Forum


rating system
AleaIactaSunt wrote
at 8:45 PM, Friday January 5, 2007 EST
As I have understood so far our system is based on ELO. It's a rather complicated formula that basically gives you points or takes it according to your expectation to win (your ELO/ranking score compared to the enemys' scores).

My questions:

1. ELO was made for 1 vs 1 fights. What score are you compared to? An average of the 6 opponents' scores?

2. That, if we assume the score represents skill, might be unfair in case there are two with a slightly higher ranking and 4 a lot below?
-> ELO gives you 1 multiplied by sth that resembles the score comparison for a win. 1/2 * CompareValue (CV) for an equal. So in my example would you get 1/2 for being 3rd since 2 are better... How does it work?

3. Wkipedia mentions a problem: points will in average of all players go higher for certain people if they play a lot. Let's say we have the over all kdice champion who will beat everyone in the world (let's name him Ryan). He plays a few games and gets a 2000 ranking. Now two less talented players (let's say "alea" and "Jaits") play a hundred games against each other: Since alea sucks he has sth like 1500 afterwards, but Jaits gets 2200, which doesn't reflect his actual skill compared to Ryan.
-> So I see (real) Ryan's point that people should play about the same number of games. (example) alea will be pissed otherwise because with stopping to play at 2200, Jaits drains points from the whole system making it harder for alea (having absolved a kdice crash-course now) to get up in ranking.

4. What can we do for "(example) Jaits" or how are the table limits involved in the system?
Without limits we had everyone having pretty much the same score probably - as soon as you got ahead with points, a loose would be a big one (1400 vs 2100: just imagine). So the limits help to maintain scores - In a game of luck (!) important to prevent massive frustration.
-> If the limits would not depend on the number of players with certain rating, we had inflation. It's not to hard to stay at a certain level, once you crossed a "table limit safety mark". More and more people climb up and then: Complain, because there are no safety marks anymore (if there were, inflation): In a game of luck that means frustration, because they go down again, if they keep playing (down to the average of all above the last safety mark).
-> If they just don't play: see above.

5. Special problem: gaining 2200+ was for the lots only possible with the old system. With some luck you could get that high from 2000 (I'm impressed by the ones making it with the new system). So we took our score away from the whole system, let therefore sink the table limits to 1900 and perfectly protected ourselves at the top. I kinda like that, especially now that I understood (or believe I did). But it's unfair. If the high ranks start playing, give some of their points to the 1900s, there might be 2000 tables again at some point I believe. And the race for 1st is open again.

6. I do NOT understand, why that -50 points was helpfull except waking us up to the problem? There were even more points taken from the system? You could've said sth Ryan, instead of making drastic moves?

7. Anyways, I'm content with everything now and impressed by Ryan who has understood that whole thing even better than me and inventing all that.

* Almost all names used are invented and have nothing to do with any existing people.

Replies 1 - 10 of 13 Next › Last »
fuzzycat wrote
at 2:49 PM, Saturday January 6, 2007 EST
points are tremendously overrated ;o)
aliaiactasunt wrote
at 3:45 PM, Saturday January 6, 2007 EST
At least somebody answered :D.
Odlanor wrote
at 4:27 PM, Saturday January 6, 2007 EST
Should just get rid of the ratings altogether, instead show # 1st place finishes and % of wins out of all games played.
Lindsay wrote
at 4:42 PM, Saturday January 6, 2007 EST
No
fuzzycat wrote
at 6:03 PM, Saturday January 6, 2007 EST
Lindsay, you are so an extremely positive guy. Especially I like that you reason your expressions.
somnambulisa wrote
at 6:08 PM, Saturday January 6, 2007 EST
I'm not a guy.
AleaIactaSunt wrote
at 6:10 PM, Sunday January 7, 2007 EST
In the middle of the night a non-reflected idea I just got from Odlanor, maybe worth discussing: How about having ratin based on the percentage of wins: 7 times #1 percentage, 6 times #2, etc.. the rankings would only differ between 100 and 700. maybe you make it 14 times #1, etc (1400 - 200)? Meanings besides lindsay's "no"?
JKD wrote
at 6:43 PM, Sunday January 7, 2007 EST
With your suggestion then beating lousy players is just as good as the "best" players. Plus it would take forever to become high ranked if you had a lot of losses in the past. Currently I believe getting first is already seven times better than getting seventh so I don't understand why you would want this to change?
TheGrid wrote
at 2:01 AM, Monday January 8, 2007 EST
Your proposal still has the Problem: finishing e.g. 4h is not by been any better than 7th. Even worse, inactive undangerous players often get higher places...
pTm wrote
at 10:10 AM, Monday January 8, 2007 EST
Ok Alea, I have read all of your questions and I try to give you some answers:

1.Yes (I think so because I saw the formula Ryan uses for calculating the ratings in the forum)

2.Well I dont know how it works but I know that it is not the original ELO-system

3.The 1000-games-problem only appears in a game with two players. It can only happen in chess were it is possible that one player wins 80% of the game or more. The luck-factor in kdice is too big to allow this.
"Jaits drains points from the whole system"
--> There are always new players who lose a few games and stop playing. They bring more points to the system.

4.I only understand the first part. Please explain "inflation"

5.Good idea but a) most of the top players won't do it. b) when someone gets a rating of 2200+ on the 2000 table he will stop playing and we have the same problem again. I have an idea how to solve it and I will post it in the forum later....

6.I already explained that at #3. You can look it up in the german wikipedia:
http://de.wikipedia.org/wiki/Elo-Zahl
KDice - Multiplayer Dice War
KDice is a multiplayer strategy online game played in monthly competitions. It's like Risk. The goal is to win every territory on the map.
CREATED BY RYAN © 2006
RECOMMEND
GAMES
GPokr
Texas Holdem Poker
KDice
Online Strategy
XSketch
Online Pictionary