A different perspective

[TorontoChelsea 4,064]

Football statistics are mostly relatively new and still developing. I was an early proponent of sabremetrics in baseball (read or even watch Moneyball to try to understand if you want) and while it will never be the same in football because it's a team sport, there are lessons to be learned. One of the things sabremetrics did is re-frame realities. Before, people thought that the key to scoring runs was getting hits. Sabermetrics actually showed that the key to scoring runs was not making outs. The equivalence in football would be not giving up possession. Football statistics will reward the aggressive and punish the conservative. If you attempt 15 long balls a game and succeed on 5 of them, are you really helping your team? Your counting stats will say “5 successful long balls”, but that player has undoubtedly had bigger a negative impact than a positive one. We measure positive impact, so why not also measure negative impact?

So, I’ve come up a statistic that combines incomplete passes with turnovers, times dispossessed, and shots minus goals gives you a total number of times when a player has conceded possession per start/appearance (for some players, it makes a big difference as they have had a lot of sub appearances . A more advanced stat would do turnovers per 90 minutes on the pitch. Premier League totals only). Chelsea totals-

Torres-14.1/11

Ba-10.2/8

Mata-14.3/12.7

Hazard-13/11.9

Oscar-16.7/11.9

Lampard- 15.3/ 10.9

Ramires-10.4/8.3

Mikel-8.4/8.2

Luiz-13.3

Ivanovic-7.7

Cole- 6.9

Azpilicueta- 10.6/9.4

Luiz and Ivanovic only had one substitution appearance so just used their starts. Cole had no sub appearances.

Obviously, you have to look at these stats in the context of positive stats as well. Ba had a very low negative effect on the team, but he also contributed very little positively. Mata gave the ball away a lot, but his offensive contribution was outstanding. You also have to put it in context of a player’s responsibilities. Someone who crosses the ball into the box a lot is going to get a lot of assists but they’re also going to get a lot of incomplete passes. I am not trying to show some perfect statistic that summarizes every player’s mistakes. This, for me, is a process of statistical discovery as a way of trying to understand football in a different way. I think people pay a lot of attention to the positives that players contribute, but not enough to the basic negatives.

I also think this goes to the way people appreciate or not players like Mikel or Luiz. The player like Luiz who is involved in many more positive plays but also many more negative. Mikel doesn't do lot that's special, but his ability to keep possession allows others to. Or, you can look at Aziplicueta and Cole. Azpilicueta had a very good first season with us and his future is very bright indeed, but I think something like this shows how undervalued Cole is. Azpilicueta and Cole had the same rough number of passes per game (Azpi 3 more) but Aziplicueta averaged 9.6 incomplete passes per start and Cole, 4.7 (despite Cole averaging more long balls and only 0.7 fewer crosses a game). Azpilicueta had 4 assists and Cole, 1 goal and 2 assists so their offensive totals were similar as well. That means that Cole was vastly more efficient on the offensive end.

Many people want all the players on the pitch to be of the high positive kind, but having players that retain possession is absolutely crucial. You need what we call in the NBA “glue players” players who move the ball and let other players be more effective. One of the reasons ManU won last year is that they were able to retain possession so effectively. They had the best passing percentage in the league and the second best possession percentage. (In fact, the big four leagues, the winners all had the best pass percentage and apart from ManU who was 0.2% behind Arsenal, they all had the best possession percentage. We were 6th in both. It’s an area I am sure we will improve in under Mourinho. Anyway, thoughts, comments, and ideas on how to improve the statistic or even suggestions for a new statistic are welcome.

Click here to view the article

[Stingray 9,441]

Intruiging concept .... could you dumb it down though, the meaning of the numbers is fuzzy to me ...

eg:

Mata-14.3/12.7

Hazard-13/11.9

What do the number refer to actually? Is it player -conceded/offensive contribution

[TorontoChelsea 4,064]

The first split is per start, the second is per appearance. So Mata gives the ball away 14.3 times per start and 12.7 per appearance. For someone like Oscar or Lampard who have a lot of sub appearances, it makes a big difference.

[Stingray 9,441]

Mata gave the ball away a lot, but his offensive contribution was outstanding. You also have to put it in context of a player’s responsibilities.

