Drawbacks of BOT Accounts • page 1/1 • Lichess Feedback • lichess.org

While the BOT accounts are good for people to test the strength of the engines they program, there are many drawbacks. First, if someone plays a rated match with a BOT account, it could have the same result as sandbagging. For example, if I played a 50 game rated atomic match with a BOT account, my rating would be deflated significantly, and then it would not be fair to my human opponents. Second, while the purpose of the BOT accounts seems to be so programmers can test the engines that they program, there are many users that do not even know how to program, are not trying to help program any chess engine, yet they are downloading Stockfish, making it a BOT account, and learning how to cheat. These accounts just deflate the ratings of the BOT's opponent with no positive result for the chess community. I think there are multiple changes that should be made with the BOT accounts. #1, how about giving BOT accounts a 2500? rating to start and make it so BOT accounts can only play rated vs other BOT accounts? This way there would not be any loopholes leading to what is currently legal cheating and sandbagging, and the community's ratings would not be affected by this. Also, if it's really necessary to have users play rated games with BOT accounts, why not make it so the human player's rating will not change, but the BOT account's rating will? As I am not a programmer, I do not know how easily this could be added, so I'm not sure how popular this idea would be with the devs. So, while these BOT accounts are a good idea so developers can test their engines, there need to be rules set in place to prevent these accounts having a negative impact on the community. Any thoughts on this?

thekid456

#2

I do believe that there are drawbacks, but some bots you can challenge casual, or some bots even if you challenge a rated match do not actually rate the game, @leelachess for example

Toadofsky

#3

Geez, so much ratings concern... the rating system is resilient, even if not quite as much as:
http://www.glicko.net/glicko/glicko2.pdf

People who don't know how to program struggle using the BOT API anyway: https://xkcd.com/1987/
https://imgs.xkcd.com/comics/python_environment.png

Geez, so much ratings concern... the rating system is resilient, even if not quite as much as: http://www.glicko.net/glicko/glicko2.pdf People who don't know how to program struggle using the BOT API anyway: https://xkcd.com/1987/ https://imgs.xkcd.com/comics/python_environment.png

FM Eldrail

#4

First, thumbs up for xkcd!
But to be fair, setting up the API can be most probably be done by everyone as there already is a video tutorial out there. But unless somebody can show some proof that it really hurts the system, why not let it be. Would still be nice to have a chance to get an overview of all bots. Not so easy to find them all right now.

First, thumbs up for xkcd! But to be fair, setting up the API can be most probably be done by everyone as there already is a video tutorial out there. But unless somebody can show some proof that it really hurts the system, why not let it be. Would still be nice to have a chance to get an overview of all bots. Not so easy to find them all right now.

BOT CDrill edited

#5

@ijh

"First, if someone plays a rated match with a BOT account, it could have the same result as sandbagging. For example, if I played a 50 game rated atomic match with a BOT account, my rating would be deflated significantly, and then it would not be fair to my human opponents."

If you intentionally lose to a bot, you can be sandbagging. If you lose normally to a bot then it is fine. Your new rating and that of the bot get adjusted to reflect your strengths.

"Second, while the purpose of the BOT accounts seems to be so programmers can test the engines that they program, there are many users that do not even know how to program, are not trying to help program any chess engine, yet they are downloading Stockfish, making it a BOT account, and learning how to cheat."

There can be a lot of reasons for creating a bot, one of those is what you said. The others can be experimenting book moves. Both bots maybe using the same engine but they can have different opening books and hardwares. LC0 could be interesting specially if your have a good GPU. Some people have a good hardware. Those people that don't have a good GPU but has a good CPU may run Stockfish and try to play against LC0 running on good GPU. It would be interesting to watch such games. Humans vs LC0 on long game like TC 2hrs + 30s inc/move can also be interesting. One of the reasons why a bot is usually stronger than human is that the bot is simply consistent. It does not blunder much. Don't expect too much from humans playing against a decent bot in a short TC (Time Control), just call it a fun play. Eventually humans will be time-pressured which would lead to bad moves/positions. Having a good hardware and opening book can be one of the reasons why bot authors who do not run their own engine created a bot. Perhaps others would just like to have fun screwing proud human players :-) Not all people that learn to setup a bot would learn how to cheat. They may learn how to cheat but does not mean that they will actually cheat. If you think someone is cheating you can report it, Lichess has a system of detecting cheaters.

