Posts:
1,032
Registered:
Dec 16, 2001
From:
Powell, OH
Age:
37 Home page
Set-based ratings formula change
Posted:
Dec 30, 2009, 1:33 PM
awesome, thanks so much partica!
now i just have to find a few minutes to make those updates.
in terms of the ratings formula, it is almost the same formula, it just isn't applied until you win/lose a set, and then the amount that your rating can change is doubled from what it used to be.
it used to be the max change in points for an established player was 32 points for a win/loss of a game. it is now 64 points for a win/loss of a set.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Help/FAQ updates?
Posted:
Dec 30, 2009, 8:03 PM
This seems a little extreme, although I guess it's similar to what used to happen if you had won/lost two individual games in a row. I think I would have left it at 32, but it's probably fine either way.
Posts:
2,220
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Help/FAQ updates?
Posted:
Dec 31, 2009, 2:51 AM
i agree but disagree with both i think, i think a P1 win should be less than 32 points, perhaps somewhere between 18 and 23 points max, and a P2 win should be 32 points max.
why? P1 is easy to win with, P2 is much more difficult.
a won set should yield perhaps 50 to 55 points max.
32 for a set is just not enough, and 64 seems a bit too generous. 50 might be a nice compromise.
you know with the glass ceiling removed now, you might want to prevent inflation by scaling down the max yield. or else we might have the middle class disappear, leaving only reds ranked around 2600 points and climbing, and greens pummeled out of ranking existence. or i may be wrong....
also a solution idea for the green class players, is to give them a annual (yearly) increase of points, like some sort of amnisty.
proposition 1 - all players under 1500 get a 100 points increas.
proposition 2 - all players under 1600 get brought up to a standard 1600.
some players who get hustled in their first 20 games and plummet into ranking disparage there after might really appreciate a second chance after getting to know what they are up against. perhaps some players wouldn't be so ashamed to log in,..
when i log in 90% of players i see on average in the lobby are green or at most 1400. this is Very telling i think.
Or is this just a band aid?
maybe some one else has another idea...
thems my 2 cents on it, thanks for reading.
z
Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
Posts:
2,220
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Help/FAQ updates?
Posted:
Dec 31, 2009, 5:41 AM
hmmm, the more i think about it the more things i see that can back fire in my proposal. what if players didn't want there points changed? what if we give these points and the reds just come and take em all further inflating their rank and sending the 1600s back to green?
and how do i know for sure that set based will create this problem i am predicting? perhaps im prematurely pointing out this theory and its better to wait and see if im right.
as far as the max point limit per set having to do with an inflation problem, would it really make a huge difference if it were 32 or 64 per set? if the sky is the limit with flawless P1 moves and the occasional P2 win, then even with 1 point per set you could still reach 2600 points.
im not sure where im going with this, but just putting points into economic terms for a second i thought might be good to possibly trigger some thoughts from others. and if not, then sorry for wasting your time in reading this.
maybe there is no problem and im just being paranoid hehe.
z
Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
Posts:
1,032
Registered:
Dec 16, 2001
From:
Powell, OH
Age:
37 Home page
Re: Help/FAQ updates?
Posted:
Dec 31, 2009, 3:54 PM
ratings calculations should be simple in my opinion, without the need for different formulas for different ratings or arbitrary year-end point increases, etc.
it is too early to tell anyway what will happen, lets wait and see. i am certainly not an expert on ratings systems so can't predict what will happen nor what the best solution is.
also, remember that 64 points is the MAX possible that the ratings can change. if you are playing another player with the exact same rating as you, the ratings change will be 32 points for the set. the actual amount of points that change is based on the difference between your rating and your opponents rating. Or if you beat an opponent that has a rating ~200 pts less than you, your rating will increase ~15 pts and your opponents rating will decrease ~15 pts.
Posts:
1,032
Registered:
Dec 16, 2001
From:
Powell, OH
Age:
37 Home page
Re: Set-based ratings formula change
Posted:
Dec 31, 2009, 6:14 PM
Thanks for the input, good to know!
The ratings formula is the same as used for chess ratings I believe. Pretty sure I grabbed it from some online chess site when I first started the site.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Help/FAQ updates?
Posted:
Dec 31, 2009, 10:17 PM
> [author=zoeyk]
> i think a P1 win should be less than 32 points, > perhaps somewhere between 18 and 23 points max, > and a P2 win should be 32 points max.
