The new ratings and statistics stuff I mentioned here is now live on the system!
The new ratings use the Elo system. This is a system devised by physicist and Chess master Professor Arpad Elo ("ee-low") something like 30 years ago, which has served well for ratings in such diverse games as Chess and Croquet (for instance, it's used by both the Fédération Internationale des Échecs [World Chess Federation] and the World Croquet Federation [yes, there really is one]). Apparently Professor Elo was a statistician as well as a physicist and spent 10 years refining and proving the system, and it's well-regarded in statistics circles as well as game circles.
In the change from the old system to the new one, naturally the "Top Players" list has changed. The old system gave higher ratings to people who played lots of games; the Elo system is less susceptible to that effect (its accuracy improves the more games you play, but your rating doesn't necessarily go up as a result). More importantly, though, the new system takes the relative strengths of the two players into account when updating the ratings at the end of the game. This is the core of the success of the Elo system. Our old system didn't do that.
To set up the new ratings, we started everyone at a certain level, and then we ran through all of the games played here, applying Elo's formulae and updating each player's rating as of the end of each game they played. This makes the numbers just like they would have been if we'd been using Elo ratings in the first place. And of course, as you play games, the new system is used to update your rating.
Note that we start all players at the same level (1600). Even if you have an Elo rating somewhere else, it doesn't apply here at Pocket-Monkey. The primary reason for this is that Elo ratings are not absolute, they're valid within a body of players, so of course your rating somewhere else doesn't apply to the body of players here at Pocket-Monkey. Even if you have a FIDE-approved Elo rating for Chess, it won't apply here as you mostly won't be playing FIDE players here.
Some other notes:
There's a new "Statistics Help" page on the site; it's a link off the main Help page.
By popular request, in addition to a player's Elo rating we also show their "simple win ratio". (Note, though, that the Elo rating is a better indicator of the player's demonstrated skill at a game on the site.)
It's entirely possible that a player can have a higher Elo rating than another player, but a lower "simple win ratio." That's because the Elo system takes the relative strength of the players in a particular game into account when adjusting the ratings at the end of the game, so someone who plays only relatively "weak" players won't have as high a rating as someone who plays relatively "strong" players, even if they win more often.
We no longer show statistics for Tic-Tac-Toe; see the Statistics Help for details.
We now count expired games as a loss for the person who allowed them to expire, rather than as a draw.
Elo ratings are ratings, not rankings. What's the difference? The short answer is that if they were rankings, two players couldn't have the same number. ;-)
Great! Now the site is truly competitive! This is a milestone in the history of the site. The site is really going to be very popular now and it membership will skyrocket. Wish you all the success :-0)
The "Simple Win %" column is the simple win ratio (percentage) I mentioned in my message, literally just the percentage of their games that the player has won. So if the player wins 4 out of 5 games, it'll be 80.00%; if they win 1 out of 5 games, it'll be 20.00%. Simple win ratios are interesting, but the Elo rating gives a better indication of the strength of the player. -- T.J. Crowder First Primate Pocket-Monkey.com
So, does this mean that if a game is a draw, both players will have their simple win score lowered?
You know, I guess it does, since a "win ratio" wouldn't include a draw, since it wasn't a win. I had to think about that for a minute, about whether draws should be omitted, but I think not. Similarly, draws would lower a "loss ratio" if one were to calculate one, so "win ratio" + "loss ratio" wouldn't equal 100% if there were any draws. Good point.
Another reason for looking at the Elo ratings instead of simple percentages! :-) -- T.J. Crowder First Primate Pocket-Monkey.com
posted by PANTERACHIK (panterachik~stacy) on 01/30 at 12:05
i must admit that im still a little confused by the new elo ratings,but im glad everythings coming together on the site!the thing im really happy about is the expired games policy.9 times out of 10 the person who let the game expire would have lost anyway.and i suppose n/a was better than a loss...lol.once tournaments are in place there wont be a reason to play at other sites :-)good job!!
Implementing the new ratings system was an outstanding move; good work Jock and TJ and all others who were involved. This site is already a favorite of many who play here and it just keeps getting better. Good going!!
Nicely done and an equitable way to handle the difficult problem of rating games and matches. My hat's off to you!! Elo ratings properly applied leave little room for grousing or complaining about ratings, and makes it less possible to pad one's ratings by playing consistently weaker opponents. Very good work, Jock and TJ!!
At the beginning of the game either player can have a rating that’s different from the same player’s rating right before the game is over. Which is taken into account when calculating the final Elo standings for both players?
Thanks! Yes, we're working on tournaments now. We were interrupted there by a couple of things, which put our timeframe back a bit, but we're back on them now. Watch This Space. :-) -- T.J. Crowder First Primate Pocket-Monkey.com
Forum
software by
Crowder Software Pocket-Monkey and the Pocket-Monkey logo are trademarks of T.J. Crowder and Jock Murphy. All other trademarks are the property of their respective owners.
