Draughts against a computer is won by shots: you give up a man, the forced capture pulls an enemy piece onto the square where you take two or three. Easy looks only two plies ahead, never sees a shot coming and plays the same moves every game, so 1.c3-d4 b6-a5 2.g3-f4 a7-b6 3.d4-e5! wins a man on move four every time. In the endgame remember the big road: three kings beat a lone king that is off the a1–h8 diagonal in at most 14 moves, and a lone king that holds it draws.
Capturing is compulsory, so a man you put en prise is an order, not an offer: the other side must take it, wherever the capture leads. A shot uses that. You give one man (sometimes two), the forced capture drags an enemy piece to a square it did not choose, and your reply takes two or three in one move — often finishing on the back rank as a king.
To see a shot coming, the side that walks into it has to look four plies ahead: its own move, your sacrifice, its forced capture, your multiple capture. That is exactly the difference between the levels here:
We let a simple player — one that, like Easy, checks only the reply to its own move — play 400 games against Easy and looked for shots. In 329 of them (82%) Easy at some point allowed a shot worth two men or more (a king counts as three), usually by its eighth move. In the same positions Hard's move left such a shot only 40 times out of 329.
Easy has no randomness and no clock. It searches two plies, scores the result by material alone, and among equal moves takes the first one it generated. So the same moves from you always bring the same answers. Whatever your first move is, it answers b6-a5; and it does not see a shot until the capture is already on the board.
This line wins a man against it on move four, every game:
The order of your first two moves does not matter: 1.g3-f4 b6-a5 2.c3-d4 a7-b6 3.d4-e5 is the same trap. We searched every line of up to three moves for you and none wins material against Easy; this four-move line, in its two move orders, is the shortest.
A second line gets a king on move six: 1.c3-d4 b6-a5 2.e3-f4 a7-b6 3.b2-c3 b8-a7 4.d2-e3 d6-c5 5.c3-b4! a5:c3:e5 6.f4:d6:b8. You give two men, take two, and your man is crowned on b8; the position before the shot is the first diagram in the next section.
Both lines work only against Easy. Hard answers 1.c3-d4 with b6-c5, and the trap never arises.
The first position is the six-move line above; the second comes from a simulated game against Easy. Both are shown after Easy's last move (ringed). Easy saw neither threat; Hard sees both.
What the two have in common is holes. In both, d6 and b8 are empty, so every jump lands on a free square and your capture runs all the way to the back rank. Before you play a quiet move, look at every enemy man next to one of yours and ask what lies behind it.
You move first and have seven possible moves. Here is how the levels answer each, and how a simple player (one move of safety, keeping the back rank) did against Easy after each, about 700 games per move:
| Your first move | Easy answers | Hard answers | Simple play vs Easy: win / draw / loss |
|---|---|---|---|
| e3-d4 | b6-a5 | b6-c5 | 64% / 26% / 10% |
| c3-d4 | b6-a5 | b6-c5 | 61% / 28% / 11% |
| g3-f4 | b6-a5 | d6-e5 | 54% / 24% / 22% |
| e3-f4 | b6-a5 | d6-e5 | 43% / 23% / 34% |
| g3-h4 | b6-a5 | f6-g5 | 38% / 29% / 33% |
| a3-b4 | b6-a5 | d6-c5 | 35% / 24% / 41% |
| c3-b4 | b6-a5 | d6-c5 | 27% / 24% / 49% |
The two moves to d4 come out on top and the moves towards the side at the bottom. That matches the old advice: d4 and e5 are the key squares of the centre, a man there supports its neighbours, and a man on the edge controls only one diagonal.
Expert's answer depends on how deep it gets in its 0.7 seconds, so it can differ from one device to another: searching eight or ten plies it meets 1.c3-d4 with f6-e5, but on the clock in our test it played d6-c5.