It would be awesome if you could operationalize this ... somehow...

Conceded/start and appearance

Offensive contribution/start and appearance

This then 'normed' or 'weighed' relative to player role ...

Uhm .... brainstorming here, mate.

[TorontoChelsea 4,064]

It would be awesome if you could operationalize this ... somehow...

Conceded/start and appearance

Offensive contribution/start and appearance

This then 'normed' or 'weighed' relative to player role ...

Uhm .... brainstorming here, mate.

it's all good. I thought about that, but it's so hard to combine the two. If you have any ideas!!! (And for some none-attacking players, measuring the positives is near impossible.)

[Stingray 9,441]

it's all good. I thought about that, but it's so hard to combine the two. If you have any ideas!!! (And for some none-attacking players, measuring the positives is near impossible.)

Well I guess you have the classics in terms of offensive contribution. But they won't do ....

What if you change concept a bit and indeed go for 'positive contribution'. Your indicator doesn't have to be so direct so direct as eg: assists. If we assume that positive contributions add to the final team result and conceded damages the team result, postitives can be very generic.

For example: number tackles, Duels won, Headers, pass completion, .... If would put that in a number somehow.

After that .. if you have a good positive contribution number besides the conceded one .... you could assign ratios per 4231 position for example. thus balancing the general number. And every player could have his generic performance 'stat'.

mmmm.

[TorontoChelsea 4,064]

I think it's a good idea...The formula would be complex and need a lot of tinkering but would be fun! (Offensive contribution positive/negative and defensive contribution positive/negative)

[Stingray 9,441]

I think it's a good idea...The formula would be complex and need a lot of tinkering but would be fun! (Offensive contribution positive/negative and defensive contribution positive/negative)

Don't we have some math wizz kids on here that could help out?!

[manpe 10,861]

Football statistics will reward the aggressive and punish the conservative. If you attempt 15 long balls a game and succeed on 5 of them, are you really helping your team? Your counting stats will say “5 successful long balls”, but that player has undoubtedly had bigger a negative impact than a positive one.

I don't really know about this. A player that tries 15 long balls and fails at 10 of them still shows more ambition and drive than the one who plays it safe and has no outstanding negative or positive impact. It would be really hard to find a player who has impressive positive impact and minimal negative impact, because you can make positive plays only by risking failure more often than not. If one of those 5 long balls led to a goal, then I would rather take those 10 failed passes along the goal instead of having a player who doesn't even try and rarely fails.

From our squad, maybe the best comparisons would be Luiz and Mikel. David loses the ball a lot, concedes mind numbing fouls from time to time, tries ambitious passes with overwhelming failure ratio, but we still rate him much more than Mikel, who retains possession well by playing safe passes, but offers little in offense and is often passive at defending, thus Luiz's mistakes get highlighted more because he tries more and has higher chance of hitting either the positive or negative side of statistics. Mikel, on the other hand, shines at passing accuracy, but nothing else - does that mean he has only positive effect?

[The Rising Sun 710]

Great concept.

What would really be interesting is to have a comparison of exactly these stats, with the Premier League Winners and Champions League winners.

That will put this exercise into perspective as well as allow us to see where we are in terms of players in the same position comparisons against winning teams.

However, i dont see Mikel really "keeping possession". His style is more mechanical, and not that effective defensively or offensively.

Clearly i dont rate him at all. So I am SORRY for putting this comment in a rather interesting topic!!

Just had to mention.

[The Skipper 20,593]

I like this... Something like NBA's PER formula. Very interesting concept. If we could somehow divide the type of players (e.g. having a slightly different formula for defenders, midfielders and forwards) that could be interesting. Would be a very, very complex formula though.

[The Mak 4,459]

Love all this stuff. @TorontoChelsea, where did you get the figures exactly? I do Statistics for Psychology and I'd love to mess with some fun figures for a change! (sad I know!)

Well I guess you have the classics in terms of offensive contribution. But they won't do ....

What if you change concept a bit and indeed go for 'positive contribution'. Your indicator doesn't have to be so direct so direct as eg: assists. If we assume that positive contributions add to the final team result and conceded damages the team result, postitives can be very generic.

