I have a 6 year old and a 3 year old who have just started chess. The older one has had a one week experience with ChessKid (the official kid version of Chess.com), which offers a range of bots starting from ELO 100. We immediately noted a great difference between ChessUP AI 1 (supposedly ELO 100) and ChessKid and Chess.com ranked bots (ELO 100 up to 1100 in increments of 100).
I ran a series of AI vs AI matchups. AI 1 will win 100% of games against all the ChessKid and Chess.com bots up to ELO 1000 (apart of one underestimated bot ranked ELO 900), and 50% of games against Chess.com Sven (ELO 1100). Either Chess.com bots are overestimated or ChessUP AI hasn’t been sufficiently dumbed down from Stockfish.
Right now the only options left for beginner kids are integrated Chess.com bots with ELO 250 and 400. I know ChessUP AI performance is expected to step up with future updates, possibly further limiting options for kids. Can you keep kids in mind while revising AI performance? And/or any plans of ChessKid bots integration?
I know this does not directly answer your question but if you use Lichess you can ‘follow’ any bot and then challenge them from the ChessUp app. If you search On Lichess for low-elo bots suitable for children, or something like bot-child that you should find something suitable. I don’t know much about these things so others might have other suggestions. I am sure someone from the ChessUp team will be by with their response.
I think that the weakest bot on the ChessUp board is actually supposed to be ELO 1000 (and not 100) – which your test more or less confirmed . I have a 4 year old son and the one time he played against the bot, we used the integrated move assistance of the board that shows you which moves are bad and which are good when you touch a piece. He plays against me in the same way and I think that he probably learns more that way than by playing against a ELO 100 bot which basically just makes random moves.
No way Level 1 on the Chessup is near 1000 ELO. I played all 12 Levels and all of them, are very weak. You can see a slightly improvement from 1 up to 12, but I think Level 12 could be somewhere around 1000 ELO, but not much stronger.
Level 1 for sure is not higher than 100 or 200, if it is even possible to guess a rating in that range.
Thank you for the Lichess suggestion, I know the Maia bots starting at 1100 but didn’t know there were weaker bots.
Apparently AI 1 is supposed to be ELO 100 up to AI 12 which is ELO 1200 in 100 increments, according to another topic and the ChessUp app interface. I only tested the lower AI 1-3 so far.
I know any ELO below 1000 might be a work of fiction in most adult players mind. I’ve seen the kids play against their cousins and the mistakes were so mind-boggling that I do believe some ELO level lower than 300 do humanly exist. My kids do play with assistance against me but I am trying to diversify the experience and let them sometimes think for themselves without hunting for hints (a hint limit would also be welcomed).
Yes for sure there are a lot of players out there with ELO below 300, not only kids. I have seen youtube videos from people showing their games online and one guy even playing on Chessup 2. It is really fun to watch and sometimes nerve wracking, when you want them to make the obvious move and instead of taking a hanging piece, they hang their queen or just move any random pawn. But hey, we all started somewhere. I find it cool, that we can watch not only the best players in ther world playing, but also those who are beginners and try to improve and learn. And to be honest, a GM watching me play will probably feel the same as me watching a 300 ELO play. And Stockfish is laughing at the super GMs.
In general, the chess.com bots are very over rated for sure. I have seen dozens, maybe hundreds of comments about them on reddit and pretty much every player can beat bots way above their elo. I think the chessup elo is more realistic than the chess.com rating. I’m 1300 chess.com and the level 12 is definitely weaker than me. It started well but seems to capitulate in the end game.
ChessUp has a built in Ai that can go up to ~1200 ELO (soon to be improved).
That is our own chess engine meant to run on the hardware on the board. No connection required.
The app runs stockfish because the hardware on the phone is much much stronger. It is best used for strong bot opponents (it is hard to make stockfish play poorly).
The built in chess Ai makes a great beginner opponent though.
Eventually we will migrate our board levels 1-12 to the app so it is a consistent experience.
But with ChessUp 2, the app usage becomes less needed over time - and we will add more built in bots to the hardware itself.
So I ran another series of AI-vs-AI matchups, this time between ChessUp 2 AI level 1 and a range of bots from ChessKid (ELO 100, 200, 300, 500, 600, 800) and Chess.com (Maria ELO 1000). 22 matches in total.
My rough estimate is that ChessUp 2 AI level 1, advertised as being ELO 100, would be around ELO 900-1000 compared to ChessKid/Chess.com bots.
Win/Lose/Draw rate:
Against ChessKid 100 3/1/0
Against ChessKid 200 1/1/1
Against ChessKid 300 3/1/0
Against ChessKid 500 2/1/1
Against ChessKid 600 2/0/0
Against ChessKid 800 1/0/2
Against Chess.com Maria 1000 1/1/0
ChessUp 2 AI losses against ELO 100-500 were exclusively through Fool’s Mate. It never lost in endgame.
Draws were mostly through totally avoidable threefold repetitions. Both AIs seemed stuck in an infinite loop of some sort with many other options available to avoid the repetition. In all cases, ChessUp 2 AI had a significant material advantage (up to +33 in one game)
Chess.com bots could be overated. But it does tell me ChessUp 2 is currently lacking on the weaker range of bots. The kid is still struggling in beating ChessKid bot ELO 300 without assistance and I think the fashion of going by 100 increment is the right way to go, as was intended with both ChessUp 2 AI and ChessUp app AI.
Hope this helps with the planned revision of AI. Think of the kids!
This time around with built-in AI. ChessUp 2 built-in AI level 1 I would estimate at ELO 900-1000 when compared to ChessKid/Chess.com bots. And ChessUp app AI level 1 around ELO 1000 by ChessKid/Chess.com standards.
In the meantime I have found a new way to teach my kids about chess. I select a Chess.com bot between my ELO and my kid’s ELO. We take turns playing our side. No hints since they are not implemented in Chess.com yet. No alpha playing allowed, but on our respective turns we can explain what we are trying to achieve and what we expect the bot to do. Cooperative chess in the making!