Bot Ratings

@JHACKIL Yeah so you expect bot ratings to relate to chess.com ratings because they're on same platform? I guess it could be done, we can rate bots against players, problem is chess.com ratings aren't accurate too (that's besides cheating issue, players would cheat against bots and should cheat detection team be distracted by these cases? if ignored this would lead to inaccuracy of bot ratings and also we can't force people to treat bot games seriously and play at full strength. many problems). I don't understand what claims you refute and why since I'm just sharing facts. What's not pragmatic, why not pragmatic? I also asked you to read about Elo and understand what that is. You refute this too? But it will help you to understand couple of things. Like you're claiming that you've defeated a 2000 bot. But you didn't specify attempt count. Time control. This is all important, you can't ignore that.

Bots traditionally are rated in relation to bot rating pool and initial bots were rated by playing games against FIDE rated GMs, so bots have received FIDE ratings. It happened in early 2000s. All bots since bear that rating because rating of new bots is determined by win ratio against older bots. Bots don't need to play humans anymore to get rated. If there is a pool of player with some rating system and ratings of players from that pool don't correspond to ratings of bots, that's not a problem of bot ratings at all. You can argue that FIDE ratings have inflated or deflated since early 2000s, but that's wasn't significant.

There is one potential source of inaccuracy: time-dependent thinking. Some engines have it, some don't, for most you can enable it. For those who have it, precise rating is not possible. I mean you can rate such an engine using another not time-limited engine with known rating. But you can't expect that engine to perform the same on a different machine (with more processor cores and faster processor speed). Komodo when used locally can be set up without time limit. So first step for you to start making some serious conclusions about bots is to start using Komodo locally in your OS (Windows, Mac, Linux, whatever you have). You don't have to use Komodo specifically, could be some other engine, like Stockfish. You can set Stockfish rating precisely from 1320 Elo and higher. In some versions 1350+. Komodo has NO rating setting, only guidelines I already explained.

If a bot's rating is stated on its profile, in the same manner in which it is presented on a user's profile, it is reasonable to expect that the ratings of both the bot and the user are directly comparable. Thus, I believe that bots should be allowed to play games against users in rated chess. With such an addition, users would not be required to play strictly rated against bots; as unrated games are available against humans, the new addition should allow that unrated games are available against bots. Additionally, I uncategorically stated that the chess game in which I defeated a 2000 rated bot is present on my public profile. By no overstatement may I be compared to such a player. My argument is simply for that of a system in which chess.com bots possess a comparable rating to that of a user; the bot ratings on this platform are significantly distorted. I can sacrifice my queen for no material in exchange against Nelson-BOT, and I may emerge victorious; Nelson commits blunders of such magnitude that he cannot possibly be of the level of a 1300 ELO player on Chess.com or FIDE. (At your request, I will provide the described game) If users are given the choice to play rated games against bots, the ratings of the bots must eventually reach a value that is comparable to a human player of equal strength. Feedback from these games may be used by the administration of Chess.com to potentially create more accurate computer ratings.

The user "xtreme2020" has stated, "How about the bots lose and gain elo but the player doesn’t? That seems like the best of both worlds, you get accurate bot ratings but no rating inflation." This is another possible solution; although, in order to affirm the validity of the bot ratings, it may be wise for the administration of chess.com to eventually allow players to gain or lose rating from their bot games.

"

I understand that people could receive unrightful ratings based on their games against bots, but:

(1) the ratings of bots will promptly reach their true value (vast amounts of people play against bots each day)

(2) the inflated ratings of the players will also eventually reach their true value (assuming the user is active on chess.com)"

Or, your previously stated solution could be implemented.

If a bot's rating is stated on its profile, in the same manner in which it is presented on a user's profile, it is reasonable to expect that the ratings of both the bot and the user are directly comparable.

Yes, it's reasonable to expect, but these expectations cannot be fulfilled. No one needs even more rating inaccuracy and rating manipulation and another thing: chess.com ratings are so far from the typical OTB ratings, for many many reasons.

Thus, I believe that bots should be allowed to play games against users in rated chess. With such an addition, users would not be required to play strictly rated against bots; as unrated games are available against humans, the new addition should allow that unrated games are available against bots.

Just think how the team would approach all the possible manipulations. Too much effort, impossible.

Additionally, I uncategorically stated that the chess game in which I defeated a 2000 rated bot is present on my public profile. By no overstatement may I be compared to such a player.

