I know there was an old thread on this, but it was closed…
I pay attention to my numerical rank, and typically in server-chosen handicap or even games, my ranking goes up or down 0.1. Sometimes, it works out to 0.2, because math is that way. But I played a couple ranked games a while back, maybe 40 games apart, and when I lost those, my score dropped 0.3. It was confusing to me, so I dug in a little and believe I figured out why. It seems that with both, the server had a rough rank for them that was several steps down from mine. When I lost to them, therefore, the server dinged me more than it would have if I had lost to someone my own rank.
Now, I don’t generally like playing with ? players regardless, because I’ve had too many games with 25k players who don’t know any better and select a game with me. But now, seeing how the server scores them, I just cancel games with ? players who accept my public match. But that’s suboptimal as well.
(For the record, I think ? players new to the server should start their own matches, stating what they believe their rank to be, and let people play with them that way.)
So I would offer two/three suggestions:
In the “restrict rank” field, add an option barring ? players.
A. Have ranked games vs ? players change the ranking of the ? player, but have no impact on the ranked player. (KGS did it this way, and it always seemed reasonable to me, and it encourages people to play with the newbie).
OR
B. If the server has some sense of the ? players rank, have it offer the handicap it believes would make for an even game.
Or maybe I’m completely wrong about why my score moved, and it was some wild coincidence that playing these new players scrolled off a game from 12 times ago that was impacting my score.
That’s not how the system works. If you want to play ranked games as newcomer you have to play in a range of 9 stones around the level assigned (6k?) at the very beginning.
Then:
1 We cannot expect OGS to encourage players to discard games with [?]. You found a way and you can warn players in the description of your offer.
2A Yes. Now it seems that the disturbance caused by innappropriate pairing with [?] Is very marginal.
2B The existence of the [?] is due to have no sense of the newcomers rank. Besides one has to consider some security policy (Sandbagging) so the system is considered till now as the best compromise for OGS (not for me but you can search there are more topics on this)
I understand that. I’m saying a ? player can create a new game, and in the description say, “I think I’m around 3K.” Then players with established ranks can choose to play with them. I find that to be more polite than just sitting at a game someone else created, not saying what you think your rank is, and playing. That’s just my preference, and how I’ve done it, not a rule I’m proposing.
Agreed marginal, but non-zero. And it could be zero, and – if I can generalize from myself – that would make people more eager to play with ? players.
In the very first game, of course. But after a few games the server has some sense, just not full confidence. Clearly, the impact on my score is based on the server’s guess at what the opponents rank was.
My point here is that I see literally zero good coming from adjusting an established players rank based on their game with a player for whom the rank is uncertain, and I do see harm.
To me this is more a bug than a feature.
The system that @Bunburyist describes seems to be the best way to get a rank without playing too easy/too hard opponents: just play people around the rank you think you are and after wining/loosing 50% of the time this is the rank the system will eventually give you.
The problem is that that’s not how it works today >3d and <15k s cannot play with people at their level as you describe…
There are not many >3ds in western servers anyway, but 15k< are likely beginners that probably do not understand how glicko works.
The reason the old thread was closed is that the discussion was extensive and IIRC possibly repetitious. Barring ? players with the rank restriction function flies in the face of historically firm OGS policy because it would make it difficult for them to get rank. As for the suggestion of a one-way impact from a game, I suspect this is impossible under Glicko.
The related question of the 6k pseudo-rank of ? players was also extensively discussed in another thread. It was shown (convincingly to many) that the 6k starting pseudo-rank is not necessary to Glicko. So perhaps some change could be made there.
Actually glicko already takes rank certainty into account… So that means that new accounts get a huge rating shift from playing established players, but the established account almost sees no effect at all from their game against a new account.
As I mentioned in my OP, the two times that I watched closely, losses to ? players resulted in a 0.3 drop, while a typical loss in a even game results in a 0.1 drop most of the time, maybe 0.2 in the outlying 10%.
Obvs, I haven’t seen the OGS code, but I’d be surprised if there wasn’t a way to calculate it like a ranked game, but consider it an unranked game for the purposes of the ranked player.
Since we were having this discussion, I accepted a ? game. The server put the player around 5k, and I’m 1k. We played an even game and I one. My rank went up 0.1, which is exactly the same I get when I beat a 1k player in a ranked game.
Maybe we just got unlucky, and that one game I just did had a strange, outsized impact. But I suspect overall that the impact on the ranked player’s rank is greater than you think it is.
Do you see no harm in your proposition? If only the ? Players rank is adjusted, players with high rank could repeatedly let the ? Player win on purpose and swindle them a high rank. Did you consider this possibility?
I don’t know why someone would do that in a server where nobody is playing for money, but it would be an option.
BUT - let’s take for example, we have a 6D player who is ranked. A new player comes in with a ? and he plays at an 8D level. If the 6D repeatedly loses to the ? player, should his rank really dip below 6D? Should he lose points because he’s losing to a stronger player who simply hasn’t been properly assigned by the server? I certainly don’t think so. So no, I don’t think that a player who repeatedly loses to a ? player should lose rank.
Does that system allow abuse for someone who wants to temporarily inflate their rank by coordinating with another player? Yes, it does. And every other system that could exist will also allow that possibility.
I would be fine with WHR, one of the consequences of which is that rating points are retroactively adjusted based on how the ratings of players you played developed, but I think many people would dislike the unintuitive effect on their ratings.
My answer was only about the fact that, with the system we have, your suggestion can’t be used for players being outside the range. A newcomer beginner cannot play a ranked game with another rated beginner for example.
I do agree (and agreed ), there have been long debates in precedent topics on this
Why should a beginner understand glicko?
Glicko is not the problem. The problem is to not let start a beginner at a beginner level but to let him being crushed in a long serie of games.
It has been established in an older thread that there is no problem considering Glicko if players would have different entry points assigned.
That’s super interesting, and glancing through the numbers - if I can read them correctly - I must be wrong about what happened? I’m looking through them now, thanks.
Maybe there’s a misunderstanding here, but that’s exactly what I said in my message.
The problem is that someone weaker than ~15k cannot play even games with players at the same level to get a rank. They need to play much stronger players and loose until they fall into that range.