I’ve noticed that tournament games affect my overall rank. I don’t think this is a good idea. Players of different ranks will be playing against each other without handicap, so it doesn’t give much incentive for dan players to join with nothing to gain and everything to lose. It doesn’t say in the tournament room that the tournament is ranked either, but I believe the 4 running now are ranked? Should have separate tournament rankings imo. I also believe there should be a scheduled handicap tournament.
Also, it would be nice with separate 9x9 and 13x13 rankings. In any case these board sizes shouldn’t affect the 19x19 ranking as they are completely different type of games. When we say someone is 1 dan, we mean 19x19, not 13x13 or 9x9. It’s also much easier to lose to a weaker player in 9x9 than 19x19.
No, you say he’s 1 dan in 9x9. Some players who are good at 19x19 can be bad at 9x9 and vice versa, you can’t group these together as they are played differently. Would you say that an AGA 5 dan who loses to kyu players in 9x9 isn’t 5 dan? No, he is 5 dan because he is 5 dan in 19x19, 9x9 and 13x13 doesn’t matter. I don’t believe any other ranking system includes smaller boards, EGF, AGA and Asian rankings only counts 19x19. You may estimate someone’s relative strength on other sizes but it shouldn’t affect the 19x19 rank. The traditional ranking system doesn’t really work for 9x9 either as 1 stone on 9x9 is like 6 on 19x19. If I lose to a 5k on 9x9 it doesn’t mean I should be playing against that player without giving him several stones on 19x19. Grouping these together just messes up the ratings.
KGS has all other games than ranked 19x19 as free games which works fine. The ranks become unstable once you mix them together. Tournament games don’t have to be ranked at all and 9x9 and 13x13 could have their own ranks. Separating on board sizes is more important than on live/correspondence/blitz imo.
I thought very hard about this and reached the opposite conclusion. I definitely think far differently under different game speeds and an informal poll of folks found a similar opinion. Separating by game speed is also something that is done with great success for other game sites, so we definitely wanted it here… maybe in the future we can also add the board sizes, but that creates a really complex rating matrix. Perhaps an alternative would be to give you broken out statistics on how you play on different board sizes as opposed to a full broken-out rating for them.
I just wanted to chime in say that I think the new time-based sub ranks are a real bother. I just had a game with a person, and we are both listed generally as 6k. However, under the time settings my opponent was a 5 and I was a 7, so I got two stones. it really should have been an even game and I won handily.
At the head of the board it said yuri(7k), opponent(5k).
In chat, yuri(6k), opponent(6k)
i think its an interesting idea but just messy and problematic when it comes down to the consequences. maybe just display the three ranks in info and in the graph because it is interesting, but just use the single
19x19 derived rank across all games as the official rank
I’ve been reading through this topic and the original thread where two ranks were suggested with some interest.
The original idea was to break out rank based on live vs. correspondence because we wanted to encourage people to play different time settings without feeling like they were risking their rank in another setting. After the switch, I found that instead of being anxious about one rank, I’m now anxious about all of them.
I don’t know if we feel like the original purpose for having multiple ranks is still a good one now that we’re on the other side of the fence. If it really has been helping to get others to play out in different areas then I guess it has. If we think it wasn’t worth it, then it seems poor to be in a situation where a feature was added that has questionable value.
I like the idea of showing statistics rather than having increasingly many ranks. I like it personally because I think it would be too confusing otherwise. Also, while it seems like it should make things easier, I suspect having a multitude of ranks only increases the developmental complexity of trying to match up players in a way that’s fair. I think yuri’s comments are a good example of how having too many ranks can go wrong.
I have a question. What happens if someone lets say registers as 15 kyu. He then only plays live games and improves over a few months time and reaches 10 kyu. What will the blitz and correspondence ranks be? And what will the overall rank be?
shimari raises a very important point here. For the stronger players you take a huge risk entering a tournament right now in terms of ranking points. It needs to be addressed somehow.
For example, games where the rank difference is big, say >4 stones then it would make some sense to make the unranked in tourneys.
Some thoughts on the difficulties presented in the last few decades in ranking players online.
Rank in Go should really be based on performance under fairly well controlled (identical) situations. Lack of control of the situation has been an issue introduced by internet go (and similarily internet chess). The impact on players has often been negative, with worry about how each and every game might influence rank. Divisions between different preferences in play result in players being afraid to dip into blitz or try a longer “more serious” game, or lately mix correspondence games in with their live games. Often players solve this by having a “blitz account” in addition to their “serious account”. Having correspondence games, and different sized games all together might encourage even more accounts, or at least a statistically complex rating system with a bazillion different ranks for each player! I think that sums it up?
In the end, to be ranked, players should be encouraged to play under situation as controlled as possible. I’d suggest a live game with fairly long minimum time controls (and a maximum as well). That rank should move rather quickly, maybe more like we tend to do irl: if two wins in a row against an even opponent, raise rank; if two losses, lower rank. Maybe not quite so quickly, but fairly quickly, so a player can opt to play just to re-establish rank within ±1 stone.
Games played under different time controls can be played at that rank, with the games having no impact on rank. Or perhaps some, but much less impact on rank.
There are obvious ways to try to cheat the system, but making statistics easily available for every player would make such cheating become quickly transparent. In addition, a serious series of wins or losses could be noted (sort of like KGS’s tilde mark) until the player finds the time to play a few ranking games, which – since they change rank for the player relatively quickly – would set the player’s rank back to something approaching correct within a few games.
Just some outside the box thoughts I had, and felt like writing down. I might not even hit reply…
One more thought, totally different direction: one single rank which is impacted more by the category of game you play more often. So if I play an occasional blitz and usually play standard live games, my rank is hardly affected by each blitz game when compared to a single standard live game. If I play a lot of blitz, the occasional long drawn out game doesn’t impact my rank nearly as much as my next blitz game will. But I’d still want a seperate rank just for correspondence games under this, I think, maybe.
Could we bring a bit of common sense into the discussion? At the moment, this is probably the only server to have multiple ranks. The current system is already confusing. There is a rank shown next to player’s time when playing a game and a different rank shown in the game information page. I think that no more than a handful of players on the server know which rank is actually shown where, which is used for handicapping, which is used for tournament pairing… The server and its players haven’t had time to get used to 4 ranks and there are suggestions to add more. Please, first make current system at least slightly understandable(like indicating next to the rank which rank it is, or making some kind of a helpfile) and then start thinking about more ranks. Thanks