New Statistic

It might be helpful if everyone's profile also had their average rating...it's a bit of a shock to find out that the 106-rated player you went after for easy points is really just last year's (inactive of late) answer to Tebb 😉. Seriously, though, it might be an interesting statistic, and I don't think too hard to implement, as all of the necessary data is in the Rating/Time graphs anyway.

I agree! It would be useful to know if you are playing above or below your average, for others and yourself as well.

Marc

This idea was brought about in my games with a very good player, whose rating is similar to mine now. Since mine is currently in a slump, I am normally quite confident against ~1300 players. Only when he busted out the lead-filled snowshoe in 3 of our 4 games did I look at the graph and see that he seems to be about a 1600 player normally. Hehe 😉. I'm learning a bit though, and quite enjoying these games.

Originally posted by royalchicken
It might be helpful if everyone's profile also had their average rating...it's a bit of a shock to find out that the 106-rated player you went after for easy points is really just last year's (inactive of late) answer to Tebb 😉. Seriously, though, it might be an interesting statistic, and I don't think too hard to implement, as all of the necessary data is in the Rating/Time graphs anyway.

You can do that now by checking a player's "rating over time" chart. A quick scan of yours revealed a 1422 high rating.

Yes, but what's the average?

Originally posted by royalchicken
Yes, but what's the average?

The rating itself is a smoothed and scaled average of the latest results... I'm not sure that an "average smoothed average" would give you numbers that you could conclude anything much from.

That is exactly what I am talking about. A smoothed and scaled average of the results of the past few games is not always a good indicator of the player's strength. The smoothed an scaled average of scores or hundreds of games is.

Originally posted by royalchicken
That is exactly what I am talking about. A smoothed and scaled average of the results of the past few games is not always a good indicator of the player's strength. The smoothed an scaled average of scores or hundreds of games is.

I would contend that if you judge the goodness of a rating by its ability to predict the outcome of games before they are played, then the current rating would come out as a pretty good indicator and the average rating over 100 games as a rather worse one.

Originally posted by iamatiger
I would contend that if you judge the goodness of a rating by its ability to predict the outcome of games before they are played, then the current rating would come out as a pretty good indicator and the average rating over 100 games as a rather worse one.

I think you're both right to an extent - a distinction though helps.

Where the player involved isn't particularly a frequent mover I'd agree with the tiger. The current rating is probably a better indicator (assuming sufficient number of games played).

I would say though that the Chickman has a point when it comes to the more frequent players. Regular movers with a certain rating (ie not 1800+ where it's hard to come by points) tend to have rather volatile ratings, whereby the current rating mightn't be wholly representative of their true strength, and that an average over 100 games might be more accurate.

Personally I'd be interested in turning the whole thing on its head, and would quite like to see in my profile some sort of average rating of the last 100 opponents I've played for example. There are some relatively very strong people that I've seen below 1400 (Tim Robinson and hypermo2001 for instance), resulting I suspect from a preference to play strong, fairly highly rated opponents.

T1000

Quite a nice analysis as always. Maybe I should take the picture out of my profile as long as I'm "Chickman" 😉.

Originally posted by iamatiger
I would contend that if you judge the goodness of a rating by its ability to predict the outcome of games before they are played, then the current rating would come out as a pretty good indicator and the average rating over 100 games as a rather worse one.

No, because there are too many variables. The biggest factor (as T1000 has mentioned before) is the number of games you've got going. For example, I play reasonably well with fewer than about 10 games, but any more than that and I'm garbage....some people have much higher "crap thresholds" than this. Latex Bishop, for example, has me pretty well beaten in two games right now, but he's got over 60 games going 🙄

Originally posted by royalchicken
It might be helpful if everyone's profile also had their average rating...it's a bit of a shock to find out that the 106-rated player you went after for easy points is really just last year's (inactive of late) answer to Tebb 😉. Seriously, though, it might be an interesting statistic, and I don't think too hard to implement, as all of the necessary data is in the Rating/Time graphs anyway.

Great idea. I was about to start a game against mmanuel when I realized he said on the forums he'd been at university for a year and was timed out.