You are not logged in or registered. Please login or register to use the full functionality of BayernForum.com

The Statistics Thread

Discussions about anything in general about Bayern, such as tactics, finances, kits, merchandise etc.
 

Re: The Statistics Thread

Postby JANCKER » Thu May 04, 2017 12:07 am

.........................................Image.........................................
User avatar
JANCKER
I tattooed BayernForum.com on my ass
 
Years of membershipYears of membershipYears of membership
 
Posts: 17591
Joined: Thu Jun 19, 2014 2:03 am
Location: 710 miles SE from Allianz Arena
National Flag:
Basque Country
Has thanked: 9162 times
Been thanked: 8471 times
Gender: Male

Re: The Statistics Thread

Postby Manchu » Thu Jul 27, 2017 8:35 pm

I wasn't planning on making another goalimpact post, but MUTU's justified love for Mario Gomez has forced my hand. To be brief, a couple months ago the goalimpact twitter account published a list of the 41 Bayern Munich players throughout all time who had the highest peak goalimpacts of all time under the current algorithms. Peak goalimpact is an estimation of how good a player is at their "peak" of 26 for field players or at ~30 for goalkeepers(although keepers generally decline very slowly, so it matters less for them) based on a players current/end of career goalimpact adjusted by the aging curve.

Anyways, here it is:
ImageImage source
This is based on a player's entire career, not just the time they played for Bayern, so there are a few odd players like Tim Borowski, who only played a single exceptionally season for us. It also doesn't include any of our signings this year; James would definitely be on there. There are also some surprises like Roque Santa Cruz actually being a good player.

Seeing Mario Gomez being rated as so much better than Lewandowski(and as the 5th best player to ever play for us is a big surprise, although their current goalimpact's are probably about the right now because Gomez is much older and hence further from his peak. For comparison, here's Mandzukic's chart:
Can't see tweet? Click here!

Then there is the Gotze question. For those interested, I have elaborated extensively on why I think he is rated so highly here.

Finally, there is the question of Lahm, who, surprisingly, does not appear on this list. A look at his chart reveals why:
Can't see tweet? Click here!

As a reminder about reading these charts, the bold red line is a player's goalimpact over time. The pale red line is the player's estimated peak goalimpact at the age of 26; it varies based on age and goalimpact. The blue line is an estimation of a players performance over the last 24 months; it has a strong tendency to regress toward the red line.

Thus, we can see that Lahm consistently out performed expectations from 2008 onward despite his not so incredible early career and stayed at about a ~175 in terms of performance until his retirement. Most players do not do this and instead have their performance regress to the expected level. Thus, Lahm is a complete outlier in that his goalimpact strongly underestimates his actual performance.
User avatar
Manchu
I'm a chatterbox!
 
Posts: 323
Joined: Wed Mar 15, 2017 6:34 am
Has thanked: 244 times
Been thanked: 256 times
Gender: Male

Re: The Statistics Thread

Postby MUTU » Thu Jul 27, 2017 9:00 pm

Not surprised with Gomez being ranked the best pure striker in Bayern history ;)
30GB free cloud storage. Click here for the referral.
User avatar
MUTU
Site Admin & EURO 2016 Prediction Game Winner
Site Admin & EURO 2016 Prediction Game Winner
 
Years of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membership
 
Posts: 31618
Joined: Mon Aug 14, 2006 11:00 pm
Location: L-Imqabba, Malta
National Flag:
Malta
Has thanked: 5582 times
Been thanked: 8319 times
Gender: Male
BayernForum.com fan club: Active member
BayernForum.com donator: Yes

Re: The Statistics Thread

Postby aterford » Fri Jul 28, 2017 12:30 am

I love Alonso but I have a hard time believing his goal impact was nearly the same as Gerd...
User avatar
aterford
I post all the time
 
Years of membershipYears of membershipYears of membershipYears of membership
 
Posts: 1595
Joined: Mon Jul 08, 2013 8:40 am
Has thanked: 1415 times
Been thanked: 861 times
Gender: Male

Re: The Statistics Thread

Postby Manchu » Fri Jul 28, 2017 2:01 am

aterford wrote:I love Alonso but I have a hard time believing his goal impact was nearly the same as Gerd...

