Is possession and shots on target statistics too irrelevant?

[Matshit]

So, I was looking through the team statistics in the Serie A (the league I am playing in), and I noticed how the shots on target and possession stats clearly lacked any sort of consistency.

To take the possession stat. The highest is Trapani. They are 18. in the league. The next teams are 6, 4, 20, 15, 1, 13, 16. Which seems about as random as possible. In real life (according to whoscored) the same stat is 3, 2, 9, 7, 1, 6, 14. Still somewhat random, but much more weighted towards the big teams.

If you look at shots on target, I am leading that myself (2. place). The next teams are 18, 1, 6, 12, 10 and 20.
If you again compare to whoscored the rank is 3, 2, 1, 9, 4, 7. 6. Pretty similar to the actual league table.

Looking at this, there are obviously two teams that really look out of place. Trapani in 18. and Atalanta in 20.
Now, this could of course be due to tactical choices, but Trapanis coach like last season has playing style: long and playing mentality: cautious. Atalantas coach has both as mixed. Neither should indicate that the coach is the new Guardiola or the new Zeman.

Now, I fully accept that possession or shots per game isn't a 100% guarantee that you play good, but I would imagine it should be more than this.
Is this a known issue? Is it just in my game? How does it look in your games?

[Svenc]

Whoscored lists average Shots on Target per game: own, against -- the latter of which would be a neat stat in the game, actually, never been in there -- analysis uses both to calculate a Total Shot Rate (which is 0.5 if that is even on both ends), as that paints a more fully picture. For instance, last Bundesliga season relegated Stuttgart averaged as many shots on target as some of the Big Boys (5.5 on average, better than Leverkusen or Mönchengladbach qualifying for the CL), but equally conceded the 2nd worst record against, which was indicative of their leakiness. FM lists the shot on target percentages (own), that is the accuracy of shots, and the total amount. You can be pretty accurate and fight against relegation (very few shots, but decent accuracy) and vice versa. Currently don't see anything completely daft in those total amounts myself (which would be linked to what Whoscored lists), certainly not relegation battlers being teams with having the most of them. If they would have as many, they wouldn't be so low in the actually tables, unless... see Stuttgart perhaps.

Possession stats is comparably irrelevant. In the EPL you've had Swansea or Southampton for years who spammed their high percentages by playing keep-ball in deep areas of the pitch where it won't harm anyone and lead top class punditry comparing Britton to Xavi despite Barcelona never getting interesed, puzzingly. Samewise, possession for the big boys improves by teams dropping off to keep it tight on many matches all by itself. And then Leicester won it all last season with the worst. It's partly also a stat that is difficult to compare, as stats outlets take the number of passes to calculate the possession ratios, whereas FM works like a chess clock method, counts the time a team is on the ball, rather.

AI managers decisions can be a bit over the place, and some roles already can have a big impact (Target Man up top, off drops the stat), and the real Guardiola ever isolating a lone forward by making him run far ahead of everyone during transitioning, effectively having a player less, which impacts quite some, would never happen in real football. AI managers are a bit of a pet peeve of mine, actually. I don't think it's a secret that a creative human manager was far better here than an AI in general, though speaking in terms of possession,  what bumps it oftenly is the polar opposite of what would stretch defenses, but that's another matter. Also in this season's the formation prefered can have quite a handy impact already, as central midfielders are in tendency isolated, and it is the central spaces that control the pitch. But I don't want to drag this there another time, you can read some about it here. You'll notice that on average, there isn't that hugely much of a difference too. We're talking about a drop from 52% mostly in the top thirds to 48% in the bottom third, with but a few teams being below and above.

As an ado: Winning possession and having more shots is not equal to "playing well" or getting the tactics "spot on", nor should the game aim to be modeled like that, but that should be most obvious, hopefully. For what it's worth on the possession front, I think it's been comparably easy to keep it highish even with lesser teams myself all along, which imo is partly connected to pressing/defending inherent to the ME, partly the modeling of technically traits, but also down to AI managers being well, AI managers. Oh, and the next time on FM's current level of overall very simplish stats you see somebody claiming "utter domination", give him a slap. (And yes, I agree it's pretty damn simplistic).
 