"I think there are multiple changes that should be made with the BOT accounts. #1, how about giving BOT accounts a 2500? rating to start and make it so BOT accounts can only play rated vs other BOT accounts? This way there would not be any loopholes leading to what is currently legal cheating and sandbagging, and the community's ratings would not be affected by this."

Not all bots are strong, there are weak bots. CDrill for example has an option to blunder and is rated around 1800. It is intended to play against humans and would try to play like humans, it is still under development. It is indeed interesting that Lichess allows bot now. This opens up a landscape for some researchers to study human behaviours comparing human vs human, and human vs bot performances. Are humans feels weaker if it play against a bot, even if the bot is weak? There is psychology here that can be studied. What if there comes a time when aliens of robot form try to invade earth, shall we send our non-chess playing soldiers because our chess-playing soldiers are already terrified hearing about a bot aliens :>)

Just play normally against a bot, if you lose you lose, perhaps the bot's level of play is really above you, there is nothing to be ashamed about it.

@ijh "First, if someone plays a rated match with a BOT account, it could have the same result as sandbagging. For example, if I played a 50 game rated atomic match with a BOT account, my rating would be deflated significantly, and then it would not be fair to my human opponents." If you intentionally lose to a bot, you can be sandbagging. If you lose normally to a bot then it is fine. Your new rating and that of the bot get adjusted to reflect your strengths. "Second, while the purpose of the BOT accounts seems to be so programmers can test the engines that they program, there are many users that do not even know how to program, are not trying to help program any chess engine, yet they are downloading Stockfish, making it a BOT account, and learning how to cheat." There can be a lot of reasons for creating a bot, one of those is what you said. The others can be experimenting book moves. Both bots maybe using the same engine but they can have different opening books and hardwares. LC0 could be interesting specially if your have a good GPU. Some people have a good hardware. Those people that don't have a good GPU but has a good CPU may run Stockfish and try to play against LC0 running on good GPU. It would be interesting to watch such games. Humans vs LC0 on long game like TC 2hrs + 30s inc/move can also be interesting. One of the reasons why a bot is usually stronger than human is that the bot is simply consistent. It does not blunder much. Don't expect too much from humans playing against a decent bot in a short TC (Time Control), just call it a fun play. Eventually humans will be time-pressured which would lead to bad moves/positions. Having a good hardware and opening book can be one of the reasons why bot authors who do not run their own engine created a bot. Perhaps others would just like to have fun screwing proud human players :-) Not all people that learn to setup a bot would learn how to cheat. They may learn how to cheat but does not mean that they will actually cheat. If you think someone is cheating you can report it, Lichess has a system of detecting cheaters. "I think there are multiple changes that should be made with the BOT accounts. #1, how about giving BOT accounts a 2500? rating to start and make it so BOT accounts can only play rated vs other BOT accounts? This way there would not be any loopholes leading to what is currently legal cheating and sandbagging, and the community's ratings would not be affected by this." Not all bots are strong, there are weak bots. CDrill for example has an option to blunder and is rated around 1800. It is intended to play against humans and would try to play like humans, it is still under development. It is indeed interesting that Lichess allows bot now. This opens up a landscape for some researchers to study human behaviours comparing human vs human, and human vs bot performances. Are humans feels weaker if it play against a bot, even if the bot is weak? There is psychology here that can be studied. What if there comes a time when aliens of robot form try to invade earth, shall we send our non-chess playing soldiers because our chess-playing soldiers are already terrified hearing about a bot aliens :>) Just play normally against a bot, if you lose you lose, perhaps the bot's level of play is really above you, there is nothing to be ashamed about it.