How much tougher are human players compared to computers of the same elo?

I've been playing bots to shake off the rust and get a feel for the game. I haven't played in years and I was always a casual player aka I have never picked up a book or learned an opening etc. I just played quite a bit as a kid. Reading through the forums its apparent that people at even the 1000 elo range seem to doing things like memorizing openings etc. I realize human players are very different from bot ones. How much tougher are humans of a similar rank to the bots on here?

Humans are way tougher

That's what I figured. I saw some posts by people in the 1000 elo range and they were talking about reading books on chess and learning openings which is way beyond where I am at as a casual player. I can mop the floor with an elo 1000 bot, but I have a feeling an elo 1000 player will destroy me. I guess there is really only one way to find out.

Yeah, I could tell just from the games I've played with a handful of bots that they play very inconsistently. At around 1000 to 1100 they seem to play relatively soundly, at least from the standpoint of a newbie causal player like myself, and then all of a sudden they just blunder a piece so blatantly it just doesn't make sense. This means you can beat them by just defending yourself and poking at them until they do something insanely stupid.

Try bots 2000+

Oh, I get straight up destroyed by even a 1300 bot. I have no doubt that their are bots capable of wrecking pretty much any player on earth let alone a casual newbie like myself. I was just wondering how their elos and play styles compared to real people's since I've been wary to try playing human players after seeing a load of posts by people in the 1000 range talking memorizing openers and stuff and the bots at this level do some pretty weird stuff sometimes which made me wonder if their elos were not very well tuned.

Meh, you don't get it at all how ratings work, what ratings are, etc.

I could tell you my cow has 1200 Elo, would you believe it, you'd be a fool. Same fool as when you believe these bots ratings.

In short: bots don't have got any actual rating, but some number someone wrote next to their name. That'll be an "evaluated rating", which is no rating at all, but usually serves the purpose of pairing players, including unrated players, in a swiss system chess tournament.

If you take chess engines (or softwares, or programms) from the 90s, they would eventually get some of an actual rating, by playing a World Chess Softwares Championship. Still something between an actual rating and an evaluation, since they'd play no human rated players in the process. Usually, at least.

And, since these engines were playing at their top capacities, one could argue an engine rated 2200 was actually playing at a strenght of a human player rated 2200. With some differences in the type of faulty moves and such, but still.

The bots you're playing now, that's a different story. They are the so very strong engines everybody knows, but somewhat depleted.

But how do one deplete a chess engine, and manage to have it play weaker than it's normal strenght? Well, one has to instruct the engine, in such ways (like) "Instead playing best move you find every move, you'll play second best move every four moves".

This doesn't really produce a type of bot that will play in a manner comparable to the older and weaker engines from the past. It'll produce a weird play, with bizarre moves every now and then, that no human and now chess software would do.

So, in conclusion, forget about that crap you're doing with the bots: it's not teaching you anything, You can't know if you're playing good or bad when playing against these frankenstein bots.

Better find engines from the past (some I had played at about 1600) than play these bots that are good for nothing.

The engines from the past (Novag, Mephisto, etc), you would beat them out of understanding chess better than them, not out of random silly blunders implemented in their play by some depleting instructions...

As of right now, it would be nearly impossible for a human to win against a similarly rated bot on the high end, and vice versa on the low end. The reason for this is simple science. AI’s are much smarter than humans, but if they have the same elo, it doesn‘t really matter. What does is the efficiency of thinking. Humans can inference extremely well of of one point of information, and don‘t need many, but bots need millions of data points. A dumber com is much easier to beat because it‘s given low processing power and data points, but one with much more of both ultimately functions better. Therefore, the higher the elo on the bot, it gets progressively harder for humans of the same elo to beat, because humans make mistakes.

AIs are no chess softwares and can't play chess unless they use a sub programm dedicaced to chess.

I challenge you to teach the rules of chess to any AI chatbot online. You'll fail endlessly.

Human ratings here are set by issuing a certain number of points, and letting them shuffle up and down the pile of people according to who wins against whom. These numbers therefore depend on the general level of ability in the pool, how many points are allocated per person entering the pool, and where you are in the distribution of players.

Bot ratings are just a number written next to the bot's name, but they don't need to be. If bots played "proper" time-constrained games against time-constrained players, there is no reason why their win-rate and the player's ratings against whom they win or lose shouldn't be allowed to dictate the bot's ratings, so that the bot rating system was brought in line with the human system. But the site has chosen not to do that. Therefore you should treat bot ratings as a totally different scheme, merely indicating the relative strength of the bots amongst themselves.

The third, odd, situation is the rating estimates given by the review engine based on individual games. This is a third rating system again, also completely different to the other two.

So, for example, I am a 450 player against humans, but the review engine generally reckons I'm around the 900 level, but I can win against bots reckoned to be 1200. I think my experience is probably typical. Three systems, not comparable in any way.

That's unfortunate. I had a feeling these bots were not vetted against actual players in order to determine their ratings. I had a feeling the strength of the low end bots was radically overestimated.

You don't have got a strenght when you are forced to play worse moves than what you actually can. Sometimes, the second best move is a losing move, that allows to take the Queen or to checkmate, which is what makes the thing absurd (to force softwares to play worse moves than what they can).

Human players are generally tougher than bots of the same Elo due to their unpredictability, strategic depth and psychological factors.