 

[Matshit]

27 minutes ago, Svenc said:

As an ado: Winning possession and having more shots is not equal to "playing well" or getting the tactics "spot on", nor should the game aim to be modeled like that, but that should be most obvious, hopefully.

Absolutely, but if you think that is my issue, you have misunderstood completely. There are tons of things that should have some sort of connection to the league table (most money spent, highest wages, highest average CA, highest reputation), but shouldn't be completely co-related. My issue isn't that there are teams who have weird statistics compared to results, it is that there seem to be no connection at all. Looking at the possession or shots on target statistics IRL, they are clearly somewhat/very co-related to doing well. In the game there seems to be no co-relation at all.

 

In England, which is ran on low detail, there seems to be a much bigger co-relation. The possession table is 2, 1, 4, 3, 7 with Liverpool who are in 8. place and bottom of the possession table as the big outlier. For shots on target it is 2, 1, 4, 7, 3. This time with Liverpool in 6. place. Newcastle is the biggest outlier here with an 11. place, but a 5. place on the real table. That seems more like real life. A clear connection between having possession and getting shots on target with getting points, but not a complete 100% connection.

[Svenc]

10 hours ago, Matshit said:

There are tons of things that should have some sort of connection to the league table (most money spent, highest wages, highest average CA, highest reputation), but shouldn't be completely co-related. My issue isn't that there are teams who have weird statistics compared to results, it is that there seem to be no connection at all. Looking at the possession or shots on target statistics IRL, they are clearly somewhat/very co-related to doing well. In the game there seems to be no co-relation at all.

As for possession, the game has never been very good with replicating what possession based teams in real football are doing. As a user, it is easy to ape their "stats", at least --- in my opinion the ball has always been too readily played out of the back, and you'd never get the cbs as involved as they are in such teams in real football, where some of them make the most passes in the game regularly -- the ball rarely moves back to them nor does it move much between them. The main coder is a good guy, but he's very obviously a Brit! As for the AI mangers..., that is another can of tunafish. However, there is limited correlation between success and possession in itself, as it imo should be. This iteration, I've argued there to be an issue (thread linked to). Whilst I haven't yet seen the bottom team having the highest possession stat, out of interest, what formation are they regularly using? If you are curious you can check that thread. On low details the game also uses a quick sim. Unfortunately that's never been updated in years. You'll find stats that are yards off pace of development, such as much fewer passes made per game, shots taken, distances run, anything. Even the goals scored by the top forwards are hugely lower. Needs an upgrade!

As for the shots, as said, there is a distinction to make. Done a fresh sim for a couple months holidaying (assistant needs to be fired :D). The ranking FM lists is Shot accuarcy, i.e. what percentage of shots taken  are being on target. It is this one, where Cagliari currently reign supreme in real Serie A whilst Roma and Inter are closer to the bottom of the table. Teams can have an above accuracy in terms of shots and be relegation fodder, and vice versa. There is little connection. In fact, in parts the top accuracy of Cagliari may be caused that teams expect to beat them and attack, which means they may have more spaces on their few attacks. In parts it may be caused that they don't get to take that many shots full-stop. What Whoscored lists is the average number of shots taken on target, which would be connected to the overall amount of shots on target FM lists next to this.

http://www.squawka.com/football-team-rankings#shot-accuracy#team-stats#italian-serie-a|season-2016/2017#0#90#any#any#season#1#all-matches#shots#desc#total



Marked in red is Shot accuracy (squawka). Marked in yellow is what relates to the Whoscored stat, which is the amount of shots on target. On FM it is season totals to up to that point, whilst on Whoscored it is averages. You'd need to calculate the averages yourself, and I'd love their to be the opposite of this, which would be SOT AGAINST rather than for too. Do you personally see relegation battlers being at the top here? At mine they currently have up to half as few as the top teams (i.e. Middlesbrough)

[Matshit]

On 1/9/2017 at 05:59, Svenc said:

The ranking FM lists is Shot accuarcy

No, one of the ranking FM lists are accuracy. There is a different ranking for total shots on target.
And listing total shots and listing average shots per game would give the same ranking as long as the teams have played the same amount of matches. Your stats look way better than they do in my game, although still a bit flakey, but it is a bit hard to tell when it is sorted by shot%

[sporadicsmiles]

1. Are shots on target irrelevant?

Here it has to be a no, with a proviso. Clearly, the most shots you have on target the better your chance for scoring a goal; by definition a shot off target is a wasted shot with no chance to score. So the SOT percentage tells you about what percentage of your shots are leading to scoring chances (which I will define as anything on target regardless of quality). If this is low, then you need to look to make it better, and if you make it better you should expect to score more and lose less. 

The proviso is that the total number of shots you have is also important. If you had 200 shots with 40% accuracy (80 SOT) this would be much better than 100 shots with 45% accuracy (45 shots). You would be able predict with quite a bit of confidence that the first team is higher in the league than the second team. So the best teams will be those who maximise both their number of shots and their SOT percentage, which is more or less what you see in the post of Svenc. 

 

2. Does possession matter?

Not really. It is what you do with the ball. I usually end up with the highest possession stats in the league, because that is how I happen to play. In particular, I tend to patiently probe when I have a league to try to draw the other team out so I can exploit the space they leave. But I have won games where I had 40% possession if that is what is required. Possession is trickier than SOT to define, since you can have plenty of possession but do nothing with it (that is my default defensive tactic), just as you can little of it but be extremely decisively when attacking and still win.

[eriktous]

2 hours ago, sporadicsmiles said:

Clearly, the most shots you have on target the better your chance for scoring a goal

Clearly not. Not all shots are created equal. Many of what are counted as shots on goal are so weak powered that even if it had been placed four yards to the left or four yards to the right, the goal keeper would always have enough time to stroll to the right position and pick up the ball. The probability of those leading to a goal would be based on the chances of the goalie going into cardiac arrest (and not dropping his body in the path of the ball) - very close to zero. On the other hand, shots off target can often only go inches wide or over, far out of reach of the goalkeeper, who wouldn't have stood a chance of saving it if it would have been a yard more to the other side of the post or crossbar.

Now of course, this can go both ways, so all I'm saying is that the stats say nothing. They can show I had 3 shots, all of target, and the opponent 21, of which 9 on target, but having watched all those shots, I could tell I was the most likely to have scored, or it can be completely the other way around, and anything in between.

[Matshit]

3 hours ago, eriktous said:

so all I'm saying is that the stats say nothing

That is just completely wrong. It is like saying a person with 5 dice isn't a favorite to get a big number over one with 2. Yeah, there will be times when the one with 2 dice win, but if you run 100 tests, the one with 5 dice is expected to win a huge percentage of the time.

Same with shots. If you look at a match statistic and see that one team has 15 shots and the other 5, you should be able to predict with a pretty high certainty which team won the match. The team with the most possession IS most likely to win, and the team with the most shots IS most likely to win. That is just statistical facts. Over a big sample size, this should also be reflected in the game.

[Svenc]

Thanks for pointing the SOT ranking out, Matshit. For some reason completely missed it! Must be getting old

 

9 hours ago, Matshit said:

 Over a big sample size, this should also be reflected in the game.

There is a piece of the puzzle missing. Teams don't only create SOT, they also concede them. Unless I've missed it, FM doesn't have such a stat. When Atletico championed La Liga, they had nowhere near the SOT as Barca or Real, however they comfortably conceded the fewest SOT in all the top leagues. Stuttgart in the Bundesliga were relegated last season despite creating more than CL qualifiers. However, they also conceded only marginally less (0.1 SOT/g) than hopeless Hannover 96. Currently Mainz -- they have almost as many as Leipzig, hower concede as much as Darmstadt, the dead last in the table. As a consequence, such teams are nowhere near the top, they may score quite a few, but concede a shitton. If you personally have a team sitting at the top in the SOT tables and being at the struggling, it would be interesting to see their scoring record (for / against). They would be expected to score quite a few, but, leak like hell. Otherwise, they couldn't be bottom of the tables! As for individual matches, that is different, not only as in such a low scoring sports where even "decent" chances are bottled regularly results go against the run of play all the time. But because if long-term things may even out some, in a single match those stats don't tell who had the biggest chances. Of those 15 shots, all can be rather poor, whilst the 5 of the opponents the biggest of the match.

