Topic: Statistics and Analytics - insight into our performance

[PhaseOfPlay]

Don't be fooled by the title...

As it was basically a competition to get picked, I went with a title which sounded like I would solve something that doesn't have a metric to measure it, which I don't think I really did, I just proposed some indicators.

Acutally, I think I probably do have something that measures how much teams leave themselves open when a goal behind, but not quite whether the shell or not when a goal up.

Interesting stuff again. I'll keep an eye on OptaPro to see if they put the videos up.

[scutty]

League Champions

1900-01, 1905-06, 1921-22, 1922-23, 1946-47, 1963-64, 1965-66, 1972-73, 1975-76, 1976-77, 1978-79, 1979-80, 1981-82, 1982-83, 1983-84, 1985-86, 1987-88, 1989-90 18

European Cup Winners

1976-77, 1977-78, 1980-81, 1983-84, 2004-055 5

FA Cup Winners

1964-65, 1973-74, 1985-86, 1988-89, 1991-92, 2000-01, 2005-06 7

League Cup Winners

1980-81, 1981-82, 1982-83, 1983-84, 1994-95, 2000-01, 2002-03, 2011-2012 8

UEFA Cup Winners

1972-73, 1975-76, 2000-01

All ways proud to use stats when talking about Liverpool

[ElstonGunn]

Hi mate - when I first wrote a piece on it (which was about a year ago) I hadn't seen anyone else looking at that particular stat. I'm sure they probably were, and as you say, they certainly are now.

Also - cheers for posting the link, Das Liverpool.

Ah, got it. I didn't realize it was something you'd been doing for a while. Sorry!

[The Woolster]

Interesting stuff again. I'll keep an eye on OptaPro to see if they put the videos up.

I think the one you might find most interesting is the one by the guy from man city. He showed how they applied some of their analytics, in this case passing networks, to opposition analysis, and then through to the training ground and then applied to the matches. It's definitely one I'll watch again.

[robgomm]

The thing about stats is that once certain successful patterns get identified, they become rote patterns in the game (which is what happened with the long ball game in the 70's and 80's). Because they are common identified patterns of play, they have some success, which means the "thought" gets taken out of that part of the game. Then the game becomes more stifled and "boring". Then along comes someone or a team or a group of coaches with a different slant - a solution to the rigid patterns - and the game changes again, and the stats become reasonably redundant, and a new set of statistical benchmarks become sought-after. It's a really interesting dialectic

That's a great analysis, contexualises the use of statistics - very easy to believe that they unveil intrinsic patterns but what they do is quantify the patterns we currently make use of. Perhaps there may be some statistical laws of the game, though.

[G1 Jockey 4(betfair)]

The thing about stats is that once certain successful patterns get identified, they become rote patterns in the game (which is what happened with the long ball game in the 70's and 80's). Because they are common identified patterns of play, they have some success, which means the "thought" gets taken out of that part of the game.

too lazy to look for the original post but this is spot on.
And its one thing you could see Rafa trying to nulify from day one here.
One season he would use wingers then change it to another way.

You have to have difefrent ideas because opposition managers (most anyway) do look at teams strong points and will mostly look to try and stop it happening.
Over a season one or two may find that way and eventualy the rest now know.


as for stats.
Well ive always said its a lazy mans way for trying to win at gambling.
Stats are usefull till everyone kops on...then that stats useless and even can cost you.

the key is spotting things before the stat men get their fingers into it.

Thats being able to spots things visually.

[ElstonGunn]

The thing about stats is that once certain successful patterns get identified, they become rote patterns in the game (which is what happened with the long ball game in the 70's and 80's). Because they are common identified patterns of play, they have some success, which means the "thought" gets taken out of that part of the game.

too lazy to look for the original post but this is spot on.
And its one thing you could see Rafa trying to nulify from day one here.
One season he would use wingers then change it to another way.

You have to have difefrent ideas because opposition managers (most anyway) do look at teams strong points and will mostly look to try and stop it happening.
Over a season one or two may find that way and eventualy the rest now know.


as for stats.
Well ive always said its a lazy mans way for trying to win at gambling.
Stats are usefull till everyone kops on...then that stats useless and even can cost you.

the key is spotting things before the stat men get their fingers into it.

Thats being able to spots things visually.

Heh, right, because no one has ever gained an advantage, either in sports or as a bettor, through superior statistical analysis.

[ElstonGunn]

Get ready for a whole bunch of pretty pictures that make Liverpool look good!

11Tegen11:



Michael Caley:



Our worst projection (Constantinos Chappas):



And, three for fun:

From Colin Trainor:



From Constantinos Chappas:



From Michael Caley:

[ElstonGunn]

Trainor also said Coutinho would be second or third on that ExpG/KP chart but he just missed the cutoff for number of key passes. So we would have the three best in the leauge!

