Besides the question of “correct” 9x9 handicap, it’s also relevant how much can be gained from a single game (on 9x9, and with the used time control).
In the rating system I use in my youth club (and also in the EGF rating system) it’s usually less than 1 rank.
I suppose it can be much more than that on OGS, as this is one of the features of Glicko2 for which it was chosen. So the rating system behaviour discussed here may just be a consequence of that choice?
4 Likes
This discussion has not tried itself. Thanks to the moderators for reading. There are problems in all discussions.
Could you clarify? I’m not even sure whether to read the last sentence as conciliatory or negative, and I’m not sure what it means for a discussion to “try itself”
1 Like
Discipline that there is really nothing to add to this discussion. Maybe I am wrong.
1 Like
I think @anoek could definitely add to this discussion.
It seems clear that removing the 9-rank difference limit has had undesirable consequences, and that something should be done about it, allowing people to play weird bots as much as they like, but not to gain rank indefinitely. Surely it would be enough to say that a victory cannot increase your rank to more than 5 (or even 3) ranks stronger than the opponent you defeat? I feel that rank definitely needs to be demonstrated against people of similar strength. Ideally against multiple opponents as well.
2 Likes
7 Likes
May be we need prohibit ranked games with bots completely?
I think if a person always playing with bot and completely avoids playing with real people it is very wrong. It is just my opinion.
2 Likes
Hah hah very funny.
And timewasting.
This account abusing a bot has been banned, and the games will be annulled.
Please don’t do that.
1 Like
Unfortunately there’s a reasonable argument that says the bots here are a useful tool for establishing a legitimate rank.
This has always been the case. There are a ton of bot exploits, this is only one of them.
We can’t prevent rank manipulation by bot exploit with solutions within the ranking system.
Bot exploits don’t tell us that the ranking system is broken.
Right now the solution for bot exploits is spotting abusers and persuading them not to do that.
What’s broken is if this can happen with human players.
I think we have established that it can, because the value attributed to handicap stones is too high, past a certain point.
Although I might add that so far this is theoretical: no-one has demonstrated that it does happen in normal practice.
anoek’s back, he’s looking at this.
If there are folk genuinely interested in helping solve the problem (proposing solutions and working through the maths, implementing, testing etc) I know anoek would welcome practical assistance.
5 Likes
how about put all the 9*9 handicap game with more than 2 stone handi in unranked?
because if I understood rightly, a 9 stone handi on 9"9 game in the equivalent of a 52 stone handi on 19"19. And a 52 stone handi game is obviously unranked no ?
because 9*9 is such a small and short game it doesn’t follow standard handi stone system, it can’t handle it. handi stone make sense in 19"19 not in 9"9
so if we agree that the maximum standard handi stones in 19"19 is 9, and If we divide this by 6, 9÷6=1.5 so in 9"9 maybe we should only allow handi stone until 2 stone in ranked 9"9 handi game ?
the point is that handicap game with handi stone was made and create to make sense for 19"19 game, not 9*9 game, and it’s dangerous for our ranking system if we keep allowing 9*9 nonsense high numbers of handi stone in ranking game
13 Likes
Yes - good call: limiting the number of handicap stones that can be ranked is at the top of the list of solutions being considered. Your suggestion of how to determine the number is a good start (that’s the obvious first question! ) 
9 Likes
If we also prohibited ranked games against humans, then all the ranking system problems would be fixed!
5 Likes
I think it was a proof of concept to prove that the issue is replicable. I don’t think any ill will was intended
8 Likes
Thank you. Yes, it was an experiment, mostly to see how big was the extent of the issue (how quickly and high one can get) and it was not meant to break the rating system and harm other players. So I used a test account which I did not plan on actually using to play with others. We can easily imagine what would be the result if a 19d (highest I could get in small amount of games) were to lose puposedly against a 10K.
3 Likes
Haven’t posted in this thread up to this point. Coming out of lurk to second @sakesoju’s suggestion.
If there is already a handicap limit at which 19x19 games can no longer be ranked, and there is already a multiplicative scaling that treats handicap as more impactful on 9x9 or 13x13 games for ranking purposes, then 9x9 and 13x13 games should also not be ranked when the handicap after scaling exceeds the limit.
This seems easy to justify under the same reasons for the ranked handicap limit existing on 19x19 in the first place. It would prevent the most extreme cases of ranking system weirdness of this kind, just like the 19x19 handicap limit already prevents the analogous extreme cases that would no doubt take place if it didn’t exist.
Of course it may still be desirable to consider other possibilities besides this. But if I’m not missing anything, this change is simple and actionable. And discussion of other options could still occur meanwhile - this change would be very easy to revert too, if further discussion eventually resulted in a consensus on other options instead.
10 Likes
To be frank, OGS main site is not the place for experimentation, especially with ranked games.
You’d think we’d have a sense of humour about it, and to certain extent that is true, but don’t forget that everyone thinks they are the funny one, and each “joke” takes mental and actual time consuming work.
I did already say it’s not allowed, and we already knew it happens: it’s show in the original post.
If you want to experiment, do it on Beta.
2 Likes
Mods may already know it happens, and some others may have made the connection, but until kosoren’s post, I had not seen any evidence that it was replicable with so few games, which is very relevant for the discussion, so I found it a very useful thing. Would it preferably have been done on the beta server? Yeah, unless I’m missing some reason it couldn’t be, but a “Thanks for taking the time to do this, in the future the beta server would be a good place for this sort of thing” would have sufficed imo
5 Likes
I thought it was a good demonstration too, even if a bit silly.
And it enlightened at least one mod, it appears
perhaps worth pointing out that @Amaranthus doesn’t play on Beta 
8 Likes
Ah. Main server it is then. I don’t see the issue in this specific case, then. It was done with the minimal side effects possible given the constraints on the proof of concept
1 Like