on whether you quick sim games or sim in full details. And thats what confuses me the most, because I can very well understand what a "key pass" might be in full sim but utterly cant in quick sim; and obvioulsy I can't full sim every single league when many are unplayable by default. 

Ok, I see your point but those are Squawka's and Opta's definitions. Both sites state that in the very first paragraph.
Not very "academic".

Not really sure about what full and quick sim means.

 

9 minutes ago, tts0 said:

All right, boys. I've just run a simple test. TOP 5 leagues full sim over a half of the season. And then used this in-game analytics tool for random matches to try to define what those stats actually are. And OMFG its all over the place! Sometimes a key pass is also a chance created, sometimes it is not; sometimes its also a cross, sometimes not; sometimes its an assist, sometimes its not! And sometimes its either or neither. And the worst part is two almost identical situations can be represented in stats either way.

That's what we are trying to tell you, it depends on the context.
So nothing worthy of an OMFG  

I doubt, though, that in *identical* situations it can be two(or more) different interpretation in terms of stats.

 

12 minutes ago, tts0 said:

Im not gonna say that this makes those stats completely useless(yet obviously badly implemented), but this way it makes it so subjective that you can even treat those stats like a stats anymore. 

I think that this matter should be revisited by the devs at some point. 

Now I'm not really sure if you are (still) serious.

Anyway...
How can you be so sure (again) that they're badly implemented when in the same time you say you want to know what is FM's "academical" definition?
Maybe it is implemented so it reflects their definition  

I'm not saying that it shouldn't be looked into, though.

[Able.Ryder]

And also, have a look at how assist are defined:

https://en.wikipedia.org/wiki/Assist_(football)

Or how they may be defined in different competitions.

How would you choose the player with the most assists in the world just by looking at each league's stats & definitions?

[tts0]

39 minutes ago, Able.Ryder said:

Anyway...
How can you be so sure (again) that they're badly implemented when in the same time you say you want to know what is FM's "academical" definition?
Maybe it is implemented so it reflects their definition  

I'm not saying that it shouldn't be looked into, though.

Ok, after another hour or so of fooling around in an in-game analytics tool I might have cracked this for good! So, "chances created" are key passes(crosses included) that lead to a clear cut chance or a half chance defined by the match engine, while "key passes" can be also passes that dont create  a CCC or HC defined by the ME, but are still concidered dangerous(assist scored from no CC or HC chances included).  it turns out that FM-ish "key passes" are RL "chances created" for the most part, while FM-ish "chances created" are RL "big chances created". Now we can talk!  

And for the quick sim and full sim. 

Full sim means that game details for certain leauges are set to "all competitive matches", which simulates all those matches in the said leagues using match engine mechanics(you can re-watch them and see all stats), while quick sim is game details set to "none", which just creates some "random" basic game stats based on teams reputation(as far as Im concerned). Those games you cant re-watch etc. 

And what really concerns me right now is the fact that full sim stats are completely different from quck sim stats. E.g. there are plenty of players in quick sim with 80+ "chances created" and only 20-30 "key passes" which is IMPOSSIBLE to achieve via ME, because - as I said before - "chances created" stat is only a component of "key passes" stat which means "key passes" >="chances created"; never the other way round. 

 

34 minutes ago, Able.Ryder said:

And also, have a look at how assist are defined:

https://en.wikipedia.org/wiki/Assist_(football)

Or how they may be defined in different competitions.

How would you choose the player with the most assists in the world just by looking at each league's stats & definitions?

Personally, I define an assist the way its done in ice hockey, meaning whichever player passes the ball to a goalscores is awarded with the assist and whichever player passes it to the assist maker is awarded with secondary assist. Thats it. No nonsense like whether it was a simple pass form FB to Messi who then driblled through 10 oposition players or simply beautiful 30 yard through ball for a CCC. Of course, you can then create tons of other more advanced stats for assists type etc.

[warlock]

1 hour ago, tts0 said:

Im not gonna say that this makes those stats completely useless

That's good... just because you don't understand the stats doesn't make them useless. And anyway, trying to reduce football to a set of stats isn't particularly helpful. 'Key passes' - which seems to upset you so much - isn't something that can be equated to strikes in baseball, or completed passes in NFL. A key pass *might* be an assist, or it might be the pass that leads to an assist, or it might be the pass that opens up the field and leads to the pass that leads to the assist that leads to the goal. Or, if not a goal, then a shot. Or if not a shot, at least a chance. Maybe.

Some of us believe that trying to reduce football to a set of statistics is not just a pointless exercise, it's a pernicious exercise designed to benefit no-one except the gambling industry.