[ElstonGunn]

Oh, and this from earlier (definitely read the whole thing):

According to data gurus Opta, in the history of the Premier League, how many players averaged one goal or more per 90 when playing at least half a season?
…
…
…
The answer: One.
Ronaldo in 2007-08.
Here’s the thing though… Ronaldo had 31 goals in 30.5 90’s that year, but 4 of those were penalties. His adjusted rate without the penalties is .89.
Sturridge’s rate this year is 1.
According to Duncan Alexander at Opta, prior to this season, there were three other players who played less than half the minutes in the season who eclipsed this mark: Ole Gunnar Solskaer  and Kevin Campbell in 98-99, and Papiss Cisse when he first arrived at Newcastle in 11-12.
That’s it.

[ElstonGunn]

[G1 Jockey 4(betfair)]

Heh, right, because no one has ever gained an advantage, either in sports or as a bettor, through superior statistical analysis.

yeah but understanding why gives you more of an insight.

and thats why spotting up and coming sides is far more lucrative

stats cant pick them out on a small sample....you can only do that visually through experience otherwise your pissing in the wind.

[ElstonGunn]

yeah but understanding why gives you more of an insight.

and thats why spotting up and coming sides is far more lucrative

stats cant pick them out on a small sample....you can only do that visually through experience otherwise your pissing in the wind.

It's just not true. You're confusing stats (which are limited, sure) and statistical analysis (which requires a lot of understanding). Lots of the stats (or, I guess I should say analytics) writers I link to here make their living using their models to bet on games.

[Blinis]

Lots of the stats (or, I guess I should say analytics) writers I link to here make their living using their models to bet on games.

Do they?

[Irakla]

That second goal from Skrtel was a 17-th we've scored form a set piece, which is the best feat for now, not only for the Premier league side but in Europe too.
And at last Sterling scored a goal without Suarez assisting him, all his previous PL goals came from LS.

[Redeo]

Oh, and this from earlier (definitely read the whole thing):

Interesting.
BTW, are you aware of any model that projects PL standings by incorporating goal difference?

[The Woolster]

Interesting.
BTW, are you aware of any model that projects PL standings by incorporating goal difference?

Do you mean basic goal difference to points/position?

If so, there is this http://5addedminutes.com/2013/01/07/not-just-a-tiebreaker-what-goal-difference-tells-us/

To project you'd want to convert it into a per game number and then work from there.

If it is something a bit more complex you are after, then maybe some kind of ELO model that incorporates goals, like this http://clubelo.com/

[barbudo]

It's just not true. You're confusing stats (which are limited, sure) and statistical analysis (which requires a lot of understanding). Lots of the stats (or, I guess I should say analytics) writers I link to here make their living using their models to bet on games.

It's a semantic minefield, but it wouldn't it be better to refer, for example, to counts such as "total number of shots" as "data", and to reserve "statistics" for the analytical procedures, which could be something as simple as "total shot ratio".

[ElstonGunn]

Do they?

Yep.

[BabuYagu]

Does anybody have an up to date mini-table for the top 7? Curious how we fare having beaten Utd & mauled Arsenal, Everton & Spurs

[rickardinho1]

Where are our goals have come from this season so far...



http://www.statsbomb.com/2014/01/attacking-styles-and-defensive-weaknesses/

Crazy how high our % of goals from outside the box is considering we've scored so many, compare that to City's and it shows you how good Suarez and Sturridge are outside the box.

Interesting that we have the second lowest % of goals scored by headers, but have the highest number of goals from set-pieces with 23

[Bird Bird Bird The Bird Is The Word]

Quite a few of our set pieces have been direct free kicks so that brings the % down, plus quite a few of our set piece goals have been "odd" finishes (Sturridge's elbow vs Sunderland, Skrtel's knee vs Arsenal etc) and then you take into account how many goals we've scored overall, that brings the % down too.

[ElstonGunn]

Does anybody have an up to date mini-table for the top 7? Curious how we fare having beaten Utd & mauled Arsenal, Everton & Spurs

[BabuYagu]

Thanks mate.

This is pretty interesting for me:-

vs Top 7   -    Goals For per Game = 2.50 / Goals Against per Game = 1.25
Overall     -    Goals For per Game = 2.52 / Goals Against per Game = 1.20

Does that shows that our performances are consistent regardless of the ability of the opposition? Or that we have upped our game for the "big games" and slacked off against some poorer opposition? Or too small a sample size to actually suggest anything?

[ElstonGunn]

Thanks mate.

