DANGER, rating system will break if bots play rated !

I see Lichess now has BOT players and is allowing them to play rated games against humans, as for example Lella Zero vs penguingim1. But the human rating system should not be affected by the computer rating system ! If that goes on, then our rating system will be screwed up forever, just like it happens on ICC, were human ratings spin up to 3700. Now please don't allow this to happen in Lichess anymore.

Lichess should keep two distinct and independent rating scales, one only for humans, another one for bots. I know we also desire human vs bot games to be rated, but there is a simple solution. Just make it part of the bots rating system. In other words, penguingim1 should have started with a 1500 provisional rating. He would then gradually build his HumanVsBot ratings.

Well, I‘ll regard this bot approach critical as well. Lichess is at the crossroads in my opinion. I plead for the strictest separation of humans and machines.

As far as I am concerned: my schedule is full with offline chess for the next years, internet chess is training and fun yet dispensable. So dealing with bots in any way will not be part of the agenda (at least for the time being).

In case you haven't noticed already, human ratings remain unaffected when they play against a bot. Only the bot's rating will change, try it out yourself.

@relevantproblem That's not true. The ratings are affected in a usual way. It is often +-0 for the human if the bot rating is provisional and much higher, but it actually changes.
@Dopaminergic_Chess Can you explain the problem more concretely? Does ICC use Glicko-2 rating system?

I don't see a problem of "screwing" human rating system. Glicko-2 rating mass is quite stable and don't change a lot due to the new players. However, I see another problem. Bots objectively play between each other much more than with humans. Thus their rating mass will indeed become more or less independent of human rating mass, and as average bot plays much stronger than average human, bots ratings will be consistently higher. Now this gives an easy legal way of sandbagging - just play a lot of rated games against bots...

In case you haven't noticed, human ratings are slightly affected when they play against a bot:
lichess.org/@/LeelaChessOfficial/all

#4 I agree, losing against bots becomes a predictable way to lose points, although you'd only lose ~1-2 Elo at a time against a 3000-rated BOT opponent.

@Wolfram_EP Human ratings are not affected. When a game is over it might say -8 to the human, but if you look at the rating itself, it remains unchanged. The bot's rating however will change.

@Toadofsky In the same link you posted, if you look at Penguin's Ultrabullet rating over the last 4 games, it's unchanged. In fact, his rating is stuck at 2912 in all those games, even though at the end it says +-n

Oops, I think Leela caused some weird glitch but in general, rated games are rated (though human opponents gain/lose 50% rating points).

That's not true. Look at my recent games instead: lichess.org/@/relevantproblem/rated

Opponent Cuxhaven's rating remains unchanged even though it says he gains/loses points. It happens in all bot games. Not sure if this is intentional or not, but this is what is happening.

Thanks @relevantproblem , I have filed an issue:
github.com/ornicar/lila/issues/4297