Topic: Moneyball, Soccernomics and Liverpool's transfer policy

[shelovesyou]

One thing to note is that the FFP come into play next week, but the short fall in any loss of Champions League revenue will not come into effect until the 2013/2014 season.

[Sangria]

That's one of the underlying questions isn't it? What are the metrics being used which could have led to that kind of statement? Let's leave aside our personal views of the player (whether we think he's great, alright, mediocre or terrible is irrelevant to this thread). But what kind of measurements would lead Comolli to think that he's an undervalued player?

One stat that's been suggested, and which seems to have some currency considering the players we were linked with, is chance creation. Downing has been prominent in the list of top chance creators for a few seasons.

[IndianaRed]

sorry lad - not trying to nit pick - its a foible that young kids have over here, they add 'est' to words - 'bestest', 'mostest' etc so to me, this side of the pond it comes across as childlike - I'm guessing thats not how you are looking to come across thats why I 'picked' on it just trying to be helpful really - no offence meant.

two countries divided by a common language eh

i see, it's no problem.  the term "winningest" is actually a really common term in sporting contexts over here.  of course there are the childish "bestest" and "mostest" and that sort of thing, but i didn't imagine "winningest" might be taken that way.  i really should have said "most successful" anyway, and i only said winningest because at that moment i had a bit of a brain block and couldn't think of "most successful" for whatever reason.  anyway... cheers for looking out for me on that, and also thanks for supporting my comments in the last 2-3 pages of this thread.

[Vulmea]

That's one of the underlying questions isn't it? What are the metrics being used which could have led to that kind of statement? Let's leave aside our personal views of the player (whether we think he's great, alright, mediocre or terrible is irrelevant to this thread). But what kind of measurements would lead Comolli to think that he's an undervalued player?

compare him with young - roughly the same fee but we're told dramatically more in wages for Young who had just a year left on his contract

27 year old about to hit his prime? Young 26.

33 england caps -  ashley young has 19 but the hype around Young is so much greater isn't it?

number of goals and assists compared to Young are they roughly similar too?

[IndianaRed]

One stat that's been suggested, and which seems to have some currency considering the players we were linked with, is chance creation. Downing has been prominent in the list of top chance creators for a few seasons.

another stat which seemed to be important to comolli (and this is one i might or might not attempt to find quotes for) is something i'll call "appearance rate" which relates to a player's injury record.  Downing has a really good injury record/appearance rate having had only one major injury back in the middle of '05 which kept him out til about the beginning of '06, and otherwise has not been unavailable due to injury very much.

i haven't checked on the appearance rates of the others included in last summer's purchases, but aside from bellamy, i don't think any of them have been unavailable due to injury very much.  i'll check on that now.

[JohnHenrysIceCreamCone]

Has it had mixed results or is the principle likely to experience diminishing returns over time? The whole principle was to look for value that others don't appreciate. ONce others start using the same measurements, the value of players with those measurements goes up. What worked for the A's is that everyone thought they were fucking nuts!

If everyone uses the same measurements, the accuracy and predictability of a player's performance will improve but the edge that gives you will disappear over time.

Moneyball (as described in the book) isn't a panacea that magically solves everything, it's just one more idea that gave a team an advantage at a particular time. The same goes for tactical innovations, for dietary innovations and so on and so on.

And I have to say that I wouldn't have minded some of the Red Sox's 'mixed results' over the last few seasons since Henry appointed a statistical wonderboy. Two World Series? Would I be happy if we'd won two league titles in the last ten years? Fucking right I would!

AS for Comolli (who I guess you're referring to) I'll say it again - I've yet to see anything written or reported about him that suggests that he uses any meaningful and measurable data to select players.

Both. It is tough to judge, at least in my opinion, just how effective Beane was before other teams caught up because all of this took place right in the middle of the steroid era. Players were plucked out of obscurity all over the league and went on to have an amazing few years not because of some obscure statistic but because they were juicing. And in a clubhouse like the A's in which steroids were practically a part of the culture ( Canseco, McGwire, Giambi, Tejahda) it is even more difficult to judge the true reasons for their success. Keep in mind the A's have not made the playoffs in 7 years ( right around when stricter drug testing was implemented).

But you are right, 'Moneyball' does not magically solve everything. It is a piece of the puzzle but a vastly overrated one. Good players have good statistics.

[Sangria]

another stat which seemed to be important to comolli (and this is one i can't/won't attempt to find quotes for) is something i'll call "appearance rate" which relates to a player's injury record.  Downing has a really good injury record/appearance rate having had only one major injury back in the middle of '05 which kept him out til about the beginning of '06, and otherwise has not been unavailable due to injury very much.

i haven't checked on the appearance rates of the others included in last summer's purchases, but aside from bellamy, i don't think any of them have been unavailable due to injury very much.  i'll check on that now.

