So I just played the Joey bot on a browser and it seemed simplistic by far. Not making the best moves every time, not being overly aggressive. I saw back in the day there was an issue with the iOS version of bots outperforming while the online version was normal. Wondering if that might be an issue again, only with the Android.
How are the bots actually rated?

I guess their rating is based on how many centipawns i player of a certain ELO loses by move and also the precision, on average. And then they program the bot that has that same elo to lose that same amount of centipawns in general and perform with that given precision. That would make the bot "correct" their precision sometimes by making random absurd mistakes or blunders, or also random great moves. That's my guess

I too think it was difficult compared to some others similarly rated bots. But I was able to win it on Android phone.
It's odd, I played the one online last night and it was a simple game for me to win. Just played one today and Joey was at 1900 and I was around 1090. I know a poor game from me will boost his rating, but sheesh I feel like when I play on my phone any game I play is poor against this bot.
There are three completely different rating schemes going on here, and there's no relationship between them. There is the human scale, which depends on who you win against, so it's a reasonably stable algorithm dividing a more-or-less fixed number of points across a large population depending on actual games. On that scale, I'm around 400-500, though I've peaked just over 700 on a lucky streak, and regularly drop to the 200 region.
Then there is the estimated rating that the review engine gives you after playing an individual game. Obviously it's not robust because it is a single game. I also suspect that it is calculated in some way that uses more than just your moves. If you're playing around the 200-300 region in human-ratings, and you play with 90% accuracy, it will tell you that your estimated rating is about 1000, but if you're playing around 550 region in human-ratings and you play with 90% accuracy, it will tell you you're about 1300. In general, it estimates my games around 1000, i.e. about twice my human rating.
And then there are bots. I can win against ghost-giri who's advertised as 1350, but I would be totally flattened by a player with a human rating of 800.
For some reason, my ghost-giri game isn't in my archive. The last computer game in my archive is me (at the time rated 301 in human terms) versus Mateo (rated 1400 in bot terms). I won, with 90.2% accuracy versus Mateo's 82.3%, and the review engine rated me at 1600 (wo! more than FIVE time my rating at the time, more than twice my highest-ever rating, and more than three times my realistic human rating), and it rated Mateo at 800, only just over half what Mateo is supposed to do.
Go figure!
I've been very curious about the actual ratings for the chess bots on here. I typically play against the bots at around a 1500 rating and win. However, recently every time I play a chess bot at 1500 and look at the game review I play around 1450-1550 level and the bot is playing at 1800+ level. Is there a reason for this?
This has really become a question after I've been able to easily win against the Logic (Ultra 85 by Logic) bot, but when I tried to play the new Joey Votto bot I lose every round like I am not even trying to play. The game review still shows I am playing around a 1500 rating, but the bot is playing at an almost 2k rating every game. So, not trying to complain, jut trying to figure out how these bots are being rated...