A Gomoku Daily Puzzle You Build Yourself, Ten Stones at a Time

Ten Stones, Black to Move, Two Different Answers

Black holds c5 and d5 on row five, e3 and e4 on the e file, and the center star at h8. White's five stones all sit off both lines, which is the point of the position: e5 joins them into three in a row with b5 and f5 empty, and three in a row with e2 and e6 empty. Two live threes through one point. White blocks one, Black extends the other into an open four, and the five lands two stones after e5.

Keep Black to move and switch the rule set to renju. The same e5 is a double three. Black loses the moment the stone touches the board. Not a bad move. A loss.

One diagram, two answers, and no book will show you both. The book prints one rule set on the cover. This board has a toggle, and the same ten stones become two different exercises depending on where it sits.

There Is No Puzzle Library Here

The board opens empty every time. No position of the day, no archive, no streak, nothing to subscribe to. If you want a puzzle a day, the position is the work, and two of the three minutes go into setting it up.

Against the computer you cannot set one up at all. Every stone you place draws an answer within twelve seconds, so a specific position you had in mind stays out of reach.

The vs Friend setting can do it. No AI, both colors belong to you, and the stones alternate black, white, black in exactly the order they did in the real game. Replay the move list, stop one stone before the mistake, and you are standing in the position.

One detail decides whose turn it is: parity. An odd number of stones means White moves next, an even number means Black. Need the other side? Add or drop the last stone and the count flips.

Twelve Stones Is the Ceiling

Six stones and there is nothing to read. Nine to eleven is the working range. Past thirteen, three or four lines are live at once, several points survive, and what you built is a game with the first half deleted.

A good puzzle passes one test: after your answer, is there a second point that also holds? If yes, the puzzle is broken. Not you.

Difficulty is the number of plausible points, not the number of stones. Two candidates is a warm-up. Five is a real puzzle, because four of them have to be eliminated for a stated reason. Nine candidates and you are guessing with extra steps.

Your Losses, and Your Wins

A lost game hands you a position and a blind spot. You remember your own idea in detail; you usually cannot reconstruct what your opponent was thinking. So set the position up and solve from the winning side. Lost as White, play Black. The refutation is easier to find than your own missed defense, because the thing in the way is your own reasoning.

Won games are the second source, and they need one edit. Take the position where your attack worked and add a single defensive stone in the spot that kills the obvious answer. In the ten-stone diagram, put a White stone on f5 and e5 stops winning: it now makes one live three and one closed three. The easy answer is gone, and a position you understood becomes one you have to solve.

That is the only edit a puzzle needs. One added stone deletes one answer.

One Question, Thirty Seconds, Written Down

Ask where you play and what the stone threatens. Not who wins. "Who wins" turns every puzzle into a fifteen-minute game, and you will stop opening the tab by Thursday.

Think for thirty seconds, then write the point on paper before you touch the board. This is not ceremony. Without the written answer you will find the right point, place it, and remember having known it all along, so the miss never enters your record.

Checking is the second half. Play your answer, then play both sides out yourself in vs Friend. Hand the reply to the AI and you learn what the engine values, which is a different lesson and a slower one.

The hint button is an answer key with the explanation removed. It marks one point, the highest-scored candidate for the side to move, and the marker fades after three seconds. Useful, and harmful if pressed early. When your point differs, the question is not "why is the machine right" but "is mine actually losing, or second-best and still holding". A second-best move that keeps the initiative is often the better thing to have found.

Tag the Miss, Not the Loss

Three tags cover almost every failed puzzle, and each one wants different work.

• Blind. You never saw the point. Volume fixes this: more positions, same difficulty, faster.
• Doubt. You saw it and took the louder point instead. More puzzles will not help, because the gap sits between seeing and playing. Write the reason for your choice before you place the stone, and the bad reason shows up in your own handwriting.
• Rule. You saw it, played it, and it was a foul. That is not a reading error. Either you moved to renju before the three bans became automatic, or the puzzle is testing a skill you are not working on.

Thirty puzzles is enough to see the shape. Whichever tag passes half of them sets the next month's work.

Six Days On, One Day Off

Three minutes a day: two to build, one to solve. Six positions a week, and the seventh day re-solves the three you got wrong. That is roughly 300 positions a year, about fifteen hours.

The same fifteen hours of games is about seventy games, and seventy games without a question attached teach less than you would like to believe. An hour at the AI is a separate exercise, worth doing, and not a substitute.

The failure mode is not difficulty. It is collecting. Forty positions saved, six solved. Keep no more than three unsolved and the queue stops being a museum.

What a Hundred Puzzles Will Not Fix

Spotting is one skill and it improves fast, which is why the first month feels so productive. Opening order is another and it barely moves. Endgame counting is a third: a puzzle ends the win on the fourth stone, and a real game makes you count to twelve with a clock running.

There is a harder limit. A puzzle freezes at the moment you went wrong, and live opponents beat you with the stone before it. If your puzzle record is clean and you still collapse at move twenty, you are solving positions downstream of the problem.

The eye you are training is real, though. The four shapes worth drilling are the material these positions are built from, and inside the wider practice loop a daily puzzle is one slot: one game a week reviewed properly, and nothing added on top of that.

Today's Puzzle

Take the last game you lost. Write the move list out, all of it, two characters per stone. Open the board, switch to vs Friend, and replay to the stone before the mistake. Count the stones to confirm whose turn it is, then give yourself thirty seconds from the winning side.

One position. Then close the tab. Tomorrow is a different position, and the point is not that you solved this one. By Friday you will have seen the shape four times and stopped needing the thirty seconds.