EPTAlso, I don't think any of the higher-ups would care enough to change the ranking system. After all, the BW server only has 600 - 700 players on a good day. And even in the best case scenario, a new ranking system won't change that by a noticeable amount.
iCCup rating system discussion - Page 2
| Forum Index > BW General |
|
Holy Check
Romania155 Posts
Also, I don't think any of the higher-ups would care enough to change the ranking system. After all, the BW server only has 600 - 700 players on a good day. And even in the best case scenario, a new ranking system won't change that by a noticeable amount. | ||
|
Black[CAT]
Malaysia2589 Posts
Iccup must implement this for evryones benefit. | ||
|
Chef
10810 Posts
| ||
jhNz
Germany2762 Posts
But I totally see your point. Iccup should offer the the opportunity to start at a higher rank after you finished pretty good in a previous season. Maybe a 2 rank penalty or something like that? Or some kind of "placement time" like you get your old rank from last season again, or maybe with a small penalty. And if you win a certain amount of your first few matches you'll keep it, otherwise you'll be demoted. You'd have to play players of the same rank of course. Don't know how it should look like in detail but I completely agree with you. | ||
|
gutshot
United States429 Posts
but I doubt any of the higher ups would even read this or they actually care about iCCup bw. | ||
|
ninazerg
United States7291 Posts
| ||
|
Phrujbaz
Netherlands512 Posts
0. Rating systems not based in statistics: Current ICCup system is an example. Pros: * Can use the rating system to motivate people, eg to play many games or to prefer certain maps. Cons: * Rating correlates very weakly with playing strength. 1. First generation rating systems: ELO. Pros: * Simple and easy to understand. * Widely used and widely known. * Easy to implement algorithm Disadvantages: * Cannot distinguish between uncertain rank and solidly established rank (does not measure uncertainty). * Tendency for players to want to play against stronger players. * Disincentivizes playing against new players. * Takes a long time to settle on your rank. * If you suddenly improve in skill the system or don't play for a while, the system won't pick your new rating up quickly. * Not very accurate: you can swing pretty far around your 'true' rank. 2. Second generation TrueSkill, Glicko Pros: * System measures uncertainty in the rating, so a lot of the artifacts in the first generation rating systems disappear. * Still relatively easy to implement. * Still relatively easy to understand. * Reasonably accurate Cons: * Takes relatively long to converge on your true rank. * Not very responsive to rapid changes in playing strength. 3. Third generation Decayed history algorithms. The go server KGS has used a decayed history algorithm with excellent results for some time. http://www.gokgs.com/help/rmath.html The basic idea is that recent results are weighed more strongly in the rating calculation than older results. Pros: * Quick to converge * Accurate * Responsive to changes in playing strength Cons: * Not widely known / not many standard implementations available * Tendency for frequent players to get 'stuck': if the rating system thinks it has measured your rank very accurately, you can get stuck and not gain any points even though you're improving. You can only get unstuck by not playing for a while. 4. Fourth generation Whole history rating. http://remi.coulom.free.fr/WHR/ Unlike other systems does not perform incremental calculations but directly computes over the entire history of a player. Pros: * Quick to converge * Very accurate * Responsive to changes in playing strength * Players don't get stuck if they play frequently Cons: * Not widely known / not many standard implementations available * Slightly more computationally expensive than other systems * Has parameters that need to be trained for the system to work optimally. | ||
|
[sc1f]eonzerg
Belgium7012 Posts
On October 31 2011 02:58 ninazerg wrote: Korean players want to play on the servers where the pros or retired pros are playing. Boxer, Nada, Bisu, Hyuk, Jaedong, Sea all played on iCCup at one time, just to name a few. They have a korean-friendly server now with progamers on it. Why on earth would any of you think changing the ladder system would 'attract koreans'? agree, the only way to kor back is fish and brain down | ||
|
medic_ro
Romania105 Posts
| ||
|
zimp
Hungary959 Posts
with iccup system, the majority, the lower level players have some kind of satisfaction of reaching something, which encourages them to play more. elo system is only good if opponents are being set vs each other, not when you pick whoever you want to play against. at fish server there is the elo system and i can tell you i dont really like it. it distinguishes players like this: bad | rest | good | very good | very very good where the starting level compared to iccup is like high C. since you get the same amount of points for a win and for a loss, and most players only want to play vs someone who has the same points as they have, like 80% of the players are stuck at ~1000 points and there are players with like 100 win and 100 loss having the same amount of points as someone who just started there. and it feels crap to play a lot of games, win half of them, and see that u didnt make a step ahead. that makes me wanna stop playing. or do you think its normal that Sziky has only 1282 points on fish with 75 win and 59 loss? actually it might be better for good players to have a system where they get to their level with less noob bashing, but youve got to create a system that ensures everyone good playing conditions. and i think iccup is good in that. i mean come on, if it was so bad to play 50 games to get on your level, progamers wouldnt have played on iccup. but they did the only problem now is that there are too few players. changing the system won't really help that... its all bad since ladder a-b channel became empty;; | ||
|
playa
United States1284 Posts
High ranks should mean something, and MOTW simply inflates ranks. And since Korean pro gamers aren't even playing on iccup at the moment, it's just annoying and an inconvenience that makes me more inclined to play Korean ladders if I want to play at a high level. | ||
|
Nakata
Bulgaria67 Posts
| ||
|
fabiano
Brazil4644 Posts
On October 31 2011 04:37 Phrujbaz wrote: (...) 3. Third generation Decayed history algorithms. The go server KGS has used a decayed history algorithm with excellent results for some time. http://www.gokgs.com/help/rmath.html The basic idea is that recent results are weighed more strongly in the rating calculation than older results. Pros: * Quick to converge * Accurate * Responsive to changes in playing strength Cons: * Not widely known / not many standard implementations available * Tendency for frequent players to get 'stuck': if the rating system thinks it has measured your rank very accurately, you can get stuck and not gain any points even though you're improving. You can only get unstuck by not playing for a while. Nice summary. But please do not go with that KGS ranking system, it is horrible. I got stuck at 3D with multiple accounts and it sucks. There is a dude there, TheCaptain if I'm not mistaken, poor guy will be forever 4D.... Maybe at higher levels the current ranking system is inconvenient, but for me, a D player, the ranking is just fine. | ||
|
BobTheBuilder1377
Somalia335 Posts
| ||
|
sataNik[pG]
Greece724 Posts
gain points even at a 30% win ratio. How accurate is that? To gain points when you should actually be losing. The rating system is supposed to try to be indicative of a players strength but here it seems to me that it just gives ways to people to look better than they are. I don't understand what is the problem of having a low rating if that means that you will play vs players who are a little better or a little worse than you. It will help you become better sooner. The game is fun only when the outcome is unclear, not when the whole game is just another routine. | ||
|
Sawamura
Malaysia7602 Posts
| ||
| ||