Please don't ignore my message about possible thinking time limit and the fact that chess.com bots run locally in your browser using resources of your PC, or run locally on your smartphone using resources of your smartphone. Consistent play strength is simply not possible. Also please don't ignore my question about how many attempts does it take for you to defeat a 2000 rated bot here, I mean what is the winrate? And also please don't ignore my question about time control you've used. Because all this means a lot when we compare ratings. Please try downloading Komodo and playing against it in any chess UI desktop app without time limit on thinking, just set the skill level to 14 (roughly represents 2000-rated player at 5|5 Blitz time control) and play 5|5 Blitz or 10 min per game with it. Play seriously at least 10 games, count all losses and wins. I will help you to calculate your FIDE-like rating then. Chess.com ratings are not realistic.

My argument is simply for that of a system in which chess.com bots possess a comparable rating to that of a user; the bot ratings on this platform are significantly distorted.

Player ratings are distorted even more.

I can sacrifice my queen for no material in exchange against Nelson-BOT, and I may emerge victorious; Nelson commits blunders of such magnitude that he cannot possibly be of the level of a 1300 ELO player on Chess.com or FIDE. (At your request, I will provide the described game)

I believe you, look up my game against 1800 bot, Wally I think. He blundered so badly. But above I explained why this might be happening. Try on desktop.

If users are given the choice to play rated games against bots, the ratings of the bots must eventually reach a value that is comparable to a human player of equal strength. Feedback from these games may be used by the administration of Chess.com to potentially create more accurate computer ratings.

Already answered above.

According to a chess.com support bot, "The strength of Li-BOT, or any other bot on https://Chess.com, doesn't change based on whether you're playing on a desktop or laptop. All our bots, including Li-BOT, are powered by Komodo chess engine. The bot's strength remains consistent regardless of the device you're using to play. When you play against a bot on https://Chess.com, you're essentially playing against the Komodo engine, which maintains its set strength level across all platforms. So, you can expect the same level of challenge from Li-BOT whether you're using a desktop or a laptop." I guess that settles the great debate. I still do not believe that Nelson is 1300.

@JHACKIL,

No, predefined agent response doesn't settle any debate. Go now start a game against bot, then shut down your internet connection. And keep playing. Observe bot is still making moves. How? In your browser's network tab observe komodo.wasm file was downloaded. That's the answer. Yes, that's the komodo engine, true. And that's the proof that it runs on your device, not on server. And there's no guarantee it's not configured with time-limit per move. With time-limit, moves will be weaker on a weaker PC. It is quite possible. You can't be sure. Try Komodo separately on your Desktop in a Desktop chess UI app. This makes total sense because users have reported many times on this forum that bots play weak moves on weak devices. I see no reason to trust blindly that there's no time limit. It's very easy to set up local desktop chess UI to play against Komodo and configure it, I could help you with that. LucasChessR has it built-in and has access to all the settings and by the way it's way more settings than chess.com allows you to use. Skill Level 14 is what should represent 2000 Elo @ 5|5 Blitz so all you need to configure really is the skill level but you can also try personality setting.

Okay. I think I'm done with the debate. What are your final thoughts? Additionally, can you help me download a 2000 rated Komodo engine? Thank you.

Sorry for posting twice. It was an accident.

What's your OS? Windows?

nvm their package is for all systems:

https://komodochess.com/pub/dragon.zip

you will see two exe files one avx, other non-avx

run avx, if it doesn't crash, type isready

it must respond "readyok"

if not, try non-avx

now a manual how to add it to LucasChessR:

download/install LucasChessR, run

play - play against an engine - configurations - new
file picker will open.
locate downloaded exe file (avx if it worked for you)
set everything at 0, click accept
to the right you will see configuration panel
scroll down to skill, set to 14 in "Value" column (2nd)
it's 25 by default
rough formula is Elo = 250 + skill * 125
except for two last skill levels where increment is 250 not 125
and for highest skill full strength depends on how many cores you enable
anyway, click close
then click on big button under opponent section, external - dragon by komodo chess 64-bit
accept
you are ready to play!

p.s. if you need to change settings later, going to configuration and editing it might not work, go to advanced tab instead and scroll down to Skill Level option. You can also try personalities

They made the bots trash and overrated so that you can feel good about yourself when you beat some bot 500 points higher than you

That answer really isn't accurate. Device types, CPU speed, available resources, other processes, etc can impact how well the engine code runs and make it play weaker than normal.

Thank you for the input.