 

Quote

Of the 181 games won in the Premier League before last weekend, the team who had the most possession only won 103 – 57% in total. The team who had more shots on target than their opponents won 128 matches – 71% of the total.

https://www.theguardian.com/football/blog/2013/feb/24/football-numbers-game-gary-neville

The thread linked to at the bottom of my first post (How accurate are the chances) as per the game in particular goes into that. The shots can hugely pile up, even on target, but few can be actually in any space, and as such, the conversion takes a fat nose dive. The only statistical clue in-game is the set piece stats. Top teams in real football average no more than 7 corners a match, if somebody ever approaches the 15ish kind, and that's doable, oftenly not a good sign. Weird AI decisions can also contribute to all those stats  -- like Barcelona managed by an AI in some matches letting the team sit all back. Sounds odd, but can happen. It would naturally also impact the seasonal possession average.

 

[sporadicsmiles]

18 hours ago, eriktous said:

Clearly not.

Except there is a positive correlation in the data in this thread showing that teams who have more shots are higher in the league. Now we can argue on why that is the case (this is the essence of statistical based discussion) to our hearts content, but there is a correlation. For example the correlation could be the other way around; better teams have more shots because they are better. I would argue that more shots = more goals = more points though.

The point you make about a single game is statistically insignificant. Firstly, a sample size of one game is not big enough and football is too random a game to draw any conclusions from. I mean clearly it is possible to have 50 shots, score none and lose to a single counter attack (how many threads have we seen like that on here?). This is why you cannot really do moneyball things in football, it is way too random (much more so than baseball) so stats really do not tell you everything. You definitely should not ignore what is happening on the scale of a single game.

By the same token, you should not ignore what is happening on a more statistical scale either. So if there is a good correlation between shots on target and finishing higher in the league, it is definitely good to try to maximize the number of shots you have on target. Like I said earlier, the simplest way to look at it is that every shot off target is by definition a wasted shot. Every shot on target has some chance to be a goal (as you point out, this may be a small chance, but on average it will be higher than zero).

You inspire me to do some statistical analysis of this kind of thing from FM actually, I might sim a season and drag out some numbers.

[Svenc]

1 hour ago, sporadicsmiles said:

By the same token, you should not ignore what is happening on a more statistical scale either. So if there is a good correlation between shots on target and finishing higher in the league, it is definitely good to try to maximize the number of shots you have on target. Like I said earlier, the simplest way to look at it is that every shot off target is by definition a wasted shot. Every shot on target has some chance to be a goal (as you point out, this may be a small chance, but on average it will be higher than zero).

You inspire me to do some statistical analysis of this kind of thing from FM actually, I might sim a season and drag out some numbers.

There's actually people making a living off this, i.e. Matthew Benham. Although, yes it is a challenge and on FM's level of stats not advisably.

If found that on FM the correlation between SOT and scoring, even in individually matches, are pretty close to football, i.e. Whoscored. Semi-regularly you can guess the score by looking at those numbers and have in mind that SOTs are converted on average at 1in3 to 4 rates -- roughly. At the least certainly is a higher correlation between the SOT and the scoreline than the dreaded clear-cuts and the scoreline, which was to be expected just looking at them. Naturally there are freak incidences, as not every shot is equal at all. The one thing that in the game can skew this hugely is that it may or may not be too easy to amass "crap shots" OT off set pieces. I call these crap shots as by definition they are in very little space, and are converted accordingly -- comparably rare when compared to most of the stuff from open play. AI managers aren't necessarily affected as hugely, as when they look for an opening, they generally approach accordingly. Which reminds me of the supposedly "super keepers" as of FM 2010ish, where there was an actually issue with a certain kind of one on one hard to convert and to be abused by the virtue of a CB bug, which only players would and could focus on exclusively (AI managers, from my experience to this day encourage multiple chance types, always). So in general, the stats compare pretty decently. That's also why the conversion stats of teams/individual players compare rather okay overall.

