Life In 19x19

A few people have now raised a dislike that a very good record can be beaten by a 0% record if enough games are played, and there have been two proposals launched to help combat the extremes of this situation. As a result, I have compiled a list of the three options (current system, proposal A, and proposal B):

Current system (50% points for a loss):

Effectively a loss is worth 50% of a win, so playing 3 people and winning each is worth 6 points, and playing 3 people and losing each is worth 3 points.

Scenarios (assuming all different opponents):
Scenario 1 - 0-13 record = 13 points
Scenario 2 - 2-11 record = 15 points
Scenario 3 - 5-8 record = 18 points
Scenario 4 - 6-4 record = 16 points
Scenario 5 - 6-0 record = 12 points
Scenario 6 - 4-0 record = 8 points

Proposal A (25% points for a loss):

Reduce the number of points awarded for losses to 25% of wins. In the above situation, the 0-13 result will be worth only 6.5 points, and a 4-0 record will be good enough to have an advantage.

Scenarios (assuming all different opponents):
Scenario 1 - 0-13 record = 6.5 points
Scenario 2 - 2-11 record = 9.5 points
Scenario 3 - 5-8 record = 14 points
Scenario 4 - 6-4 record = 14 points
Scenario 5 - 6-0 record = 12 points
Scenario 6 - 4-0 record = 8 points

Proposal B (50% points for a loss, but capped loss points):

Same as the current system, but the total number of points awarded for lost games cannot exceed the total number of points awarded for wins. So, with a 3-10 record, 6 points are available for the 10 losses. If the player then wins another game, they can pick up an additional 2 points for their losses too. (EDIT: There are also discussions about the possibility of a lower bound to this cap - e.g. a player can always get 5 points for losses, but to get more than that they need to get more than 5 points from wins)

Scenarios (assuming all different opponents):
Scenario 1 - 0-13 record = 0 points (5 points with the 5 point lower bound)
Scenario 2 - 2-11 record = 8 points (9 points with the 5 point lower bound)
Scenario 3 - 5-8 record = 18 points
Scenario 4 - 6-4 record = 16 points
Scenario 5 - 6-0 record = 12 points
Scenario 6 - 4-0 record = 8 points

Scenario results in descending points order

1 - Current system (13 points), proposal A (6.5 points), proposal B (0 points - 5 with lower bound)
2 - Current system (15 points), proposal A (9.5 points), proposal B (8 points - 9 with lower bound)
3 - Current system (18 points), proposal B (18 points), proposal A (14 points)
4 - Current system (16 points), proposal B (16 points), proposal A (14 points)
5 - Current system (12 points), proposal B (12 points), proposal A (12 points)
6 - Current system (8 points), proposal B (8 points), proposal A (8 points)

Ranking of scenarios in proposal order

Current System - 1st (5-8), 2nd (6-4), 3rd (2-11), 4th (0-13), 5th (6-0), 6th (4-0)
Proposal A - =1st (5-8), =1st (6-4), 3rd (6-0), 4th (2-11), 5th (0-13), 6th (4-0)
Proposal B - 1st (5-8), 2nd (6-4), 3rd (6-0), =4th (4-0), =4th (2-11), 6th (0-13)

This change is what will make me join the league eventually. B seems to be a favorable change in my opinion.

There are other ways B can be tweaked of course. For example, instead of "points for wins", the cap could be whichever is higher out of 5 points and "points for wins", to make sure that all players can at least get to 5 points.

Taken from another ASR thread on this topic, I posted the following that seems relevant:

daal wrote:

topazg wrote:Another criticism of the current system is that it is way too generous to people with bad records. Currently, 4th place in the top division has achieved it just by playing lots and losing every game. To make a contrived example...

Now, a few days later, the highest placed player in Alpha with a negative record is you in 5th place. In the entire league there are only two players with a negative record that currently occupy a promotion slot. do you have any statistics on how that has played out in the past? I think that that would be more relevant that a "contrived" example.

Another question is who would benefit by a rules change and what would that benefit mean to the structure of the league? Is it that stronger players who would otherwise be beaten out by "hyperactive" weaker players would be able to promote easier? Wouldn't that make the classes more homogenous in the long term? Seems undesirable.