One way of looking at injury records that someone suggested was to look at the type of injuries they suffered. Impact injuries tend to be bad luck more than anything, and aren't indicative of injury-proneness. Recurring muscle injuries, especially those in certain areas, may indicate injury-proneness. Again, a descriptive rule of thumb rather than a measurable stat, but the reasoning behind it is open to peer review. Personally, I wouldn't bother with the appearances stat for players like Downing. I'd prefer to look for someone younger and hence without much of a track record, with pace as their primary asset, and exploit the hell out of them before they burn out. I don't rate the tactical thinking of English players, so I wouldn't look for anything sophisticated that may come with age. Get them young, sell them when they're around Downing's age, having got the best years out of them in the mean time. If they amount to more, then pat ourselves on the back for striking lucky.

[Alan_X]

Couple of points:

I don't see that Barca would give useful individual player's stats. You could learn a lot about the way a group of great players work as a team. But drop Xavi and put in Fabregas and what does that tell you? There might be a marginal change in the overall performance but how does that feed into an understanding of what either player contributes individually? As Barca would probably still beat most teams without either player the logical conclusion would be that both players are worthless.

In theory, it would be easier to judge a player's individual effectiveness, and isolate what they actually do if they are the one decent player in a shite team. But we all know from experience that it's more complicated than that. The perfect example is Charlie Adam - the best player at Blackpool and it was clear exactly what he contributed. But... it seems that what he contributed to Blackpool is not directly transferable.

Which brings me back to the problem I see with trying to apply Moneyball to football. Baseball is an incredibly complex game but the starting position of the players at the start of each play are effectively the same. The variable is where the batting team's players are on base as the inning develops. But even that has a limited number of possibilities.

In football there is no equivalent. Unless of course you take the Charles Hughes approach and eradicate the influence of 90% of the players on the pitch - long ball, knock down by the big man and the striker scores (also seen being played at QPR recently by a visiting team in red).

What would be good to try and do is forget about the pitch and analyze the relationships between any given player and the nearest half-dozen players (his own and the opposition). I started sketching this out after QPR. I called them 'Downing diagrams' because he was the player who fascinated me most during that game. Some people thought he had a good game but what was evident and what happened very time he was in the attacking third and being closed down, was that he offloaded. He always looked to pass or stopped to think ad lost the ball.

I think you could measure that ad come up with useful information. By contrast, im the same situation, Luis will hold on to the ball when he should probably off-load. In some ways you could say that both are measurably as bad as each other because both waste possession. But Luis looks like a world class player (dribbling and running at defenders is valued) and Downing looks like a lemon. Both are reducing our ability to win games.

Anyway. Need to get home.

[iamrobk]

No mate, my gut reaction (and the limited evidence of this season's transfers) is that whatever measures are being used are virtually worthless in determining the performance of the team.

One of the most interesting things about the book was that the A's improved after the midseason trades. They didn't just buy with a view to the long term. They were able to identify players that would have an immediate impact based on the measures they were using.

I think it would be interesting to come up with what we think are useful metrics. The ones being used aren't working.

And I think that's what statisticians looking at football are trying to do right now.  To either take the data we already have (chances created, passes completed, etc.) and find a way to make it more useful, or to find new stats to measure that are inherently more useful.

[Alan_X]

One orher particular problem with football is that the most valuable statistic (scoring a goal) is so infrequent.

For strikers, one in two is a decent scoring rate. Which means that you can expect them to fail at what they are paid to do 50% of the time on average.

[Vulmea]

What would be good to try and do is forget about the pitch and analyze the relationships between any given player and the nearest half-dozen players (his own and the opposition). I started sketching this out after QPR. I called them 'Downing diagrams' because he was the player who fascinated me most during that game. Some people thought he had a good game but what was evident and what happened very time he was in the attacking third and being closed down, was that he offloaded. He always looked to pass or stopped to think ad lost the ball.

the italians have drilled their players this way for decades - depending where the ball is thats where you should be - if it in zone xy then you need to be in zone ab, if you are in ab then the ext player has to be in bc etc.