However some "statistically dominating" download/ME busting approaches are prone to encourage them, which is why you see that moans popping up every year, been following this on many sites where frustration is dumped and there are clearly connections. It's not happening by random chance, it's happening by going extremes. That way you can dick Barcelona in terms of possession, but stretching and challenging them you do not always, which is visible. Depending on the OP formation, the players as well as their current tactics, this looks more pronounced, sometimes less so. There are also supposed to be matches where results go against a run of play, which considering that outright penalties are converted at ca. 80% rates should be expected. Still if you ever have 15 SOT and convert miserably, it's time to take a real critically look at all of those and check how much that stat is getting inflated by the aforementioned ilk. At that point, real life football comparisons go out of the window, as outside of FM I have never seen a match of football where off 35 "chances" 95% were directly off throw-ins, free kicks or corner kicks in all of my life. It should spring real questions how decent your approach is either way immediately if it creates nothing from play.

[Matshit]

9 hours ago, Svenc said:

Quote

Of the 181 games won in the Premier League before last weekend, the team who had the most possession only won 103 – 57% in total. The team who had more shots on target than their opponents won 128 matches – 71% of the total.

 

57% is a lot. In last season in the Premier League, only one team managed more. 

 

And yes, there are tons of other statistics that will impact as well, but again there is a huge correlation between shots on target and winning matches. This correlation is a lot lower in my game, and if that is more than a statistical fluke it should be something to worry about.

And this isn't me not winning, or anything. I am not looking at my own matches at all. I have the lowest possession and the highest amount of shots on target, and I have no issue with that. 

[Svenc]

5 hours ago, Matshit said:

This correlation is a lot lower in my game, and if that is more than a statistical fluke it should be something to worry about.

I see what you're getting at as that is two bottom teams making the SOT ranks, one of which coming in 2nd place. As posted, I'd be interested in those teams actually goals scored / conceded. If the team with the 2nd most shots is placed 18th, they must be leaky as hell! All kinds of stuff could lead to this, including weird AI tactics too, or simply random chance. But I still think you are confusing stuff. The only thing you could gauge from that ranking is who may be expected to score the most goals, statistically based on the number of shots on target. If you're first in the SOT rankings and thus have the most shots, you would be expected to score the most goals or be thereabouts. That is the correlation. Often those teams scoring the most tend to be the more successful teams, but not always.

The reason for which is that unfortunately football tables don't tend to rank teams by the number of the goals they score, but by points. If teams score a lot of goals, that is entertaining, but if they ship even more, they end up like Stuttgart last Bundesliga season. In other words, there is a Shot On Target rating AGAINST too somewhere, as football isn't merely about attacking, but defending too. As I'm curious too, I did something like that for the some leagues in Europe last season, which overall doesn't look that different to mine (sorry for the formatting). As said, would be interesting to see the actually tables including goals/scored of yours, and what match day that actually is.

Teams ranked by SHots On Target (FOR)  /// position in actually league table

Bundesliga
1 Bayern Munich, 1st
2 Borussia Dortmund,  2nd
3 VfL Wolfsburg,     8th
4 B. Moenchengladbach  4th
5 VfB Stuttgart         17th
6 Bayer Leverkusen     3rd
7 Koln             9th
8 Schalke 04         5th
9 Augsburg         12th
10 FSV Mainz 05         6th

EPL
1 Tottenham       3rd
2 Manchester City       4th
3 Arsenal          2nd
4 Liverpool       8th
5 Leicester       1st
6 West Ham       7th
7 Chelsea          10th
8 Everton          11th
9 Southampton     6th
10 Newcastle       18th

La Liga
1 Real Madrid     2nd
2 Barcelona     1st
3 Sevilla        7th
4 Rayo Vallecano     18th
5 Atletico Madrid     3rd
6 Malaga     8th
7 Athletic Bilbao     5th
8 Las Palmas     11th
9 Celta Vigo     6th
10 Deportivo La Coruna 15th

[Matshit]

4 hours ago, Svenc said:

The reason for which is that unfortunately football tables don't tend to rank teams by the number of the goals they score, but by points. If teams score a lot of goals, that is entertaining, but if they ship even more, they end up like Stuttgart last Bundesliga season. In other words, there is a Shot On Target rating AGAINST too somewhere, as football isn't merely about attacking, but defending too.

