It can also have to do with the size of the games you play. If you play nothing but 1v1s and win 80% of the time, your ranking will be astronomical but your G will still be quite low.

That doesn't follow.  Winning a 10 player game gives (almost) identical rating points as winning 10 1v1 games against the same players,  and also provides the same G Rating.

Maybe I misunderstand. If a player plays and wins one 10-player game, their G Rating is 9.00.

If a player plays and wins 9 (or ten, or fifty) two-player games, their G Rating is 2.00.

No?

I do believe asm is correct. (Or at least more correct, I get easily confused when attempting to think/calculate G rating.)

G rating sounds like a pretty useless statistic then.

asm wrote:

Maybe I misunderstand. If a player plays and wins one 10-player game, their G Rating is 9.00.

If a player plays and wins 9 (or ten, or fifty) two-player games, their G Rating is 2.00.

No?

It would actually be 10.00 for the 10-player game.  But yes, the G rating for a particular game caps out at the number of players for that game, so someone who plays only two player boards will never have a G rating above 2.00.

Are you sure about the 10? From the looking around I did it seems that you get the G rating for the number of players defeated, not the number in the game. Although that doesn't make sense for 2 player games, does it. Hmm.

It should be as I defined it earlier (# of players (including self) in games you won) / (# of games played) which is equivalent to Yertle's definition. It does have a meaning and you can see that most players have a G-rating of less than 2 even though they play all different types of boards with differing number of players.

A G-rating of over 1.5 certainly says that you are a great player (provided that you have played sufficiently many games). A G-rating of less than 1 says that you are winning less than is expected.

As Yertle said earlier, A g-rating of 1 means that you exactly as many games as is expected by probability (50% of 2 player games, 33% of 3, ...).

There is of course some skewing that occurs from playing large player games, but this becomes minimal as you play more games. Certainly, as a board specific stat it has meaning even with only a few games played.

In my opinion, it is a much more meaniful stat than win percentage and is really trying to measure the same thing.

It's a useful stat to an extent, but you have to compare it against the other stats for it to have much meaning.

First, it's easily skewed if you haven't played many games. A person can come in and play 1 ten player game, win, and now have a g rating of 10. But if they continue to play games, you can pretty much guarantee it's not going to stay there. It will average itself out over time.

BD was also onto something in that you need to look at the users global ranking. If you beat a lot of people with low global ranking (compared to yours) you will have a high g-rating but a lower global ranking. If you beat a lot of people with high global ratings, you'll have a high g-rating and a higher global ranking.

Then Yertle pointed out that you need to look at what type of games a person plays. A person who only plays 2 player games will be partially capping themselves. This again, evens out for people who play all kinds of maps.

I know there are more things, but I think Alpha had it right. It a more meaningful stat than win percentage, but it's affected by to many things to be the most important.

To the point of the original post, losing a ton of points once your rating gets really high. I'm still quite a ways off from BD's 2500+ but I can feel it. I've won a 5 player game and only collected 40 points or so to then go on and lose a 6 player game and lose more than that. I'm wondering if instead of a straight percentage (mine/theirs)*20 if something a little more scalable would work.

Maybe something along the lines of (mine-theirs)/1000*5+20. It would keep the extremes from being as high of a point swing.

Another thought would be a cap. So you can never win more than X or win less than Y. Just a thought.

Alpha wrote: 
There is of course some skewing that occurs from playing large player games, but this becomes minimal as you play more games. Certainly, as a board specific stat it has meaning even with only a few games played.

In my opinion, it is a much more meaniful stat than win percentage and is really trying to measure the same thing.

G rating is trying very hard to do something, but I don't think there is a noisier stat on the site.  As stated, the original thread had little to do with this, so I'll start a new one, and hopefully you will see that it is more skewed and less meaningful than we would like.

Seige07 wrote: To the point of the original post, losing a ton of points once your rating gets really high. I'm still quite a ways off from BD's 2500+ but I can feel it. I've won a 5 player game and only collected 40 points or so to then go on and lose a 6 player game and lose more than that. 

I don't think so. Let's say everyone is a 1000 kind of player going into a game. When you win a five player game you take home 20 points x 5 people = 100.  When you lose in a game, it doesn't matter how many people are playing.  You only lose 20 points.  I'm pretty sure that's how it works.

M57 wrote:

Seige07 wrote: To the point of the original post, losing a ton of points once your rating gets really high. I'm still quite a ways off from BD's 2500+ but I can feel it. I've won a 5 player game and only collected 40 points or so to then go on and lose a 6 player game and lose more than that. 

I don't think so. Let's say everyone is a 1000 kind of player going into a game. When you win a five player game you take home 20 points x 5 people = 100.  When you lose in a game, it doesn't matter how many people are playing.  You only lose 20 points.  I'm pretty sure that's how it works.

M57 is right, but Seige07 is valid too.  If you have a high score on a map (or global ranking) and you win a game against 4 others that have a significantly lower score then it is possible to only grab 40ish points.  But if you lose a game to a person with a lower score (like 700) and you have a higher score (like 1500), then it is possible to lose more than you would have gained total (lose like 42ish).

It almost seems like there is a perception that there is no limit to how many points you can accumulate. I don't see it that way. I see it as not unlike a chess rating.. 3000 just may not be reasonably attainable.

My perception is that it's not about accumulation. Rather, it's about relative playing strength.

M57 wrote: It almost seems like there is a perception that there is no limit to how many points you can accumulate.

You can definitely get up to a 20,000 score :P  If a 20,000 player beats a 1,000 player then the 20,000 player moves up to 20,001 and the 1,000 drops to 999

Yertle, you're a funny uhmm.. -- ..turtle.

Of course, we know the reality of the situation is that luck of the dice and the quality of the competition should keep even a phenomenally good player's rating in check.

And what happens when that 20,000 loses to the 1000? Can you say "minus 400 points"?

Yertle wrote:

M57 wrote: It almost seems like there is a perception that there is no limit to how many points you can accumulate.

You can definitely get up to a 20,000 score :P  If a 20,000 player beats a 1,000 player then the 20,000 player moves up to 20,001 and the 1,000 drops to 999

And if that same 20,000 rating player loses say a 10 player game to that same player. Even though he may have had no control over it because of the number of players, he's now going to lose 400 points.

Dang, M57 beat me to it.

The maximum loss is actually capped at 100 points.

I didn't know that! In that case, I will begin my climb to 20000 ranking points.

tom wrote: The maximum loss is actually capped at 100 points.

Oo... I don't know that I like this.. Parity is lost at the top end.  Good players who play long enough can just rack up the points.  I take it maximum gain is capped at 100 also?  At 20,000, all you get is 1 point per win (against 1000 kind of players), but you could play only good players, say someone in the 5,000 to 10,000 range to make it worth your while.

That's true... but it's insanely hard to get a score of 20,000.

In fact I don't think I've never seen a score much higher than around 2,500.