I know that I presented an opposite view before entering the league, but now that I'm in it, I see it's main function as being to provide a good learning environment for the most players possible, and the current system seems to be doing that quite well. If it ain't broke, don't fix it.

Yes, I agree, and I feel very guilty about being 5th, as I think I should be demoted this month, my record has sucked! I don't think this is likely to be an issue for people promoting unreasonably, as there are always strong + active players in a class, but there are a number of people displaying aggravation that people aren't being demoted with incredibly weak records. I didn't feel particularly strongly either way before, however, my personal experiences have changed things:

Now, I am slightly in favour of a change purely because I feel guilty about my record putting me in the top half of the class, as I think it shouldn't. I've found myself playing games "just because", which is against the spirit of the league. My first 3 games were played with a high fever because I didn't want to fall behind in the first week of the class, but they were all epic disasters. The majority of the games I've played have been either tired, between midnight and 2am, or in a tabbed window at work, because I haven't had a great deal of time to play this month, I've just been too busy in the "normal" part of the evening. Some games I've taken on knowing I'll lose but thinking "Ah well, it's all points!" - I should be taking on expected losses as serious games with the intent of learning and having, in effect, a teaching game, but actually the point reward has been enough just to play regardless of how tired I am or how I'm feeling.

More activity is a core tenet of the ASR, but "junk games" aren't supposed to be. Ok, I'm guilty here of being a hypocrite by advocating the league and then playing games without the seriousness they deserve, but the reward system sure made it tempting to do so. If the limit system proposed were adopted, I'd currently be on 13 points, and to get more, I'd need to win games. That will encourage me to take the games more seriously, and encourage me to seek out weaker players for wins (an advantage for them RE: teaching as well, because I always offer to review properly), so that I can enjoy my games against stronger players for learning. I will then make the most of those games against stronger people, because I won't be getting an endless supply of points for those losses, so finishing quickly to play the next one has no reward.

This is only my perspective, and it's one I've only really formed over the last few weeks, but I am steadily becoming of the opinion that the current system encourages junk losses for the sake of simply fitting games in, even if only a minority use it for this.

Proposal B - 1st (5-8), 2nd (6-4), 3rd (6-0), =4th (4-0), =4th (2-11), 6th (0-13)

this proposal does not promote activity purely because someone who played 13 games is behind someone who only played and won 4 games. I think someone who played 13 games in the league should have benefits because playing only 4 games and being in front of them is just wrong.

stalkor wrote:this proposal does not promote activity purely because someone who played 13 games is behind someone who only played and won 4 games

This is false logic. Someone who played 13 games still has more points than someone who played 3, it just increases the relative reward of wins in these two examples. Activity is still clearly promoted by having rewards for simply playing games.

stalkor wrote:I think someone who played 13 games in the league should have benefits because playing only 4 games and being in front of them is just wrong.

Remember that this is just your opinion, that is shared by some but not all of the league members. I think there is a fine balance to be gotten between activity and score, and it will probably never be reached for everyone. I am beginning to believe from received comments that it could currently do with tweaking, and my own experiences this month make me inclined to agree.

If those 13 games were just quickly thrashed out with little thought to get 13 points, would you feel differently?

I voted change A, but I'm fine with change B also. I feel that you are right about the amount of junk games being played. I've played against a couple of weaker people that didn't seem to really care about the game and just wanted to get it over with (for the points I suspected). So yeah, I agree on change, what change I'm not sure.

I like proposal A.

I don't like the idea of capping losses with wins because that puts more emphasis on winning rather than playing games as a fun experience. Of course, there is a lot of weight on winning as it is, but I feel this cap would pressure players to worry about winning even more, which is not what I think go should be about. Proposal A would make sure those who are losing many many games aren't gaining much from them, but ARE gaining something. This way, a weaker player would be more likely to play a stronger player even if he doesn't think he can win. With proposal B, it could be harder for stronger players to get games and even harder to inspire participation from new weaker players to join the league because of the difficulty in attaining points.

