Gomoku

Yes. Gomoku is the Japanese name, short for gomokunarabe, and five in a row is what the game is. It has local names in many places: wuziqi in China, omok in Korea, gobang in older English books, and luffarschack in Sweden.

On the points where two lines cross, as in go, which is where the board and the stones come from. On squared paper the same game is played with marks inside the squares. Nothing else is different.

That depends on the game you chose. With "Five or more" it does. With "Exactly five" it does not: the row counts for nothing, the board tells you so, and the game goes on.

Three pieces in a row with the points at both ends empty, and room beyond them. One more piece at either end makes four in a row that is still open at both ends, and that cannot be stopped.

So an open three has to be answered on the very next move. It is the row a newcomer most often leaves alone, and then loses to.

Four pieces in a row with the points at both ends empty. There are two ways to make five, and a single piece can only take one of them. An open four wins.

A four with one end closed is a different matter. It has one way to become five, and a piece on that point stops it.

The side that moves first. In 1993 Victor Allis showed that in the free game, on a board of fifteen lines each way and with no restrictions on the opening, the first player can force a win. His program searched for chains of threats, each of which has to be answered, and found a win against every defense.

This is why tournament play adds forbidden moves or opening rules: without them the first move is worth too much. This board has none, so going first is a real advantage. You can give it to the computer under Settings.

In the middle. A piece there can grow into a row in every direction, with seven points of room each way. The computer starts in the middle, or right next to it when it varies its opening.

Only if the whole board fills up. That is 225 pieces without a winning row, and in practice somebody wins long before.

Because 225 points are too many for a fingertip on a phone. There the first tap brings the board closer around the place you touched, and the second tap puts the piece down. A single tap never places a piece on a phone, so a stray touch costs nothing.

A tap below the board shows the whole board again. On a screen where the points are far enough apart, or with a mouse, a move is a single click.

Yes. "Five or more" and "Exactly five" each have a saved game and a record of their own. Switching under Game leaves the other game where it was, and it is still there when you switch back.

No. Go uses the same kind of board and stones, but it is a game about surrounding territory, and stones are captured. Renju is gomoku with forbidden moves for the side that starts. This page has neither.

Two kinds of row cannot be left alone. A four, because the next piece makes five. And an open three, because the next piece makes an open four, which is already won.

Before every move, look for both, yours and the computer's. If the computer has one, your move is decided: stop it, unless you can make five yourself.

Nobody leaves a four unanswered on purpose. Games are won by a piece that makes two threats with one move, so that answering one leaves the other.

• A four and an open three: the four has to be blocked, and then the three becomes an open four.
• Two open threes: only one can be stopped, unless the other side has a four of its own to play first.
• Two fours: two points make five, and one piece cannot take both.

The point to look for is the one where two of your rows cross.

Two pieces, an empty point and one more piece on the same line are a three: fill the gap and it is four in a row. Three pieces, a gap and one more are a four: fill the gap and it is five.

Rows like these are easy to overlook, for people and for the lower levels here. Beginner and Easy see only unbroken rows.

A threat that has to be answered also decides where the other side plays. A player who makes one such threat after another places every piece where it suits, while the other side only follows.

A four is the strongest of them: there is exactly one answer. Strong players win by a chain of fours and open threes that ends in two threats at once. The same holds in defense. When you have to block, block with a piece that also builds a row of your own.

An open three can be stopped at either end. Pick the end where your piece does more: where it lies in line with other pieces of yours, or where it sits on a point the computer needs for a second row.

A row needs five points on a line. Near the edge of the board there are fewer of them, and a row that runs into the edge is closed at that end. Pieces in the middle have room in all directions.

When only exactly five wins, a row that would grow to six is no threat at that end. Four in a row with one more piece of the same color a step beyond the empty point cannot be finished there: the piece that fills the gap makes six.

It works both ways. Check that your own four really has a point where it becomes five and not six.

The Hint button asks the engine at full strength what it would play. When its point is not the one you had in mind, look for the row it stops or the two rows it joins before you play either.

For every empty point the engine keeps track of what a piece there would make, for each side and along each of the four lines through the point: a five, an open four, a four, an open three, and so on down to a lone piece. A three with a gap in it is found the same way as an unbroken one.

Its moves come from those threats. If it can make five, it does. If you can, it blocks. If you have an open three, it tries only the points that stop it, or a four of its own. Otherwise it tries the empty points within two steps of a piece, the most threatening first, then your answers and its answers to those, as many moves deep as the level allows.

The two strongest levels also run a search of their own for forced wins: rows of fours, or of fours and open threes, where every move has to be answered. A win found that way is certain, because the answers it considers are the only ones there are.

All of this happens on your own device, in your browser.

Looks at its own move only, with a beginner's eye: it sees unbroken rows and nothing else, and no double threats. It blocks an unbroken four and an unbroken open three. In every other move it only glances at the board, and then it misses the three.

Has the same eye, but looks at your answer as well and builds on its own rows with a purpose. It is careless in about one move out of five, and it never sees a four with a gap in it.

Sees what is really there: rows with gaps, and double threats once they are on the board. It still looks only two moves ahead, so it does not see them coming. It is careless in about one move out of twelve.

Looks four moves ahead and is never careless. It also searches for short forced wins, four threats long.

Gets one second for every move in place of a fixed depth, and searches for forced wins ten threats long. On the computer where it was measured it used 0.4 seconds a move on average and looked seven moves ahead. It varies its first few pieces a little, so that games do not all begin the same way.

Each level has played the level below it in the free game. The games began from random openings of one to four pieces near the middle. Every opening was played twice, so that both levels got to start from it, and in every other opening white placed first. No game was drawn.

• Easy against Beginner: 902 wins and 98 losses, over 1,000 games.
• Medium against Easy: 867 wins and 133 losses, over 1,000 games.
• Hard against Medium: 365 wins and 35 losses, over 400 games.
• Expert against Hard: 34 wins and 6 losses, over 40 games.

With the exactly-five rule, Medium beat Easy 348 times in 400 games. The other steps were measured in the free game only.

The last figure rests on few games, because Expert uses its clock. And all of these are engines playing engines: a simple model of an ordinary beginner, one that makes its longest row longer and blocks only fours, beat Beginner in 658 games out of 1,000.

The proof that the first player wins was found by a program built for that one task. Expert has one second a move and no book of openings. It does not know the winning line, and it does not play without error: it lost six of its forty games against Hard.

So the proven result does not decide games on this board. Going first is an advantage, against Expert as against anyone, and Expert can be beaten from either side.

The Hint button is the engine at full strength: one second, and the search for forced wins ten threats long. That is Expert without the variety in its opening.

Using it does not make the level you are playing any stronger: every search starts from nothing. Hints are counted, and the count is shown with your record at the bottom of the page.