Rating drift on LG

Hi fellows,

I took a look at the Street Soccer ratings today and was a little bit surprises, how many people have a rating over 1500 compared to those below this value.

Then I copied the list of players and their ratings and got a sum of 860166 points for 522 players, which gives an average of approximately 1648 rating points.

Being surprised, I checked the values for WYPS Italian as well (less players = less to copy) and found 35669 rating points for 23 players, which gives an average of 1551 points. The same for WYPS Polish gives an average of 1512.

Where does this deviation from 1500 come from?

Maybe you win more points on a win than you lose on a defeat, on average.

I also think is has something to do with the new players. They win points faster, to reach their real rating more quickly.

Crelo, I think that was changed long ago. Now, the winner gains as many points as the loser loses. But this deviation COULD be aftershocks of this earlier rule, I suppose. But it will not cause further drift.

Does the list of players include everyone who has ever played a game here, or are people who have quit the site excluded? That could also explain the daviation from 1500.

Marius, you think players with lower ratings are much prone to leave LG? So not taking them into account we have a list of winners :-)

I think so, yes. However, SOME lists here contain every player who has ever played, I think. But then again, some do not. So I suppose this effect depends on what list has been used.

Yup, biased, and self-selecting sample.

How to lie with statistics, page 1.

I took the "List of players" from the main page and chose a particular game: Street Soccer, WYPS Italian and WYPS Polish. From each of the pages, that data was copied to Excel. This took a while, but changes during data collection can not explain the observation.

I will check in a few months, when more Street Soccer games are played. When the excess points stay the same (total rating points - #players*1500), then it is due to historical reasons.

I took a look at those lists, and it seems that it doesn't show every player who ever played here. I thought it did, though, but I miss several names. For instance, I can no longer find wccanard among the dots players.

This strenghtens my belief that "old" players are eventually removed from the lists, and that such players more often have below 1500 ratings than above 1500 ratings, and that this is why you observe the reported effect.

This is my impression, too.

I think tasuki's list has all the players. Perhaps we can convince tasuki to make one more page with statistics per game (average rating, deviation, etc.)

Yeah, tasuki has the full list! If you include players like Hot_Sync_Karen http://www.littlegolem.net/jsp/info/player.jsp?plid=2788, I'm sure the average rating will get closer to 1500 :-)

Yup, one wcanard leaving removes 500-1000 points per discipline he excels in.
One Hot_Sync_Karen departing adds 200-500 points per discipline she was signed up for in quantity.

Statistically, one would expect fewer to reach the peak and leave, as fewer will reach the peak. So rating inflation is what I would expect.

WCcanard might be reïncarnated? I mean, Lovebaby hasn't been around heer for ages, too. (He's now better known as slaapgraag)

Oh, I SHOULD read more carefully before reacting...

Are you sure this does not follow the skill inflation of LG players?

After all these points rising the level around 1550 or 1600 were earned by the mean players...

I think globally LG population is ageing as from the definition, this occurs when the median age of a country or region rises due to rising life expectancy and/or declining birth rates (here being new player arrival). Are older player more skilled than newbies or are we already less efficient due to early degenerescence factors?

Ah, Hot_Sync_Karen, such memories!

Maybe this has been discussed before:
If two players of the same rating play two games and split the games (each winning one) then the first person to resign would come out on top of the other player after he resigns. Correct?

p.s. Caroll: I have only been playing 10 months but I am also 67 years old. Does that make me a newbie or an older player?

Yes, mtbikesman, that is correct. See here . If you lose first then win in your scenario, your rating will go up by roughly 1.5 points. If you win first it will go down by 1.5 points. In general the ELO system assigns more weight to recent games, so it's better to lose early and win late, other things being equal.

@mtbikesman, you are clearly a newbie and a young man!