Proposal A would lower the points from losses enough that it would be difficult to "unjustly" be promoted or stay in a higher level. But it would also push players to play more games, even if their opponents are stronger.

does the score need to be a linear function of wins and losses? What if you tried something like wins + sqrt(losses) ? The value of each successive loss decreases, so playing without trying to win has diminishing returns. For the examples you gave, the result would be:

6-4 : 8.0
5-8 : 7.8
6-0 : 6.0
2-11: 5.3
4-0 : 4.0
0-13: 3.6

This seems simpler and harder to cheat than a system based on caps.

emeraldemon wrote:does the score need to be a linear function of wins and losses? What if you tried something like wins + sqrt(losses) ? The value of each successive loss decreases, so playing without trying to win has diminishing returns. For the examples you gave, the result would be:

6-4 : 8.0
5-8 : 7.8
6-0 : 6.0
2-11: 5.3
4-0 : 4.0
0-13: 3.6

This seems simpler and harder to cheat than a system based on caps.

The only problem with something like this is tying it into the current system, which already has diminishing returns on both wins and losses when against a player you have already played. One thing I intend on doing in the back-end code is separating out "points from wins" and "points from losses" so that we can implement a number of the suggested ideas - once that's done, the scope for what else we can do improves.

Now that i have been looking how the system goes, the current system works very fine to really encourage activity among all people, this is a good thing, but some people dislike it beacause people wich good win record but few activity dont go well, claiming that someones are superactive, and they are right, but you have to be superactive to play the 39 available matchs you can in a month, that means playing at least a match in a day or play many matches in few days, this also makes you play only league matches instead a regular rated matches, wich is also good for more serious playing.

So this high level of activity makes you keep playing and improving more. But well some people(strong people) dont want to play that much and still keep playing in the league, and they have their point. So one of the few viable solutions to have everybody happy(superactives, strong-but-underactives and everything else) is to only have one match per player, so you only have to play 13 matches, there will be no need for superactivity and this single match may rise in quality beacause of its importance since you can only play 13 times, so better you win all of them. This of course improve the quality of the matches and keep strong busy-in-other-things people happy.

About improving, mmm i guess both ways are good quality and quantity are both very important, still the current matches are already of high quality . So this is one more option that can be added, but since people prefer 3 matches instead of 2, i bet also prefer 3 instead of 1, this tells people like activity.

Also making the losing points going half his value, seems half right , but looks like the proposal is not very popular, still i like more this proposal of half points than capping the losing points wich seems will make the strong players play less, if people want play less, just reduce the amount of matchs available or make go to half the value of losses point, but this cap loss points thing doesnt seems good, also others ubermambojumbo systems looks very strange and seems solving nothing.

So in resume, yes a change may be good, be it: one match per person or half the value of losses, otherwise keep the current system wich is already great. By the way the 3/6 promotion demotion is a very good admin move, thanks to all the admins for they very great work. =)

At the time of this posting, 60% of the voters favours change. I'd say: viva la revolución!

I've seen point systems before, and I like this system the best...

3 points for a win, 1 point for a loss. If you play the same opponent, then the points decrease by half each time.

so A vs B and A wins = A 3 points and B 1 point.
Then A plays B again, A gets 1.5 which brings 3 up to 4.5, and B gets .5, which brings him up to 1.5.
Third round, A wins, A gets .75, and B gets .25= A 2.25 abd B 1.75.

As for if B wins the third round, the same math can be applied from the system that's in use to provide the divisions. But the start or main being a 3/1 ratio is my favorite.

Instead of punishing the many loses players, you should award the many wins! This gives the same result but better feeling for those who play actively.

On how to do it, you can do as how goverment taxes you. If someone's win points is higher than 10 then they have 20% bonus for points above 10; above 20 50%; etc.

it seems over 50% has voted for the current system to carry on.

I think the current system has gotten better when we decided to have promotion/demotion at 3/6 so in hindsight we can better keep the current system and "tweak" the things around it. I would like to thank everybody who decided to post their own ideas, this helps us have an open mind about what is possible and keeps us from getting tunnelvision. Feel free to use this thread to have a discussion about different systems a league can use (for instance compare korean insei league vs ASR league, or even real life leagues)

I still think this system is best when considering online go, keeping in mind that internet does not bind players like a real club does. Though i think i am gradually turning ppl around in thinking there is merit to being part of an online community, but thats a painful and longterm goal. KGS is still the most social place to play go on and i will continue finding it a good place to play go:)