Read this, it might help:

https://www.nytimes.com/2014/07/08/science/soccer-a-beautiful-game-of-chance.html

[warlock]

5 minutes ago, tts0 said:

And what really concerns me right now is the fact that full sim stats are completely different from quck sim stats

Well, yes. Because if quick sim stats were the same as full sim stats they wouldn't be quick sim stats, would they? You can't 'quick sim' something to the same degree of detail as 'full sim' because then 'full sim' and 'quick sim' would be the same thing.

[tts0]

3 minutes ago, warlock said:

That's good... just because you don't understand the stats doesn't make them useless. And anyway, trying to reduce football to a set of stats isn't particularly helpful. 'Key passes' - which seems to upset you so much - isn't something that can be equated to strikes in baseball, or completed passes in NFL. A key pass *might* be an assist, or it might be the pass that leads to an assist, or it might be the pass that opens up the field and leads to the pass that leads to the assist that leads to the goal. Or, if not a goal, then a shot. Or if not a shot, at least a chance. Maybe.

Of course, football is way more complex that baseball or american football, which are basically 100% set pieces, and Im well aware of that. But It doesnt mean that stats in football arent helpful, they are. E.g. said "key passes' mesures creativity and vision of a player for the most part. Then you can create your own advanced stats like total amount of passes to key passes, which can measure a players' likeliness of creative play(flair?). But in order to create those kind of advanced stats you must know the exact definition of the basic stats components, dont you? And Im here to get a clarification on some of these; especially in the quick sim. An why is that? Because I play FM with completely hidden attributes.

[GerdMuller]

Still Football is very subjective on ist stats and the game even more doesnt really care to have a proper representation of "exact" stats. This sport just isnt meant to be to Project "Moneyball" ike stats were everything seems to be able to be numbered and computed down. I see the American Sports is very crazy for those Things now, but will it down to a nearly accurate science in Baseball , i see it has brought a lot of advancement to Hockey too, but reached ist Limits also rather fast. In American Football ist basically denied in usefullness, because the flags now dominate the game more than anything else really.  If you come from a Baseball, everything is calculabe via stats Approach you jsut wont find the reliable data for it in the sports media itself and even more in FM. TBH i dont hire any more data analysts anyore in FM at all and go purely by eye test and the attributes. I habe seen the pros and cons of stats in hockey very much , it was a hot topic. It can never be reduced down to numbers as much as Baseball and Football will achieve that even less, it is a too fast and situational game to crunch down the action into numbers that make a purely scientific valuation possible, So i know the crazyness for advanced stats , but yu cant create calculated values of Quality out of data that  in itself is so lacking.

I would say statistics in  their current state in Fm can show you some hot action zones, distribution and general values of quality of play, but mostly are more a part of general apperance of a representation of the world ist taken part in, then hard data you can calculate a winning strategy out of.

I am more romantic than mathematical though :-D , but i think especially for that Approach and advanced stats you need exactly this hard facts data ground you can base your calculations on and i dont see that.

 

And in a way ist very good the way it is , because the game is trying to hide the fact its only numbers and try to portrait a living world. I dont think it should be a priority to be so exact that you cna crunch it down to exact realiable numbers agaun, when even the sport itself doesnt achieve that.

[warlock]

7 minutes ago, tts0 said:

And Im here to get a clarification on some of these

That poses a question that has a two-part answer.

1. What is the real-world definition of some stats? As pointed out up-thread, it differs depending on who is doing the measuring.

2. Does SI use the same definitions in game? Often (usually?) not. See the endless yearly threads about crossing, and clear-cut chances.

Regardless of that source of confusion, I would suggest that you're hunting the snark... even if you had accurate insider information (which SI is reluctant to provide*) it's not going to get you very far. Take a very straightforward, easily understood, stat like possession. In the real world, it usually equates to control of the ball, which usually equates to number of passes made and completed by each side. But in FM it's much more like a chess clock. On a real-world football pitch, it's often the case that neither side has possession; in FM one or other team must have possession.

Of course, I accept that there are meaningful stats, both in FM and IRL, but I think they mostly come down to the individual player level. If I'm looking at two potential signings, their stats can provide a useful insight - passes completed, shots on target, interceptions made, clean sheets for goalkeepers (but even then allowing for a world-class defensive unit in front of them).

But at the team level, most of the stats available - both in game and IRL - are pretty much meaningless, as that NYTimes article makes clear. More possession does not equal better results; more shots does not equal more goals.

*I believe SI is reluctant to reveal the 'wizard behind the curtain' and for very good reasons. The more information they reveal, the easier it is to 'game' the game. In the real world football is almost always about hunches, best-guesses, experience and a host of intangibles. Which is why I despise the whole industry that is trying to reduce the game to a bunch of numbers - they only measure what can be measured, and ignored everything that makes the game what it is.

