That is a good point.
Itâs not against OGS policy to beat a bot that does not know ladders. It is against policy to exploit that by repeatedly playing against that bot. The reason is the two situations (bot with a bug and a weak human) are not comparable. Bots with bugs donât improve and are often neglected (not fixed). Humans learn basic situations like that, usually very quickly, so there is, very soon, no longer such a weakness to exploit.
There are much more simple bugs that have existed and been exploited that are completely different from anything a human would do even once. It might make it more clear if I described them, but I wonât, lest it help cheats in the future.
Nooneâs arguing about what is or what isnât against policy. I just see no reason why bots deserve to be protected from their own gameplay weaknesses. If those weaknesses are so bad that they wreck the rating system, then either the rating system is broken beyond repair because it apparently makes the implicit assumption that the players in it are not exploitable, bots need to be banned from ranked, or (my preference) we should trust the rating system to overcome noise and to rate the bots correctly exploits and all and accept that weak Go bots are notoriously bad as they have always been and no bandaid of pretending that theyâre not in a particular area is completely missing the point
Nobody is protecting the bots, not at least when I was moderating. The purpose is to protect the ranking system. SDKs, for example, may not want to waste their time playing pseudo-SDKs who are really inflated 25k players. As I already noted, fixing the bot is what needs to be done. Usually, however, it takes more than the normal time to come to the modsâ attention. When it does, the bot is usually suspended until fixed.
Several years ago, I suspended a whole large family of very popular bots due to a simple, functional flaw (not something any human would do). It had already been exploited, IIRC, by three or four players. It turned out that the fix was very easy, done in one day I think, but it took weeks to get the administratorâs attention. When the fix was made, I lifted the suspension.
1 Like
Problem is not about protecting bot or Rating System. Problem is that 1 human user will have very incorrect rank.
Only because the botâs rank was inflated, because people arenât allowed to use the exploits to keep the botâs ranking at what it deserves to be. If the botâs rank was at itâs correct value, it wouldnât be exploitable in that way because you wouldnât earn many rating points from playing it
I think you are both conflating the intent of the policy, with what the policy does. I keep saying the policy protects bots. I am not saying the policy intends to protect bots, Iâm saying it does
Surely itâs not controversial to point out that making it illegal to exploit a botâs weaknesses within the rules of the game (not score cheating or stalling or anything else which would be illegal against a human), is protecting the bots. You can claim that thatâs not the intent, and Iâve no reason to doubt that, but surely you canât claim that itâs not protecting the bots
That is not illegal, and the policy does not do that. The problem here is the broad us of the term âweakness,â which I have already explained does not mean a playing quality in normal human terms. For example, letâs say a person had a stroke during a game, but was still capable of playing again, but with a brain that was seriously damaged The opponent plays and wins 10 or 20 more games with the brain-damaged person. That is the situation that is more properly comparable to âbot exploits,â although still not exactly comparable.
As I have said and will repeat, in most instances, exploits consist of simple flaws that are not analogous to anything humans would do, so no valid grounds for comparison exists.
You have cannot play out ladders as an example. You keep implying that I donât understand what you mean by exploit in this context when I understand perfectly well. In Arimaa pieces are shuffled in order to bait a bot into getting its elephant blockaded: this is an exploit: the moves make no sense from a human perspective, and induce behavior that even a very weak human would be able to avoid, in order to win
It is very simple. The bot has a weakness. This makes certain moves very strong against it. Playing those moves, is illegal. This protects the bot from being properly rated against non-cooperative opposition
That has nothing to do with it, which tells me you very much donât understand it. I give up.
Actually I wonder if this could largely be solved in the rating system, by silently reducing the effect of repeated games with the same result between the same pair of players.
If A beats B 100 times in a row, whether or not either of them is a bot, we should not count that the same as A winning against 100 different players, right?
4 Likes
Excellent idea. It could also be solved by making all bot games unranked.
2 Likes
Or require to play 4 rated games against a human before playing a bot again. (Rated games against bots are necessary, otherwise bots canât get a rank.)
2 Likes
Or it could be solved by making all human games unranked!
2 Likes