Not really a counter argument, but again sort of a related comment: I believe the benefit of ranked games vs bots is that it’s supposed to be a quick way of getting to your “proper” rank without having to potentially sandbag or airbag against humans.
If you’re over ranked you might legitimately play a 25 kyu bot, lose a fair amount of the time, and then settle on a proper rank before matching up with humans again.
If you’re under ranked you might quickly rank up with a bot and then go back to roughly even matches with humans and not be accused by opponents of sandbagging along the way.
So if one makes bot games worth half as much, or a quarter as much, then you’re potentially slowing that process down.
There could be some hybrid idea, but I don’t have an idea whether it would be better or not off the top of my head.
I wonder how many players actually “set” their ranks solely based on vs AI games from the start. Is it common? I can also see ranking by AIs as a loophole for people who want to create multiple sandbag accounts by deliberately losing to bots, which nobody would bother to check or report vs AI games. It all depends on the portion of players actually use AIs as to set their ranks. (besides, as others had said above, not every bots is properly adjusted, also their rankings can flucturated a lot, so if you just play a few games in a very short amount of time, it might set a player at the wrong starting points as well)
I mean it is true that people can use ai to sandbag, but we also have an option to choose a starting rank so it’s not even as necessary to do that.
Equally you might think you can fix the problem by making bit games unranked but people can still probably just sandbag with multiple accounts playing each other. Losing rating against a bot account or another one of your own accounts probably will take a similar amount of time to be reported and noticed, essentially when you start playing ordinary players.
Equally playing humans can mess up your starting rank quite badly. If you’re a beginner and you match with an SDK on a 19x19 board, occasionally you see the experienced player resign because they become aware a few moves in that the player isn’t near their level and they don’t want to play out a full game. Those “big upset” results can also take a quite a while to be undone.
I don’t think there’s one thing you can do to fix all the problems, and I agree it does matter the relative proportions of players using or abusing features.
It’s actually a bit different, because bots played hundreds, if not more games daily, their ranks although fluctuating, stay within a range, hence it is harder to detect one side sandbagging. Manual sandbadding will leave records on both accounts, and the number of manual input limit the number of games can be utilized between two accounts, hence easier to detect unusal up and down (the artificially deflated account will boost the other side making it less useful afterward). They need to either registered a lot more accounts just to create the effect, making them easier to detect.
This is exactly the reason why I proposed different weights for different types of games, making the impact of abandoned games, etc., less “impactful”, even exploits less impactful, while if you abandoned “tournament games”, you are effectively out, and players tend to be more thoughtful about joining or abandoning these games. This is also why I’d want vs AI games very less impactful, keep their ranks less fluctuated, and take much longer to have the effect when someone exploit it (like if it takes 0.1 of the weights, it would be very tedious or easy to catch the exploit, if they just resigned like only a few moves in, and even making the game weighted associste with how many moves had been played). The key is not to eliminate problems; even OTB games and tournaments had their flaws, like players can also cheat, or sandbagging, but we put more effort on those in large tournaments than in club games or local games. There is no fool proof ways, just try to find a better way.
I think bot ranking could be interesting if it worked well. Unfortunately, I don’t think the types of bots that exist are well-suited for gauging rank.
I also don’t think the moderation burden goes down, as it would still be possible to airbag/sandbag/cheat in human games. AIranking system possibly makes this easier.
I disagree. They often take sadistic pleasure in humiliation; this is especially true of sandbaggers. Alternatively, some are vandals who enjoy destruction for its own sake.
