The rules, and the one thing that decides the game
You have ten pegs in one point of a six-pointed star. You win by getting all ten into the point directly opposite. There are no captures. Nothing is ever removed, nothing is ever lost, and no position is ever unrecoverable — it is a pure race, which is why it works for six-year-olds and still has real depth.
On your turn you either step one hole to an adjacent empty hole, or you jump. A jump goes over exactly one adjacent peg — either colour, yours or theirs — and lands in the empty hole immediately beyond. From that hole you may jump again, and again, as far as the board allows. A good chain carries a peg most of the way across the star in one turn, and building a ladder of your own pegs for a later peg to run along is the whole strategy.
The rearmost peg decides the game. Every beginner runs their leading pegs forward and leaves one behind, and a single straggler costs more moves at the end than every clever jump saved at the start. We know this precisely, because our own bot could not finish a game until its evaluation was changed to punish its own rearmost peg — it went from finishing 3 games in 12 to finishing 12 in 12 on that one change. So the board rings your rearmost peg in red. It is the number the game is actually scored on.
Every jump chain is shown
Tap a peg and every hole it can reach this turn is marked — steps as small dots, jump landings as rings. Chains twenty landings long exist and no other site we could find will show you one; they all make you trace it in your head and then punish you for missing it. Hiding a legal move is not difficulty, it is a tax on people learning the game.
Ten bot levels, and none of them cost anything
The strongest site in this category gives away ten difficulty levels for free. Shipping three and calling it generous would be a downgrade wearing a wedge's clothes, so there are ten here too, and there is no paid tier on this site at all.
How we know the board and the rules are right
This game has no published solve — no perfect-play result, no standard position counts, nothing external to check an engine against. That is unusual for the games we build, and rather than grade ourselves we check three things that cannot be argued with.
The board is exact. A Sternhalma star is a central hexagon of 61 holes plus six triangles of ten, totalling 121. Our engine generates the six triangles by rotating one triangle five times rather than by listing coordinates by hand, and the suite asserts all three counts, that the triangles are disjoint, that adjacency is symmetric, and that all 37 strictly-interior holes have exactly six neighbours. A board with one missing corner hole would play almost normally and quietly lose about one game in fifty.
The jump generator is checked against a second one. Chained jumps are where this game is genuinely hard to implement: chains can revisit holes and cycle, so a naive recursion either loops forever or silently drops landings. We wrote the generator twice, in two different ways, and the suite runs both over 23,729 random positions and requires them to agree exactly. Two independent implementations agreeing everywhere is the strongest claim available when nobody has published a number to check against.
Invariants. Peg count is conserved across every move, no jump ever lands on an occupied hole, and cyclic chains provably terminate — checked over thousands of plies of self-play on every build.
It is not Chinese
It was invented in Germany in 1892 as Stern-Halma, a star-shaped variant of the older English game Halma. The name "Chinese Checkers" was invented by American manufacturers in the 1920s as marketing. The board has no connection to China, and we would rather say so than repeat it.