Again, it's for Alonso's entire career, not just the time he played for us when he was on his last legs. Also, goalimpact is always a relative measurement to some degree, which makes it hard to compare people across such long lengths of time.

Additionally, the average goalimpact in the Bundesliga was probably much lower back when the league was just founded, so Gerd and the rest of the team might have had a quality advantage equal to or greater than the current team has over the rest of the Bundesliga, even though they were theoretically "worse" than the current team. Of course, there have been a lot of things which have happened since like the hyper-concentration of the best players into a few teams or the modern German youth program which raised the average quality of teams in the Bundesliga, so it's hard judging stuff like that.

Ultimately, it doesn't matter that much because Gerd is still a legend and one of the players most responsible for bringing Bayern into dominance, and he will remain to be that forever. Being a legend is more than being a good player.
User avatar
Manchu
I'm a chatterbox!
 
Posts: 323
Joined: Wed Mar 15, 2017 6:34 am
Has thanked: 244 times
Been thanked: 256 times
Gender: Male

Re: The Statistics Thread

Postby MUTU » Fri Jul 28, 2017 5:22 am

Correct me if I'm wrong but I am understanding that the Goalimpact is more of an indispensability measure than anything else. In a team with a bench as strong as the starting XI the Goalimpact ratings would therefore drop, right?
30GB free cloud storage. Click here for the referral.
User avatar
MUTU
Site Admin & EURO 2016 Prediction Game Winner
Site Admin & EURO 2016 Prediction Game Winner
 
Years of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membershipYears of membership
 
Posts: 31618
Joined: Mon Aug 14, 2006 11:00 pm
Location: L-Imqabba, Malta
National Flag:
Malta
Has thanked: 5582 times
Been thanked: 8319 times
Gender: Male
BayernForum.com fan club: Active member
BayernForum.com donator: Yes

Re: The Statistics Thread

Postby Manchu » Sun Jul 30, 2017 7:12 am

MUTU wrote:Correct me if I'm wrong but I am understanding that the Goalimpact is more of an indispensability measure than anything else. In a team with a bench as strong as the starting XI the Goalimpact ratings would therefore drop, right?

No. In fact, most of the highest rated players play for teams with very good benches.

In order to understand why, let us consider a hypothetical team with three strikers, A and B. A and B are equally good. However, A is favored over B because the coach likes A's beautiful hair more, and he is the first choice starter. When A is playing, the team preforms very well because A is very good. B only plays garbage minutes against worse teams, but the team still preforms very well(which means that the team has a very good goal-difference because it's expected that they would already do very well against bad teams), as well as they would if A played. Thus, both A and B are rated as equally good and given high goalimpacts; the algorithm knows that the strikers are responsible for the good performance because the rest of the team is also being rotated around them, and the good performance continues whatever the rest of the team looks like around them.

There are a couple problems with goalimpact, though. One of them is that when two players are almost always on the field at the same time, it's impossible to distinguish which one is responsible. That's one of the reasons why goalimpact is career based rather than season based because eventually the paths of all players diverge through injuries, transfers, national team play, or retirement.

However, the biggest problem with the algorithm is that it can't account for tactics. It can't tell if a team is preforming well because its players are really good or because the manager has figured out really good tactics, and it can't tell if a player is being played out of position.

Thus, you can see that this method of rating players isn't perfect. However, it is very good at its [ur=http://www.goalimpact.com/blog//2014/03/how-well-does-goalimpact-fare-in.htmll]job of finding undervalued players[/url], I'm not aware of any other statistical method of objectively rating players which delivers comparable or better performance(and to be honest, I don't think it can be done because we have no real understanding of which actions on the field contribute to winning), and I'm inclined to trust it over "popular wisdom" in general because that seems to driven by the bandwagon and flashiness more than anything else.
User avatar
Manchu
I'm a chatterbox!
 
Posts: 323
Joined: Wed Mar 15, 2017 6:34 am
Has thanked: 244 times
Been thanked: 256 times
Gender: Male

Previous

Return to General Bayern Discussion

Who is online

Users browsing this forum: No registered users and 2 guests