Why do I handily beat bots =< 1300 Elo, but lose 24hr games to players of 600 Elo?

Hi,

New here, first post, don't know if this question gets asked a lot (apologies if it has been.)

When I play the bots, I can win up to ~1300. I took a few goes at learning how to beat Nelson, for example, but I can beat him almost every time now. Up to ~1100 I managed to beat the bots mostly first go. In 10 minute games I've peaked at 1063 (now I'm lower, 960ish) BUT in 24hr games, I lose to players of 600 Elo. I don't get it. If I review the bot games, sometimes I win after even 3 blunders and 5 mistakes. I recently lost a 24hr game against a 660 player and the review had me down as no mistakes, no blunders.

My current theory is that it is all about continuity of focus. In a short game, the game is in my head in totality, from start. to this point, including the reason for the choices I made. But in 24hr games, I look away, sometimes for 20 hrs, then come back, and I have no context. Is the trick to "re-play" the game up to that point, and try and figure out what I was thinking?

Many thanks in advance if you have experienced this anomaly and have tips to make progress

Any bots under 2000 on chess.com will sometimes make some horrible blunders, also when you are playing a real player which is a lot rating lower then you, never underestimate them. I am 1900 rapid and lost to a 1100 4 times in a row…

Also 900 rapid is a good start.

Having played much more now, I have to ask: In what way is Vinh or Nelson etc 1300? And so on. I have played 800s, 900s, 1100s in 1 day chess, where a single blunder, and accuracy <80% is a lose. And yet I can beat a 1400 bot whilst making 6-7 blunders in one game. So why are the bots rated in such a way?

It could be the humans I am playing are underated (as in they are on their way up to a their current rating, passing through 800 or 1100) but isn't it more likely that the bots are just massively over rated? Isn't it misleading?

I think the bots are massively over-rated. Even up to 1400 they play completely random moves which resembles a chimp playing. Also, they frequently just leave a piece to be taken for no reason at all - like just move it to avoid being taken by my pawn?! And this is not in complex positions. I have beaten up to 1700 bots as both black and white. Today I barely mustered a draw against a 950 human. Go figure. I've started playing the 2000+ bots figuring it will prepare me for 1100-1200 rated human beings. (?) So far zero wins but learning lots.

I agree. Many years ago I played some over-the-board tournaments and got my official rating. Now, my chess.com rating is FAR below that (on chess.com). Granted, there is an obvious drop from not playing for so long, but my subjective experience is exactly as you describe it. When I play someone who is 1000-1100, based on my past experience, if feels more like I'm playing 1300-1400. This is entirely subjective of course, so go ahead and dispute it. I'll stick with my gut feeling on this.

Question: Why would the ratings on chess.com be deflated? How are experts and beyond, who play on chess.com, rated in-the-real-world vs. here online?

I have an ELO of 800 in Blitz and 1100 in rapid chess in chess.com. I have an ELO of 1200 and 1400 in Lichess and an ELO of 1400 and 1600 in HIARCS. I beat chess.com bots of 1400 and 1600

I think the Chess.com ELO is a bit distorted in Blitz given the habits of players. Meaning, in weekends is much harder to win while in weekdays I win many more games! You see many players trying new traps and tactics and, once you know the trap and avoid it, easily winning against them. It looks to me like we have many players that are playing to test their one trick pony newly learned skill (meself included) and that distorts the sample. On Monday, they learn the trick and win 100 games. On Wednesday, everybody learned the trick and they lose 100 games. I guess it is part of the ELO system when you have many occasional and opportunistic players (meself included) in low and medium levels.

More on this...some players are just really good. I played a 1400 bot, and won! But our accuracy (me and the bot!) was in the mid 60s%. But I lost to human, who waas the same exact 1037 rating as me, and I played 90% accuracy, they played 95%!

Yeah, the bots are fun, but they are in no way representative of the Elo of actual humans you play on this website. I can't beat a 900-1000 rated player with less than 75% accuracy

I paid extra so that I can review all of the games that I play. At first I assumed that the bot would always play at their rating and so I only tracked my game rating. But now I also track the bot's rating. What I have found is that Laura's (1100) individual game rating varies from 100-1500. I also found that my rating ranges from 100-1100. against Laura. My rating seems to correlate with hers. When she is good I rate higher, when she is bad I play bad. What I do no know is if the bot is programed to adjust to my play or if I play better when it plays better.

Bc bots are programmed to make as many blunders as a 1300 player would), but they don't know how to make them in a 1300 style. Their blunders are way easier to spot. Consider bots as a simulation of how your games would go if all blunders were obvious to you

Thanks lassus_dinnao yes their blunders are really obvious. Like leaving a queen to be taken. But what is interesting to me is that their rating is always close to mine. I have found that my game rating is as variable as theirs. So my question is: are they programmed to match my level of play or do I play better when I am challenged ?

I don't know

Thanks Daddy_Chillimao: I am trying to get good and I probably analyze too much. I practice playing the bots and people online and in person at different chess meet-ups. Glad to see my rating reach 950 against a bot but frustrating to see it drop to 100 on the next game. Just have to persevere until I get more consistent whether playing people or the bots.

Don't worry about the bots, their ratings are horribly inaccurate. I just played two 400 elo bots (I'm currently 1500 rapid) and although I won both games, they were surprisingly decent. After reviewing each game, both bots played with decent accuracy in weird positions (400 bot openings) and their game "ratings" were incredibly high for 400 rated opponents. The first opponent played like a 1200 and the second opponent played like a 1600 according to chess.com, and that's accounting for their low ratings which apparently skew the ratings a bit.

Also, I've beaten the Alexandra Botez bot, but not as often as I've beaten real humans at similar ratings. Her bot is incredibly accurate until it isn't. It makes the games very hard unless you can survive until the required blunders are inserted.