This is pretty interesting for me:-

vs Top 7   -    Goals For per Game = 2.50 / Goals Against per Game = 1.25
Overall     -    Goals For per Game = 2.52 / Goals Against per Game = 1.20

Does that shows that our performances are consistent regardless of the ability of the opposition? Or that we have upped our game for the "big games" and slacked off against some poorer opposition? Or too small a sample size to actually suggest anything?

It's probably too small to suggest much, but I'd guess it comes to a combination of motivation/"getting up" for the game, and that the big teams are more likely to give us space to break into, which we love.

[ElstonGunn]

Beautiful:



By the way, notice that "Team+" score--Michael thinks we're the second best team in the league, just barely behind CIty.

[Roblaar]

Beautiful:



By the way, notice that "Team+" score--Michael thinks we're the second best team in the league, just barely behind CIty.

It also suggests that ourselves and West Brom are probably the most unlucky team, in terms of performance stats not turning into a fair amount of points.

As a predictive tool it may be less useful when considering fragility of a team/squad, 25games probably isn't long enough to build in 'black swan' type events like long term injuries/suspensions to one or more key players.  But then any model would struggle with that.

[Juan Loco]

Not a particularly smart stat, but I'm throwing it in here anyway.

This Liverpool team is the third highest scoring of the "Premier League era" after 25 matches played. It's scored more at this point in the season than any of Ferguson's, Mourinho or Wenger's teams had. Only City (11/12 and this current season) have had teams who've scored more after 25 games in.

It's not the most worthwhile stat in the world, but a nice indication of just how ridiculous this attack actually is.


Edit: By way of a Liverpool comparison, this current team has scored more 4 more goals after 25 games than the 87/88 team.

[ElstonGunn]

Not a particularly smart stat, but I'm throwing it in here anyway.

This Liverpool team is the third highest scoring of the "Premier League era" after 25 matches played. It's scored more at this point in the season than any of Ferguson's, Mourinho or Wenger's teams had. Only City (11/12 and this current season) have had teams who've scored more after 25 games in.

It's not the most worthwhile stat in the world, but a nice indication of just how ridiculous this attack actually is.


Edit: By way of a Liverpool comparison, this current team has scored more 4 more goals after 25 games than the 87/88 team.

Nah, goals scored are plenty smart stats. That's just incredible.

[jdpapa3]

Via Michael Caley.

I think this is why everyone around here is freaking out. A genuine chance at the title if we collect all 3. Even if you slash a few percentage points away since MC_of_A's projections inflate our chances more than a lot of other models, it's still a decent shot.

[BabuYagu]

Beautiful:



By the way, notice that "Team+" score--Michael thinks we're the second best team in the league, just barely behind CIty.

Nice viewing that.

One thing though, how can our Top4% be higher than Arsenal and our Title% lower?

[ElstonGunn]

Nice viewing that.

One thing though, how can our Top4% be higher than Arsenal and our Title% lower?

Seems possible. It's a Monte Carlo simulation, so he actually sims the season 10,000,000 times. So there are both more seasons where we win the title than Arsenal, and slightly fewer where we come top 4. I'd guess the main reason is that Arsenal have slightly higher projected points, but we have a GD superior to everyone but City. I bet the difference is because of the times where we end the season level on points with City for the title but lose out, if that makes sense.

[Pistolero]

Seems possible. It's a Monte Carlo simulation, so he actually sims the season 10,000,000 times. So there are both more seasons where we win the title than Arsenal, and slightly fewer where we come top 4. I'd guess the main reason is that Arsenal have slightly higher projected points, but we have a GD superior to everyone but City. I bet the difference is because of the times where we end the season level on points with City for the title but lose out, if that makes sense.

Really? I read it more as a Santa Monica...ie, the underlying differential between a) goal distributions b) potential penalty area fractions c) DFD projections, and d) exponential transfer smoothing - make the random variables impossible to call...the Pearson coefficient even more so. ......As such it can be seen as a 4%, 2, 1%, 3* season.. but obviously that remains to be seen.

[BabuYagu]

Really? I read it more as a Santa Monica...ie, the underlying differential between a) goal distributions b) potential penalty area fractions c) DFD projections, and d) exponential transfer smoothing - make the random variables impossible to call...the Pearson coefficient even more so. ......As such it can be seen as a 4%, 2, 1%, 3* season.. but obviously that remains to be seen.

Got as far as "Really" then you lost me

I´m out of my depth. But I will take both your words for it that it is correct.

[ElstonGunn]

Really fascinating article, basically arguing that it's impossible to identify finishing skill, at least over a one-season sample, and that a scout should only pay attention to the quality of positions the striker gets into. Here's the extremely convincing graphs:

Shot quality is very, very repeatable for strikers year-to-year:


While there is literally no correlation year-to-year in finishing:


Michael Caley has shown that finishing *is* a skill for the very, very best strikers if you get a giant sample, but even then it's pretty small. Remember when Suarez was a historically bad finisher in 2011/12 and a historically great finisher this year? I basically think finishing is a dramatically overvalued skill that's much less important than a player's ability to get into good positions.