All of this is completely irrelevant.

I ran a quick calculation of how big the correlation was.

For possession: In real life in Italy there is a correlation between the possession table and position in the table of 0,66. Meaning there is some correlation between them, but not an extreme one. In my game this number is -0,15. That means the possession table is about as related to the actual table as just putting the teams in random order. The small correlation there is, actually say that having possession is actually a negative thing, but that could very well be influenced by my own team.

 

For shots per game: In real life there is a correlation of 0,83. In other words a very large correlation, meaning that number of shots on goal per game have a huge influence on where you end up in the table.

In the game this is down to 0.46. That is a gigantic difference, and if other tests can show similar results for both possession and shots that should be a concern.

[Svenc]

1) Possession. I see your point, the purely correlation between possession and the rankings, which are more random in FM for sure. I'm personally more interested in the shots, as due to AI / ME issues -- I've seen it myself how picks in formation can sadly influence this quite some already, and AI is... pretty bad on this. On my quick runs for about half seasons on full details on my slow CPU I've seen Napoli dropping as low as 46% of possession (rank 18). That doesn't keep them from winning, as it should be, as the correlation between winning and possession in football is marginal, which is the prime reason why I'm not overly sussed about the randomness in the possession rankings here. It's worth pointing it out, though, mind as those stats are over the place because of AI decisions and ME. You misread the Opta quote I think, it said that in their sample size of matches, in 57% of the cases the team with more possession won (with shots that is, unsurprisingly different: 71%). The German Roland Loy conducted a similar study based on several classes too, and in his sample size of 3000 matches he found that in a third of them the team that had more possession won (the standard of the league as the Telegraph noted below can influence). Here's more EPL seasons of this, and Pinnacle go a bit more in-depth (Pinnacle is a place for guys betting, so are interested in  performance indicators, rather than crowd wisdom perpetuated by general media):

http://www.telegraph.co.uk/sport/football/competitions/champions-league/10793482/Do-football-possession-statistics-indicate-which-team-will-win-Not-necessarily.html

Quote

In the A-League, the team with more possession loses 57 per cent of the time. More remarkably still, the team that completes more passes in the opposition’s final third – proper passing, not just knocking it around the defence – loses 59 per cent of the time.

Whilst FM's match engine doesn't despict a big difference between classes here, that's still a completely reverse of the EPL figures.

https://www.pinnacle.com/en/betting-articles/soccer/possession-in-soccer
 

Quote

Possession figures tell you little of how effective a side may be, although possession based teams may also be good teams, creating an illusion of a general widespread correlation. It merely helps to inform you about how they tactically chose or were compelled to play the match.

2) Shots/Shots On Target. Whilst solely looking at the shots/for is a simple metric, there is obviously a higher correlation between the total shot rates taking shots / for+against into account and the tables, why not. At the moment nobody knows what you're doing though. Is this based solely on that one standing? I went through all of my previous saves, and it's roughly in line what I found for the previous European seasons and the above screen shots. I thus went with Serie A as I thought it may be league specific. Not seen a team at the completely bottom topping stuff yet. edit: With full details disabled things appear more random on mine, with several struggling teams having statistically the more shots. That quick sim doesn't at all simulate matches, but simply churns out results and numbers though, including fewer shots in general, so it's hard to compare that. I had crunched another off real football last season with was Ligue 1, which is a bit closer to yours, again, ranked by sot / for (and the ginal league standing next to it). Several sides with decent shot on target counts finishing short off expectations,  barely scraping through or getting relegated even.

1    Paris Saint Germain    
2    Lyon     2
3    Marseille     13
4    Caen     7
5    Monaco 3
6    Nice     4
7    Reims     18
8    Toulouse 17
9    Lille     5
10    Lorient 15