Why are bots always easier to play against than people with the same ELO?

I was noticing this for a while - bots are easier to play against than people (same ELO). For example, if you go and play a 800 ELO bot and then play against a 800 ELO human, the human will be harder to play. Here, for example, are analysis of my recent game with a bot, which level I out to "Intermediate II - 1200 ELO". Though my ELO is around 600-900, and I not an easy time gaining more ELO, the bot was easy for me to beat. To notice, I did NOT reverse any moves or take any hints.

https://www.chess.com/analysis/game/coach/5745005/review

Or https://www.chess.com/game/coach/5745005

Same happened when I played a 1000 ELO bot - a bit less predictable then the coach, but still easy to take advantage of, even when I blundered a few times.

Maybe that happens because of the algorithm — to simulate a lower skill level, the bot intentionally makes mistakes from time to time, the more his ELO is, the less it makes obvious mistakes. I dived into the topic, - and it appeared I was partially right:

Sources like:

https://maximagamingstudio.com/how-ai-personalizes-difficulty-levels-for-each-player/

https://www.houseofstaunton.com/chess-blog/how-does-a-chess-bot-work/

https://palospublishing.com/ai-driven-game-difficulty-scaling/

https://aicompetence.org/train-ai-for-perfect-game-difficulty-balancing/#:~:text=AI%2Ddriven%20difficulty%20balancing%20is,creates%20a%20fluid%2C%20personalized%20experience

and other proved that bots don’t naturally have a skill level. A chess or gaming AI can usually play far above any human level if left unrestricted. To make them “fit” a specific ELO (like 800, 1200, etc.), developers adjust the algorithm so it intentionally miscalculates, plays suboptimal moves, or randomizes decisions. However, the mistakes are often mechanical or predictable, not the kind of nuanced errors humans make under pressure or due to strategic misunderstanding.

As a result, bots actually might feel easier to play against even if they’re rated the same, because they do not have human-style pressure and unpredictability.

What should we do in this situation - while bots haven’t learned how to bluff or blunder naturally yet? They still feel way easier to play against than people with the same ELO.

Huh interesting because chess.com elos of bots lie...

I am 1000 elo. But I beat Samay Raina on my alt account.

But when I clicked Game review,

Samay played like a 1200 and I played like a 1700!!

Bot ratings are very inflated and the game review rating is meant to make you feel good

Yes bots are easier to beat than humans with the same ratings. I beat Francis Bot and Danya Bot on my iphone and I suck lol. All i did was just kept playing over and over everyday until I won. Seriously sometimes they make weird moves. Even chess streamers on YouTube that play against bots notice how easier they are compared to humans. I guess the bots are just programmed to make dumb moves lol

Bot ratings are strange, and they don't seem to be linked to player ratings but more bot vs bot matches. Almost everyone can beat Martin (though I did lose against him a couple times in my first weeks of learning chess haha) even if rated under 250, but GMs and influencers on YouTube tend to be very close to their own bot levels (in Rapid/Blitz at least) - there are YouTube videos of real players vs their bots. I think I also read a comment here a while ago, maybe by Martin Stahl, that said bot levels depend a lot on your own equipment too - players playing on a phone will play a weaker version of the bot than a player on a PC.

the ratings are suspiciously given to the bots who are actually 2000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000006 rated

What I think is it's useless to play below maxim bot. At the end human perspective is very distinct from bots. Also just because someone is at 800 ELO doesn't mean they never reached 1200 level. Bots should be seen as the way to found best response for all the moves also know what's the best way they can response. I also started with with minimum level then go for maxim recently. Some players are just off-book players and they still manage to defeat the book players.

It's because its harder to make a bot that makes mistakes than one that is stockfish level (and making an elo that actually matches performance)

You'd be surprised just how many titled players are playing under accounts rated around 400.

It's a lot.

If you check the code for every bot you will realise that if you play the opening better than the elo of the bot, it will play worse throughout the game

(shoutout to my friend for telling me this)

Very well said

Do you remember the mittens tho? LMAO

I don't play bots, but people have been complaining about it for years. Bots are overrated and chess.com does not seem to want to do anything about it. Why? Your guess is as good as mine.

Bots below Stockfish's rating are programmed to randomly blunder. A higher rating just decreases the rate of the blunders and (usually) makes them more likely to be positional errors, which are generally much trickier to spot than the tactical errors low Elo bots tend to make.

Humans obviously cannot be programmed. Therefore, they can't play a dumb move simply because they were coded to do so.

Yeah! I noticed that as well! I can easily beat 1400 boys but I’m having a hard time playing as 900 rapid!

Human vanity.

Meaning, bots are a marketing feature. They're assigned an Elo rating by the product managers of this website. Bots don't earn a rating. It doesn't fluctuate. So, the product managers assign an Elo rating that flatters human vanity. Makes the human player think they're a stronger chess player than they actually are.

S-s-s-such beautiful words!