[PhaseOfPlay]

Really fascinating article, basically arguing that it's impossible to identify finishing skill, at least over a one-season sample, and that a scout should only pay attention to the quality of positions the striker gets into. Here's the extremely convincing graphs:

[snip]

Michael Caley has shown that finishing *is* a skill for the very, very best strikers if you get a giant sample, but even then it's pretty small. Remember when Suarez was a historically bad finisher in 2011/12 and a historically great finisher this year? I basically think finishing is a dramatically overvalued skill that's much less important than a player's ability to get into good positions.

Are you sure that's not confusing "skill" with "outcome"?

Because you most certainly CAN identify finishing "skill"

Edit - I agree about the positioning, btw. But perhaps it is better to first define what they mean by "Skill", because there is a great misunderstanding of its definition that might lead to some confusion

[ElstonGunn]

Are you sure that's not confusing "skill" with "outcome"?

Because you most certainly CAN identify finishing "skill"

Edit - I agree about the positioning, btw. But perhaps it is better to first define what they mean by "Skill", because there is a great misunderstanding of its definition that might lead to some confusion

Well, it's measuring outcomes, obviously, but then so is every stat. And yet by measuring outcomes of other actions--i.e. passing, or dribbling--you get pretty substantial repeatability, which is often how statistical analysts define something as a "skill" rather than "luck."

Now, of course that's a different perspective on how you define skill than someone more tecnhique-minded such as yourself would use, and both are legitimate IMO. I'm sure finishing technique is very identifiable to those who know what to look for. What this suggests, however, is that it's just not that important. If the good finishers don't score goals from the same positions at a predictably higher rate than the bad finishers, then it should only hold a slightly higher position on your list of skills to value when signing a player than "ability to do the seal dribble."

Now, like I said, there's some other evidence that finishing "skill" does show up in sufficiently large (read: giant) samples, so it's not the same things as a seal dribble or some other irrelevant ability. But even then, it's much, much less important than the positions the shots are taken from.

[PhaseOfPlay]

Well, it's measuring outcomes, obviously, but then so is every stat. And yet by measuring outcomes of other actions--i.e. passing, or dribbling--you get pretty substantial repeatability, which is often how statistical analysts define something as a "skill" rather than "luck."

Now, of course that's a different perspective on how you define skill than someone more tecnhique-minded such as yourself would use, and both are legitimate IMO. I'm sure finishing technique is very identifiable to those who know what to look for. What this suggests, however, is that it's just not that important. If the good finishers don't score goals from the same positions at a predictably higher rate than the bad finishers, then it should only hold a slightly higher position on your list of skills to value when signing a player than "ability to do the seal dribble."

Now, like I said, there's some other evidence that finishing "skill" does show up in sufficiently large (read: giant) samples, so it's not the same things as a seal dribble or some other irrelevant ability. But even then, it's much, much less important than the positions the shots are taken from.

But the problem with the article is that it uses Messi as an example. It essentially (to my reading) says that just because Messi scored an out-of-this-world number of goals in one season, it doesn't mean that he will do so in the next. Am I right about that? And because of that outlier example, you shouldn't focus on goals scored, but on positions shots are taken from. Correct?

[ElstonGunn]

But the problem with the article is that it uses Messi as an example. It essentially (to my reading) says that just because Messi scored an out-of-this-world number of goals in one season, it doesn't mean that he will do so in the next. Am I right about that? And because of that outlier example, you shouldn't focus on goals scored, but on positions shots are taken from. Correct?

The Messi example is just a way to make it more interesting. It's the graphs above that are the meat of it. So, no, it's not because of the outlier example. It's because the positions the shots are taken from are highly repeatable--look at that first graph, how close the dots are to the best-fit-line, and how the line has a slope of almost 1--while the finishing, when holding the positions constant, is essentially random from year to year.

He's talking about the stats of a huge number of players here, not just a handful.

[PhaseOfPlay]

The Messi example is just a way to make it more interesting. It's the graphs above that are the meat of it. So, no, it's not because of the outlier example. It's because the positions the shots are taken from are highly repeatable--look at that first graph, how close the dots are to the best-fit-line, and how the line has a slope of almost 1--while the finishing, when holding the positions constant, is essentially random from year to year.

He's talking about the stats of a huge number of players here, not just a handful.

Nevertheless, the point he's making is that Messi's reduced scoring is because he isn't positioning himself the same as he did before, no?