A plan matters even against a two-ply engine. We gave a simple player — it checks only the reply to its own move — two habits and played 5,000 games against Easy for each combination:
| Your play | Win | Draw | Loss | Games where Easy crowned a king |
|---|---|---|---|---|
| Random legal moves | 0.2% | 0.1% | 99.7% | 90% |
| Check the reply, nothing else | 36% | 26% | 38% | 65% |
| … and keep the back rank | 46% | 26% | 29% | 54% |
| … and prefer the centre | 48% | 30% | 22% | 51% |
| … and both | 79% | 15% | 6% | 20% |
Each habit alone is worth about ten points; together they more than double the wins. The back rank is why Easy crowns less often: a man on your first row stands on a square an enemy man needs in order to crown, and there is rarely a reason to move it before the middle game. The centre is the reason your men stay connected: from d4 and e5 a man covers two diagonals and is hard to get behind.
Taking material first: a player who just grabs the biggest capture and otherwise moves at random loses 99.7% of games against Easy, the same as random play. Captures are compulsory anyway; what loses is the quiet move that lets the other side shoot.
The long diagonal a1–h8 is called the big road. It is the only diagonal that crosses the whole board, and in the endgame it decides most results. We solved every position with one, two or three kings against a lone king exactly, with the game's own rules:
The order of play follows from that. With three kings, put one on the big road before anything else; then the lone king cannot get there, and the rest is a technique known as Petrov's triangle. With a lone king, run for the road, and once there move along it — never off it.
The 15-move limit matters here. The game declares a draw after 15 moves by each side with kings only and no capture, counted from the last capture or man move. The slowest win of three kings against one takes 14 moves, one inside the limit, so there is almost no move to spare.
Against this computer the endgame is kinder to you than the theory. We gave Expert's lone king 100 positions where it could still reach the big road and draw; against three kings playing correctly it saved only 42 of them. The other way round, with three kings of its own, Expert won 76 of 100 theoretically lost positions against a defender who just avoided capture and ran for the road, but only 49 of 200 against the most stubborn defence — the rest ran into the 15-move limit. So with three kings, press; with a lone king, never resign: count the moves.
Here is how three kinds of play did against each level, a few random opening moves apart (simple play against Easy: 5,000 games; everything else: 32 to 192 games):
| Your play | vs Easy | vs Hard | vs Expert |
|---|---|---|---|
| Simple: check the reply, keep the back rank, prefer the centre | 79 / 15 / 6 | 4 / 2 / 94 | 0 / 0 / 100 |
| Calculating six plies ahead, like Hard | 97 / 1 / 2 | 34 / 35 / 30 | 18 / 27 / 55 |
| Six plies plus finishing every capture, like Expert | 97 / 2 / 1 | 48 / 29 / 23 | 34 / 16 / 50 |
Read the table as win / draw / loss for you, in percent. Expert was simulated at six plies, the least it searches; on a fast device it often looks seven to twelve plies ahead, so the real Expert is somewhat stronger than its column.
Play for shots and do not walk into them: before every quiet move, check which enemy man has an empty square behind it and what the forced capture would pull where. Keep the back rank, take the centre squares d4 and e5, and keep your men together. Against Easy those habits alone turn about 36% wins into 79% in simulation.
A move to the centre: e3-d4 or c3-d4. Against Easy, a simple player won 61–64% of games after them and only 27–38% after the moves towards the side, a3-b4, c3-b4 and g3-h4.
Yes, if the lone king is not on the big road, the long a1–h8 diagonal. Then it is a forced win in every one of the 49,676 such positions, in at most 14 moves, inside the game's 15-move rule. If the lone king holds the big road, it is a draw.
No, as a rule. Two kings win only when the lone king is already caught; with nothing to capture and the lone king to move, about one position in a hundred is lost for it.
The long diagonal from a1 to h8, the only one that crosses the whole board. A king on it controls the most squares, and a lone king on it cannot be beaten by three kings.
Easy does: it has no randomness and no clock, so the same moves from you bring the same answers, and a winning line wins every time. Hard and Expert think against a clock, so on a slower device they may search less deeply and occasionally answer differently.
Yes. Easy loses a man to the same four-move trap every game, and simple habits — keep the back rank, take the centre, check the reply — beat it four times in five. Hard needs real calculation: a player looking six plies ahead wins about a third of games against it. Expert is the toughest, but its endgame is weak: with three kings against its lone king you won 58 of 100 positions it could have drawn.