The Human Perspective: Analyzing Positions with Maia 2
How predicting human candidate moves calibrated to your rating provides essential context alongside engine calculation.
Structured chess puzzle training
ChessKinetics is a web-based puzzle trainer that replaces open-ended solving with bounded sets, gradual difficulty scaling, and mandatory recovery breaks. Built on sports science principles, it paces calculation practice to prevent mental fatigue and rating tilt.
96,000 Curated Puzzles
Filtered from real Lichess games across opening, middlegame, and endgame phases.
900 to 2700 FIDE Range
Organized in 50-point difficulty buckets with 2,000 randomly sampled puzzles per bucket.
Free Daily Practice
Every feature is available for free for up to two training waves per 24 hours.
Standard puzzle trainers serve an unending stream of positions and update a visible rating after every attempt. Without a natural stopping point, rising fatigue or a single mistake often leads to rushed calculation, guessing, and chasing lost points.
In physical athletics, training is rarely pushed to exhaustion in a single stretch; instead, workload is divided into sets that alternate effort with recovery. ChessKinetics brings this structure to chess calculation, pacing difficulty and enforcing rest before frustration sets in.
Each wave steps through up to six puzzles: three easier positions to build rhythm, two closer to the target rating, and one final challenge slightly above it.
A mandatory two-minute break runs after completing a wave or making a mistake, serving optional easy puzzles while giving players a pause to clear their head.
Mistakes cut a wave short so players do not grind through fatigue. Target difficulty updates only between waves, rising after solving all six and stepping down only if the first two are missed.
Puzzle ratings remain hidden while solving and only appear afterward, keeping attention on the board rather than the number.
The Weekly Personal Record tracks the highest-rated puzzle solved in the last seven days and never drops after a mistake.
Puzzles are filtered for rating stability and sampled at random rather than by popularity, avoiding a bias toward flashy sacrifices.
Post-puzzle review pairs single-line Stockfish 18 evaluation with Maia 2 neural network predictions, showing what human players at that rating level actually consider.
The free tier includes every training and analysis feature for up to two waves per 24 hours, enough for a consistent daily practice habit. An optional monthly subscription removes the daily limit for longer sessions.
How predicting human candidate moves calibrated to your rating provides essential context alongside engine calculation.
Why losing a game or puzzle often leads to rushed play, and how mandatory breaks help interrupt the cycle of loss chasing.
How 96,000 puzzles were filtered, bucketed across the 900 to 2700 rating range, and sampled to prevent pattern-recognition bias.