[GerdMuller]

Thats basically what it comes down too, that new advanced stats Industry says you are a dinsoaurs if you cant see that this is the future and everything just has to be measured correctly to be used in an advanced stat. Problem is i cant see a lot of really telling measurements thta could be used to paint a telling picture thats translates the huge complexizy of the game into a simple to use calculated end product that describes quality of play and makes it possible to base decisions on.

[sporadicsmiles]

5 hours ago, GerdMuller said:

Thats basically what it comes down too, that new advanced stats Industry says you are a dinsoaurs if you cant see that this is the future and everything just has to be measured correctly to be used in an advanced stat. Problem is i cant see a lot of really telling measurements thta could be used to paint a telling picture thats translates the huge complexizy of the game into a simple to use calculated end product that describes quality of play and makes it possible to base decisions on.

Potentially you can use some stats to analyse individual performances, and for player by player analysis this is probably useful in many ways. Tracking the position of a player, for instance, must be extremely useful for spotting his tendencies, what he does wrong and right, etc., without having to watch and catalogue everything manually.

The biggest problem for statistical analysis is that randomness is huge in football. Since football is in general a low scoring game, this means chance is often the deciding factor, especially when games are close. No amount of statistical analysis is going to be able to predict a weird bobble of the ball on the pitch, or a defender tripping over his feet, or a midfielder misplacing a pass. Events like this will always defy statistical analysis; you can say how much randomness there is, but not when it will happen.

[tts0]

Guys, I completely agree with the fact that football just isn't the game of stats, so to speak. If it had been the case, Americans would have loved it already. But - as some of you metioned - stats still can be helpful to some degree. E.g. If i take total number of GK's saves made and divide it by total number of clean sheets kept, I can see how many saves GK has to make in order to keep a clean sheet. That then can indicate how good or poor is his team's defensive play(fewer saves required = better defense). Of course, it doesn't provide me with a winning formula. In fact, it doesn't provide me with any sort of formula whatsoever. But it can help me define whether the next match-up against said team would likely be a struggle(good defense), so Im more willing to analyse tactics/weaker points etc. or it should be an easy-peasy, so i just go with it, saving a lot of time and work. 

Stats are also very helpfull for players abilities evaluation. You can go with any sort of stuff, like tackles made per interceptions(shows whether a player is more inteligent or more agressive in his defensive play), total number of fauls given per dribbling/run(shows whether a player is more likely to draw a set piece rather than provide a meaningful individual run with the ball), and so on and so on.

Thats why Im so into statistical aproach to the game> It gives you whole new perspective on things. E.g. In this test save Ive created yesterday there is Huddersfield sitting dead last in EPL at January 1st. If you go with traditional aproach and just look at the EPL table you are more likeli to say that "they just suck, there is no hope". Where in fact they DON'T. They sit middle of the table in almost every aspects of the game except finnishing, which means the addition of a good goalscorer(or two) could change things dramatically. On the other side of things is Crystal Palace, sitting 5th in the EPL, while being completely shambles in almost every aspects of the game. That means they are less likely to keep that pace for long period of time, and despite this seasons mild success, they might find themselves fighting against relegation as soon as next season. So, probably, if we were offered a DoF job at both Huddersfield and Crystal Palace, 99 out of a 100 would immediately go to CP and I would be the one that goes to Huddersfield, because they are one player away from being mid-table, while CP is one "come down to earth" from being prime relegation candidate.

[GerdMuller]

I think the argument is over time luck will be a non factor  which is a statistical basis , its like saying in gabling in the end the bank wins, but basing individual avanced stats to coe to quality conclusions of a player in football sees reaching. As i said in baseball were circumstances are a lot more predictable it i might make some sense.

[GerdMuller]

vor 7 Minuten schrieb tts0:

Guys, I completely agree with the fact that football just isn't the game of stats, so to speak. If it had been the case, Americans would have loved it already. But - as some of you metioned - stats still can be helpful to some degree. E.g. If i take total number of GK's saves made and divide it by total number of clean sheets kept, I can see how many saves GK has to make in order to keep a clean sheet. That then can indicate how good or poor is his team's defensive play(fewer saves required = better defense). Of course, it doesn't provide me with a winning formula. In fact, it doesn't provide me with any sort of formula whatsoever. But it can help me define whether the next match-up against said team would likely be a struggle(good defense), so Im more willing to analyse tactics/weaker points etc. or it should be an easy-peasy, so i just go with it, saving a lot of time and work. 

Stats are also very helpfull for players abilities evaluation. You can go with any sort of stuff, like tackles made per interceptions(shows whether a player is more inteligent or more agressive in his defensive play), total number of fauls given per dribbling/run(shows whether a player is more likely to draw a set piece rather than provide a meaningful individual run with the ball), and so on and so on.

