Are the bot characters’ ELO ratings accurate?

funny i just posted the same question 2 minutes ago, look 2 spots below. i beat 1500 antonio easily but struggle against 1100 real ppl

Bots are a weaker than real players. Against a 1300 you can expect to win 15% of the time if you are 1000. If you have a win rate of 33% against a 1300, it means that your ELO is higher than the actual one (near 1200) or that the rate of the BOT is 1100 and not 1300. To check this, what don't you play against strongest opponents (1200 for example) and you will see... 

That's just because with real person you have time constraint, with a bot, you don't have time constraint.

Yesn't

It’s more than just time constraint.  I can win fairly quickly and consistently against any of the intermediate bots (1000-1400) and I am 1098.  I have beaten up to Arthur (1700) however I do find any bot 1500+ to be less than 50% winrate.

I just played antonio bot (1500) 100 times in 10 minute chess and saved them to my archive so I only have to write down results once, I won 58 times and lost 33 times and 9 draws I think his rating is super accurate for 10 minute if I played him in daily I would do a lot better because my rating is a bit higher in daily with more time to think.

You may have been playing with a ten minute time limit but the bot played the same as it always plays, whether you've signed up for bullet or a two hour limit.

Look at this whopper played by Antonio in one of your last games.

I don't think many 1500's would play such a move.

The 1500 and under bots can blunder their queen...
Some comments on ratings and bots

Sometimes it feels like real people do that too in daily chess.

Rd3 going after the h pawn? I make a lot worse mistakes and blunders in 10 minute.

I have't played in a chess tournament for many years, but in the 1980's and early 90's anyone rated below class C in the USCF was considered a weak player.  I got my rating up to 1450 in over the board play which got me over the threshhold of being a class D player.  Class C was 1400 to 1600.  I can't imagine any chess program really being that accurate.

http://www.uschess.org/archive/ratings/ratedist.php

I'm confused now
How can bots be related to elo? For example, I'm around 400 rapid and yet i can easily beat bots that are 1200

<

they are 1200 ELO for classical chess (slow chess) maybe, don't you think?

This bot was assigned a rating of 850. 

That doesn't make sense. If you move instantly and can play 1200 when the opponent gets 90 minutes on the clock, you probably play at 1800 when they have 10 minutes on the clock. If 1200 is that slow chess rating, then their bullet rating should be like 2000+.

from my experience bots are absolutely not their ratings, it's almost always the case that they're less than their ratings say, this goes in the other direction for mitten, the 1 elo engine cheater which my sf 15.1 engine needed 83 moves to beat at 40~ depth 

The bot ratings aren't accurate to human ratings. Antonio for example plays closer to a stronger 1200 on average.

They better indicate the difficulty, as higher-rated bots are almost always stronger than their lower-rated counterparts, so you can expect to struggle more against higher-rated bots than previous ones, but not in the same way as struggling against players of that rating.