The checkers engine enforces two rules that most casual players skip. If a capture exists anywhere on your side, you must take one. If several capture chains exist, only the longest chains are legal. Read the jump code and the opponent becomes predictable, because its replies to any sacrifice are forced.
How the jump rules work here
Your red men sit on rows 5 to 7 and move up. The white AI men sit on rows 0 to 2 and move down. A capture jumps an adjacent enemy diagonal and lands on the empty square beyond it.
The depth-first search then keeps going. If your landing square enables another jump, the chain continues within the same move, and every enemy piece you pass over is removed at once. When a man reaches your back row mid-chain, it crowns and the same chain continues with king movement in all four diagonal directions.
Two consequences follow.
1. A single move can remove several pieces, and the engine counts it as one move.
2. A crowned piece keeps jumping in the same move, so coronation can be the setup for a second capture.
Why the forced capture helps you
Because the AI must capture and must prefer the longest chain, you can compute its reply exactly. This converts checkers from a guessing game into a trade calculator.
Use this sacrifice pattern.
1. Offer one man so the only available AI capture drags an enemy man into your diagonal.
2. Verify the forced landing square sits inside a chain for your next move.
3. Take the multi-jump and collect two or more pieces for the one you gave.
Check the longest-chain filter before you commit. If the AI has a two-piece chain elsewhere, it takes that instead of your offered piece, and your trap silently fails. Count every capture on the board, then move.
What the AI values, in points
The evaluation assigns numbers you can play against. A man is worth 100 and a king 300. Advancement pays 5 per row. Center squares inside rows 2 to 5 and columns 2 to 5 pay 10 for a man and 20 for a king. A back-row man pays 15. Each legal move pays 5. Every piece on columns 0 and 7 pays 8 because edge pieces are harder to trap.
The AI searches two plies on easy, four on medium, and six on hard, and easy adds a 30 percent chance of a random legal move. Two rules follow from the numbers.
1. Trade men when ahead on material, because a 100-point swing per man scales with the count difference.
2. Push toward the center early, because the AI gains 10 to 20 points per piece there and loses that same score when you occupy those squares instead.
Protect your back row while advancing. The AI keeps its row 0 men for the 15 point bonus, and your row 7 men do the same job by blocking white coronation routes.
Crowning is worth three men
A king at 300 points equals three men at 100. When you calculate a sacrifice, price the crowning square accordingly. Giving up one man to guarantee a coronation on the next move is a 200 point swing in your favor before counting mobility.
Edge columns are the safe lanes for kings. The 8 point edge bonus exists because a piece on column 0 or column 7 can only be jumped from one side. Park a king on the edge when you need it to survive, and move it to the center when you hunt.
Checkers checklist
- [ ] Before every move, list all captures for both sides.
- [ ] If you have a capture, remember the move is forced, so plan the chain.
- [ ] Verify the longest-chain filter before relying on a sacrifice.
- [ ] Offer baits that force AI pieces into your diagonal chains.
- [ ] Keep two men on row 7 until the AI stops threatening coronation.
- [ ] Push men through the center squares for the 10 and 20 point zones.
- [ ] Trade when ahead on material.
- [ ] Price crowning at three men when evaluating a sacrifice.
The honest downside. The engine has no draw condition in code, even though its stats reserve a draws counter, so a blocked king endgame can grind on until one side runs out of legal moves. Hard at depth six also sees three full exchanges ahead, which means shallow two-move baits fail there, and the 0.4 to 0.8 second think delay per AI move makes those long endgames slow to resolve. Open [checkers](/en/checkers), play one game per difficulty with the sacrifice checklist, and track how many forced chains you predicted correctly. If your trap success rate on medium is below half, send the positions, because the longest-chain filter is the usual culprit.
