Players by average rating

[ariotofmyown]

The people who run this site have done well to provoke debate by creating automated users who can keep threads going with seemingly crazy talk. I used to believe that 'bris vegas' was a real person, such was the exotic backstory, but as it's arguments become so contradictory and insane most people have realised the truth. Which brings us to 'albert'.

Regardless of whether it was supposed to be a happy clapper or a wrist slasher, it clearly seems that project albert has become self aware. No human could churn out the stats like we've seen over the last few weeks and it's fun too see some of the earlier creations struggle so much with numbers.

Cheers dcfcfan overlords for giving us this new statistical approach and i hope u can still control it. Makes a change from reading "sammon is terrible" five mins after full time regardless of the result.

[RamNut]

pFhvO

I'm sure that this means something .

I've no idea what, but it reduces me to fits of giggles.

Long live Albert.

[maydrakin]

Personally, I love poring through stats. One has to be mindful that they don't paint the whole picture, nor do they offer nothing.

The variables such as which teams did each player play against, the form of those teams at the time and which players were in the side with them at the time can't be factored.

Plus there are elements such as the Atwell factor which will have cost/given points in different places.

One just has to review McEverley's career trajectory to appreciate he wasn't our best defender.

Nevertheless, Albert's threads are always nourishing, so keep the stats coming...

[Albert]

Personally, I love poring through stats. One has to be mindful that they don't paint the whole picture, nor do they offer nothing.

The variables such as which teams did each player play against, the form of those teams at the time and which players were in the side with them at the time can't be factored.

Plus there are elements such as the Atwell factor which will have cost/given points in different places.

One just has to review McEverley's career trajectory to appreciate he wasn't our best defender.

Nevertheless, Albert's threads are always nourishing, so keep the stats coming...

The key point is that for a player with sufficiently large numbers of games, factors such as the opposition, which players they play with etc. should cancel themselves out. That's why I included the standard errors, although they are almost certainly a underestimation for players with smaller numbers of games. Of course you need to know what you're looking at, but ultimately there is a good argument for a cutoff of a certain error margin rather than a certain number of games. For example, the error margin could be done as a percentage of the value or simply have a higher minimum number of starts cutoff. But yes, you're exactly right, for smaller numbers of games such unwanted factors could distort things quite drastically. Here's something interesting:

FJJvI

This is a plot of points per game average against starts, with error bars included. Notice how for players with larger numbers of appearances they tend to jump around a lot less. Also notice how it's just basically a soup as you head left.

A second, much more annoying point is that the change of seasons can bring new players and so a sudden jump (for better or worse) for the team as a whole, this could also skew the values drastically. Barker's points per game (as with others) was heavily influenced by that late half of the 2010-11 season, whilst new players this season will have a huge advantage over those in years before.

Long story short, it's not the best system in the world, in fact it's not that great at all, but it does show something in a comparative sense if taken in context of what it is.

One last thing, consider the McEveley zone, that is, the area that error bars cover, notice how many players' averages are contained within it:

g3syw

Also, for those wondering who Mr. Fantastic over there on the 10 start line that has an average ppg of 1.8 is, it's Kilbane, another artefact of playing minimal games.

[Albert]

Yes, but I’m confused what that something could be.

It’s a team game and I’m not sure what valid conclusions can be drawn by rating individuals according to the team’s results.

For example, Leacock and Barker are similar according to these numbers. However most of us know that Leacock was OK and held his own for a brief spell in a decent, but very lucky, side. However Barker in his first season was immense and, despite being surrounded by utter dross, probably kept us in this division.

Or am I Brissing the point?

As above, over sufficiently long periods of time, assuming that every player isn't playing every single game, they'll have a mix of good and bad around, good and bad form, good and bad opposition and it should even itself out. To be honest, 100 odd games isn't even that much in this sense though. It can give an indication of players who the team plays better with, but ultimately it has some limitations, the largest of which is that the dataset just isn't big enough.

[RamNut]

So.....

1. Any player-rating is 90% determined by who else was in the team.

2. No player / team achieved more than 1.5 points per game.

3. Therefore no-one played in a team than was any more than average.

4. Therefore we've been a bit crap for quite a while now.

There is an interesting table of stats published which shows team ratings rather than player ratings and captures the relative merits of the all the different teams put together.

Its called the league table and it seems to be generally accepted as the definative guide.

However any good system should allow us to make verifiable predictions, therefore we could alternatively deduce from the above player ratings that......

1. the best defenders always have first names that start with a J.

2. The best midfielders tend to have six letters in their surnames +/- 1

3. The best strikers tend to have lots of o's in their names - the exception being Alberto Bueno who - despite having two o's - seems to have under achieved.

On that basis we can predict that....

Jake Buxton and James O'Connor should both be decent defenders.

Buxton might yet make an effective midfielder and O'Connor might yet be surprisingly good up front.

Conor Sammon is likely to be our best striker at the moment.

Callum Ball will never achieve anything.

Hughes is likely to be a very good midfielder.

James Bailey might yet be a good defender

Jamie Ward is too short to play in midfield, and is even less effective up front, but might make a good full back.

Etc.

Seems a pretty good system to me.

It definately tells us something.