Thats why Im so into statistical aproach to the game> It gives you whole new perspective on things. E.g. In this test save Ive created yesterday there is Huddersfield sitting dead last in EPL at January 1st. If you go with traditional aproach and just look at the EPL table you are more likeli to say that "they just suck, there is no hope". Where in fact they DON'T. They sit middle of the table in almost every aspects of the game except finnishing, which means the addition of a good goalscorer(or two) could change things dramatically. On the other side of things is Crystal Palace, sitting 5th in the EPL, while being completely shambles in almost every aspects of the game. That means they are less likely to keep that pace for long period of time, and despite this seasons mild success, they might find themselves fighting against relegation as soon as next season. So, probably, if we were offered a DoF job at both Huddersfield and Crystal Palace, 99 out of a 100 would immediately go to CP and I would be the one that goes to Huddersfield, because they are one player away from being mid-table, while CP is one "come down to earth" from being prime relegation candidate.

That makes some sense, i am re the guy to look at the squad and the attributes and say...yeah they have ****** strikers, we could iron that out.

 

On an FM level heres my experience in my Everton saves: I have a lot of gaes of 40-2 shots i always have to to hope to get a draw or better...so is there a lot of conclusion i can take from that ...my strikers are top ..i can just say it ust be y possession based tactics and hope for the best..not much to really find a eak spot here and inmprove it.

[MBarbaric]

1 ora fa, tts0 ha scritto:

If i take total number of GK's saves made and divide it by total number of clean sheets kept, I can see how many saves GK has to make in order to keep a clean sheet.

what? without considering what kind of shots he is facing? maybe the team was playing really well, sat deep and allowed only long shots resulting with a keeper saving 10 shots. In another, his team might attack high allowing the opposition no time on the ball except those 2 counter attacks when they conceded 2 goals. so what do you conclude from the numbers? he needs to save 6 shots to keep a clean sheet when we already saw he conceded twice from only two?

 

1 ora fa, tts0 ha scritto:

You can go with any sort of stuff, like tackles made per interceptions(shows whether a player is more inteligent or more agressive in his defensive play),

or maybe it shows that in one match the team was pressing more aggressively while in another they were instructed to contain the opposition keeping their positions and denying space?

 

what I am trying to say is that those numbers need context. If you don't actually see what, why and how things happen, numbers won't tell you a lot. Once you see things happening on the pitch, i.e. a goalkeeper mentioned above, you might actually see that he lacks decision making, speed, acceleration, positioning... when coming off his line on counter attacks. Youd then check his stats and see that he is prone to conceding on counter attacks and you'd have some proof and hard numbers to decide not to sign that player because your team is set up to play with high line and faces a lot of counter attacks.

[tts0]

MBarbaric, good point on that football stats needs a lot of context to be concidered relevant. Putting them in said context is what top analysts do all the time. You can even do that in FM to some degree, but its so time consuming due to the lack of proper collection and often presentation of data required.  Taking into acount the fact that FM is only a video game I would rather go with my own less in-depth 'advanced stats" for the most part and - sticking with the GK saves made example - Im well aware of that its out of context and as you mentioned it doesnt really tell the whole story, but in the long run(in the course of the season) it "evens out" a little bit and you can really see that teams who allows way more shots on goal before recording a clean sheet generally defend worse and the context doesnt really matter that much(especially in FM).

[Svenc]

Data in Football is becoming some more useful (even in itself), but FM is lacking far, far, FAR behind the curve (wasn't always the case). Real world examples of data Analysis / usage....

https://www.theringer.com/soccer/2018/6/27/17511596/2018-world-cup-germany-south-korea-mexico-sweden-elimination
http://11tegen11.net/2015/01/05/the-best-predictor-for-future-performance-is-expected-goals/
https://statsbomb.com/2018/11/a-new-way-to-measure-keepers-shot-stopping-post-shot-expected-goals/

 

etcetc

Additionally, what's in the game is oft a tad fishy and of subjective SI interpretion. Additionally, how the game uses the stats that are there… is quite woefully simplistic. F'r instance, if a Team had more shots come the end of the final whistle, it was the "deserved" winner (or unfortunate loser) according to the game's match reports. That's not how Football works -- and it's certainly not how the game's match engine works. It's not even how the game's artificial intelligence works, as the on the more defensive Ends of the tactical AI spectrum, the AI doesn't at all care About how many shots it ever has. Plus, as soon as a Forward bags a Goal/or two, he gets a great Rating. Whilst that mimics Mainstream Football media to an extent -- why should he get a great Rating for scoring Goals? Most of the time he's simply expected to do that, as it is his Job -- similar to that ball Winning midfielder Winning all those key tackles, never at all appearing in the danger Zone in front of the Opposition goal….

That said, as Football is such a low scoring Sports in particular, data in footie tends to be more useful over the longer term. Football is a Sports where even the most dominating Teams in the most unbalanced top leagues don't bag a fistful of Goals  really every other week after all...