For example: number tackles, Duels won, Headers, pass completion, .... If would put that in a number somehow.

After that .. if you have a good positive contribution number besides the conceded one .... you could assign ratios per 4231 position for example. thus balancing the general number. And every player could have his generic performance 'stat'.

mmmm.

I was thinking about this, but I'm unsure how you would combine each stat into an overall stat exactly.

Still, it'd be interesting to see stats like: passes complete in opposition half vs. possession given away in own half!

[TorontoChelsea 4,064]

I don't really know about this. A player that tries 15 long balls and fails at 10 of them still shows more ambition and drive than the one who plays it safe and has no outstanding negative or positive impact. It would be really hard to find a player who has impressive positive impact and minimal negative impact, because you can make positive plays only by risking failure more often than not. If one of those 5 long balls led to a goal, then I would rather take those 10 failed passes along the goal instead of having a player who doesn't even try and rarely fails.

From our squad, maybe the best comparisons would be Luiz and Mikel. David loses the ball a lot, concedes mind numbing fouls from time to time, tries ambitious passes with overwhelming failure ratio, but we still rate him much more than Mikel, who retains possession well by playing safe passes, but offers little in offense and is often passive at defending, thus Luiz's mistakes get highlighted more because he tries more and has higher chance of hitting either the positive or negative side of statistics. Mikel, on the other hand, shines at passing accuracy, but nothing else - does that mean he has only positive effect?

The problem is that you don't know how many of the other possessions would also have led to a goal or saved a goal conceded. If the player had passed to better targets and completed a higher percentage would it also have led to a goal? Obviously, in the case you site above, it's worth it because a goal a game is enormous, but hitting 5 out of 15 long balls in reality is never going to be worth it.

[CHOULO19 24,332]

Brilliant stuff mate.

I have suggestion, what if the stat is 'per zone'. Meaning the stat to be divided into first, middle and last third. Because obviously losing the ball in the final third is not as bad as losing the ball in the middle or worse in defense. But that will take very accurate stats.

[TorontoChelsea 4,064]

Love all this stuff. @TorontoChelsea, where did you get the figures exactly? I do Statistics for Psychology and I'd love to mess with some fun figures for a change! (sad I know!)

I was thinking about this, but I'm unsure how you would combine each stat into an overall stat exactly.

Still, it'd be interesting to see stats like: passes complete in opposition half vs. possession given away in own half!

I used whoscored. @choulo19. Thanks! Ideally, the statistic would be even more advanced to be divided by position in the field as well as by per/90 minutes. @theskipper. Yes, the PER for the NBA is a very useful stat. Any stat that isn't baseball (i.e. team sports) is going to be flawed but I think it really does help the understanding of the game. For example, how efficient is Real Madrid's strategy of using Ronaldo as their go-to guy for everything. Yes, he scored an enormous number of goals but...he takes 7 shots a game. Ronaldo took more shots than our top three shooters Lampard, Mata, Hazard combined (and they scored more). 34 goals looks less impressive when you factor that in. Would they be better off trying to spread the shots around the pitch? The NBA still has that "Player X was huge, he had 35 points but then you look at the boxscore and see he was 13-35 shooting so he actually had a terrible game).

[Stingray 9,441]

We could slowly build this and let it evolve to bypass the complex formula problem....

I think the Conceding part is is generally a furtile concept.

Further questions:

1: What would constitute Positive contributions.

2. How do we balance Pos/neg according to the three zones (like Choulo said ...)

Then we could maybe take it from there?

[Stingray 9,441]

Still, it'd be interesting to see stats like: passes complete in opposition half vs. possession given away in own half!

Interesting ... the weight factor then should be much higher for defenders than for attackers

[The Skipper 20,593]

We could slowly build this and let it evolve to bypass the complex formula problem....

I think the Conceding part is is generally a furtile concept.

Further questions:

1: What would constitute Positive contributions.

2. How do we balance Pos/neg according to the three zones (like Choulo said ...)

Then we could maybe take it from there?

3. What would exactly constitute a negative contribution?

4. Some sort of points system e.g. passing in the final third contributes to the higher end whilst losing it there is less punishable than losing it in a dangerous zone yanks off a greater number of points?