Lobonovsky had taken it a step further earlier implying that any rigid application of a system is probably flawed......which turns into the old level 3 football thread and where the hell do we go from there -

however 'Downing diagrams' would depend not only on your own team's relationships but those of the opponents and players are increasingly encouraged to have game intelligence which means they can adapt to circumstances meaning anomalies could be either productive decisions or mistakes and you would need to place them in context - it all becomes very messy not to say it can't be modelled but the greater depth of detail the greater the level of inaccuracy

what we need to do is use our eyes but at the same time remove the bias - use stats to give pause for thought - identify patterns -  identify  significant moments - the sophistication of the models and stats required are very difficult for me to imagine - which brings us back to football being a simple game complicated by idiots

the idea of 'chance creation' as  a key metric has interested me for a while however it requires definition and context -

and as we've found - if you make it an objective be careful what you wish for because we've created oodles of chances (which is one of the criteria we used to select the players allegedly) so that worked - its an old adage that any system will change to support the metrics gathered -  unfortunately you dont win games on chances created - we could easily argue the next link in the chain is  to buy players with high chance conversion ratings and we are home and hosed once we find the right balance.........which brings you back to having to find players that compliment each other not plucking them at random because of their stats -

[Red Genius]

One orher particular problem with football is that the most valuable statistic (scoring a goal) is so infrequent.

For strikers, one in two is a decent scoring rate. Which means that you can expect them to fail at what they are paid to do 50% of the time on average.

That's not entirely accurate though is it?

As you are saying that scoring 1 in 2 is decent i.e acceptable. So by holding down a 1 in 2 record they are succeeding. If they were scoring 1 in 4 they would be failing in 50% of their matches as per a 1 in 2 ratio being acceptable.

[Giono]

So there is no doubt Moneyball is effective in baseball because the Red Sox hired a guy?

They hired THE guy.

[Giono]

That was me.

It was in response to The Carroll Law which states that any Luis Suarez shot will hit the post and go in instead of out at the mere sight of the Big Yin's pony tail.

The Seemingly Divine Pony Tail !

[Carlos: Very Kickable]

I think it would be interesting to come up with what we think are useful metrics. The ones being used aren't working.

Sorry about the semantics - because of my background i tend to get extremely pernickity abut these things!

It's an incredibly interesting topic. I would love to know what metrics Rafa used to such great effect. Also clubs like Porto, Lille and Napoli seem to get some amazing results.  You're right, the key factor is deciding which metrics to use; I think a lot of clubs have got caught up using a lot of less useful data when they could ask fewer, better questions.

[Carlos: Very Kickable]

That's not entirely accurate though is it?

As you are saying that scoring 1 in 2 is decent i.e acceptable. So by holding down a 1 in 2 record they are succeeding. If they were scoring 1 in 4 they would be failing in 50% of their matches as per a 1 in 2 ratio being acceptable.

I thought what he was saying is that you pay them to score so, on average, even a decent strikeforce is failing half the time.

The underlying question is fascinating; i truly believe a club that spent a pound on sports science for every ten pounds on transfers would absolutely storm the league.

[IndianaRed]

my god... i had typed 75% of a somewhat long post and then i accidentally hit refresh and lost it all... ugh

edit:  well, at least this has led to me installing an extension for firefox called Lazarus, which auto-saves text filled into web forms...

i'll post it later today

[Not A Scouser]

I thought what he was saying is that you pay them to score so, on average, even a decent strikeforce is failing half the time.

The underlying question is fascinating; i truly believe a club that spent a pound on sports science for every ten pounds on transfers would absolutely storm the league.

In baseball, where all this started, the absolute best hitters fail about two thirds of the time.  However, there are about 500 discrete events that are easy to measure a year, so there is a ton of clear data.  In football you might get 100 events (shots?  chances?) but those events are so much more varied than in baseball that all the data will be muddy.

[steveeastend]

the italians have drilled their players this way for decades - depending where the ball is thats where you should be - if it in zone xy then you need to be in zone ab, if you are in ab then the ext player has to be in bc etc.

Lobonovsky had taken it a step further earlier implying that any rigid application of a system is probably flawed......which turns into the old level 3 football thread and where the hell do we go from there -

...I agree, love that . We should recreate it (level 3 thread), too much fun to not have it around

[Melbred]

Didn't know where to put this as I couldn't find the Comolli thread, but interesting read.

http://www.nesn.com/2012/03/liverpool-fc-director-of-football-damien-comolli-feels-club-is-headed-in-right-direction-excited-to-.html

[IndianaRed]

Alan Kayll works for NESN now?  did not know that.

[iamrobk]

Alan Kayll works for NESN now?  did not know that.

Yeah, this.  Wow.

[Gainsbarre]

Didn't know where to put this as I couldn't find the Comolli thread, but interesting read.

http://www.nesn.com/2012/03/liverpool-fc-director-of-football-damien-comolli-feels-club-is-headed-in-right-direction-excited-to-.html

good read thanks