This makes no sense. With set-based ratings there is no longer a concept of a P1 game and a P2 game and separating the ratings for these would totally defeat the purpose. The whole idea is that a SET is a perfectly balanced unbiased entity, there is no advantage to either player. If one player wins a SET, they should be rewarded with a ratings increase determined only by the difference in the ratings of the two players. The inherant unbalance of the game itself no longer comes into play at all -- that problem has been solved.
My thought was to keep the constant K at 32 since that's what used to be used when a match (one game) was completed that resulted in a ratings change. So, it would make sense to continue to use K = 32 for when a match (a set of games) is completed that results in a ratings change. Only the definition of a match that should result in a ratings change has changed.
> a won set should yield perhaps 50 to 55 points max.
The value of this constant is arbitrary and could be anything. Changing it, however, impacts the volatility of each player's rating as they win and lose matches. It also might affect where the ratings end up overall. For example, if we chose K = 5000 it is clear that many ratings would quickly exceed 2100! How much impact might there be from changing from 32 to 64? It's hard to say.
> you know with the glass ceiling removed now, you > might want to prevent inflation by scaling down the > max yield. or else we might have the middle class > disappear, leaving only reds ranked around 2600 > points and climbing, and greens pummeled out of > ranking existence. or i may be wrong....
Ratings simply do not work this way. If you review the formula, the ratings changes are based on the current rating of both players. So, if a 2000 rated player beats a 1600 rated player each player's rating is hardly changed at all. It might be about 1 point (well, everything roughly doubles with the new change from K = 32 to K = 64).
So, while a select few Reds might separate themselves from the pack and achieve ratings of 2600 by beating up on other REDS, not ALL reds will be able to achieve this by always playing against 1600 players because as your rating increases you get less and less benefit from continuing to play against 1600 players until eventually it is practically zero and your rating is pretty much stuck until you get up the courage to start playing against other REDS. The whole point of the system is that people really should be playing matches primarily against players with a similar rating. That's where there's good competition and fun and the rating system rewards victories against opponents with a similar rating. It does not reward lots of victories against players with significantly lesser ratings, BUT does provide huge rewards if you can manage to beat an opponent rating much higher than you, which will be extremely rare in set-based play now that the playing field has been leveled off. Ratings inflation like you are describing might happen to a small extent if players try to abuse the system in this way, but there's a limit to it, and it's not as high as you might think. On the flip side, Green class players will not be hurt by this in the way that you are thinking. First of all, they should be playing against players that are rated similar to themselves anyway. But an occasional match against a red will result in a very minor penalty for the learning experience.
> some players who get hustled in their first 20 games > and plummet into ranking disparage there after might > really appreciate a second chance after getting to > know what they are up against. perhaps some players > wouldn't be so ashamed to log in,..
This thought continues to be brought up and I honestly don't understand why people think this is happening. There are enough beginners that they should be playing against each other anyway and the rating helps to identify each other. When I first joined this site I spent many months at 1000 - 1100 and had a ton of fun playing against many other 1100 players. Increasing your rating comes with improving your game. There is no reason players should be ashamed of their rating and/or refuse to log in because they are ashamed. Contrary to popular belief, most players don't care about their rating at all and use it just as a minor tool to identify other players that would be appropriate competition -- as it should be. I doubt very much that this "problem" is widespread. Lots of people out in the real world will try out a gaming site for a day, be mildly entertained, and then just never log on again -- not for any reason in particular.
> when i log in 90% of players i see on average in the > lobby are green or at most 1400. this is Very telling > i think.
This is how it normally is and has always been. In fact, when I first joined several years ago average ratings were even lower, mostly in the 1100 range. Average ratings probably have increased due to more and more new signups and the provisional rating of 1600. The provisional rating of 1600 is probably the main problem as it is totally inaccurate for most new players and represents a relatively advanced player. The idea is supposed to be that this provisional rating adjusts itself quickly and by the time they have played 20 rated matches it becomes pretty accurate. But for most new players this means taking a nosedive from that original 1600 -- people don't realize that this provisional 1600 was never really their true rating in the first place -- it's more of a placeholder that adjusts wildly for the first 20 matches until your true rating can be determined. Maybe it would be better if this provisional rating started at 1000, who knows.
Re: Help/FAQ updates?
Posted:
Jan 2, 2010, 9:20 PM
It may become important at some point in the future to decrease the K values at at least one (perhaps more) levels as theskald mentions happens in chess. Perhaps a single decrease at a 2200-2300 figure would be sufficient, perhaps a couple would be needed. This helps to make the top ratings more accurate and stable- otherwise ratings may have a tendency to continue to climb based primarily on the number of games one plays- which leaves the good newcomers forever playing catch up.
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
