A humanlike playstyle chess bot for every level player?

Imagine that you are looking for a chess bot to practice against. However, you only find bots that blunder everything and make zero sense, or bots as strong as Stockfish that crush you every time without caring about your gameplay experience. If you've had enough of this, you should try SE-MCTS bots, which play exactly like real human players at any level!

We've all been there. You play against a Level 3 engine, it plays like a Grandmaster for 15 moves, and then suddenly hangs its Queen for absolutely no reason. It doesn't feel like practicing with a human; it feels like playing a slot machine.

Why traditional chess bots strange?

Modern chess engines are built for one goal: play the objectively best move. They are optimized to calculate deeper and more accurately than any human ever could.
But here’s the problem: when these engines need to simulate a lower rating player, they don’t actually change how they “think.” Instead, they simply reduce their thinking time or limit their search depth. And that’s where things start to feel off. Limiting search doesn’t make an engine think like a 1200-rated human. It just makes a super-strong player think for half a second and then move anyway. The result? Fifteen moves of flawless opening theory, followed by a completely senseless blunder.
That’s not human chess. That’s just a genius playing while distracted.

How SE-MCTS works

This is where SE-MCTS comes in to fix the mess. SE stands for "Strength Estimator." Instead of forcing a supercomputer to play dumb, we trained on millions of real games from everyday players on Lichess. During its search, SE-MCTS uses "dynamic strength adjustment." It actively estimates how a player of a specific rating would evaluate the board. It doesn't just randomly weaken an engine—It aims to replicate the tactical logic and natural mistake patterns of real players at your specific rating. No artificial piece hangs, no sterile computer moves. Just pure, authentic human playstyle.

Try it yourself (and maybe grab some free ELO(X)!)

This is an ongoing research project aiming to capture human cognitive blind spots, and I need your help to test them out! I've just deployed the full lineup of SE-MCTS bots on Lichess. And Here is a little secret for you: because they just went online, they are currently in the process of rating convergence. This means their Lichess ratings haven't fully stabilized yet. Some of them might be a bit overrated or still adjusting to their target strength. If you want to catch them before their ratings fully stabilize while helping a CS student gather research data, now is the absolute perfect time!
Choose your sparring partner here:

• ELO 1000 - 1199: https://lichess.org/@/SEtheChessBot8
• ELO 1200 - 1399: https://lichess.org/@/SEtheChessBot7
• ELO 1400 - 1599: https://lichess.org/@/SEtheChessBot6
• ELO 1600 - 1799: https://lichess.org/@/SEtheChessBot5
• ELO 1800 - 1999: https://lichess.org/@/SEtheChessBot4
• ELO 2000 - 2199: https://lichess.org/@/SEtheChessBot3
• ELO 2200 - 2399: https://lichess.org/@/SEtheChessBot2
• ELO 2400 - 2599: https://lichess.org/@/SEtheChessBot1
• ELO bruh : https://lichess.org/@/SEtheChessGod

Check out the full SEtheChessBot lineup, and let me know in the comments: Do they actually feel like real humans? Which bot gave you the most realistic sparring experience?
(Developed by @Dr_Kiwi. Read more about the SE-MCTS research here: https://rlg.iis.sinica.edu.tw/papers/strength-estimator/)