Chess vs the Computer: How to Build a Winning Streak

Updated October 2026

The Chess game on this site is a streak challenge: you always take white, the computer always takes black, and your score is the number of games you win in a row. That changes how you should play. A loss or a draw ends the run, and the computer looks further ahead once your streak grows. This guide explains how the opponent actually thinks, where its blind spots are, and how to build a long run of wins one careful game at a time.

How the streak is counted

Every finished game ends in one of three ways. If you checkmate the computer, the streak goes up by one and your best is updated on this device straight away. If the game is drawn, by stalemate, insufficient material, the fifty-move rule or threefold repetition, the run is over, exactly as if you had been checkmated, and the streak goes back to zero. After a win the result card offers a Next game button, which keeps your streak and sets up a fresh board.

Two details are worth knowing before you start a long run. First, pressing New in the top bar or on a result card while a streak is running ends the run: a Run ended card appears with your streak and a Save score button. If you want to keep your run, always continue with Next game. Second, the button that sends a score to the Hall of Fame appears whenever a run above zero ends, whether by a loss, a draw or pressing New, and it submits the streak that just ended, not your all-time best. Your best streak is still remembered locally, so nothing is lost if you never press it.

How the computer chooses its moves

The opponent is a classic search engine. It looks a fixed number of moves ahead, scores the resulting positions, and plays the move with the best guaranteed outcome. The scoring is simple: material (pawn 100, knight 320, bishop 330, rook 500, queen 900) plus a small bonus or penalty for where each piece stands, from a table that likes centralised knights, pawns that have advanced, and a king tucked in a corner behind its pawns. At the end of every line it keeps following captures until the position goes quiet, so it does not stop counting in the middle of an exchange.

How far it looks depends on your current streak. At a streak of zero or one it searches three plies deep, where a ply is one move by one side. From two to four it searches four plies. From five onward it searches five plies. There is also a hard thinking budget of 1.5 seconds per move, after a short pause of about a third of a second so the move does not appear instantly. Search depth is increased one step at a time, with the previous best move examined first. When the budget runs out in the middle of a depth, the computer keeps any better move it has already confirmed at that depth instead of throwing the unfinished depth away.

On a desktop computer the five-ply level finishes well inside the budget, even in busy middlegames. A phone is slower, and in our test with the budget cut to match a phone, about three moves in ten were cut short at five plies; thanks to the partial result those moves are still a little stronger than the four-ply level. In matches between the levels, each one beat the level below it both at the full budget and at the phone-speed budget, so the opponent really does get harder as your streak grows. The computer also shuffles its candidate moves before each search, so when two moves score the same it does not always pick the same one, and repeating an old game move by move will not reliably reproduce it.

Exploiting a short horizon

Anything the computer can see inside its search window it will handle well. It never leaves a piece hanging to a one-move capture, and it will happily take one that you leave unprotected. What it cannot see is a plan that only pays off beyond the horizon. Three practical consequences follow.

Opening and middlegame plan

You move first, which is a small but real advantage. A steady system works better than memorised lines: occupy the centre with one or two pawns, develop both knights before the bishops, castle, and connect your rooks. The engine's table discourages it from leaving pieces on their home squares too, so expect it to develop sensibly, but it has no concept of tempo beyond what it can calculate.

Before every move, ask two questions. What does the computer's last move attack, and after my intended move, is any of my pieces defended fewer times than it is attacked? Almost all losses against an engine of this type come from material left loose rather than from deep combinations. The captured-piece rows above and below the board help here: each shows the pieces that side has lost, and a +N figure appears next to whoever is ahead, using the familiar 1, 3, 3, 5 and 9 point values.

When you are ahead, trade pieces, not pawns. Every exchange reduces the number of tactics the computer can find to swing the game back, and a simple ending where you are a piece up is a guaranteed win if you know the technique. When you are behind, the opposite applies: keep the position complicated, because the time budget means the computer searches less deeply when many pieces are on the board.

Converting a winning position

A won position is only worth a point once the king is actually mated, and three things can stop that from happening.

  1. Stalemate. If the computer has no legal move and is not in check, the game is drawn. When you have a queen against a lone king, leave the king a square to breathe on every move until your own king has arrived to help deliver mate.
  2. Repetition. The game counts every position, including whose turn it is, castling rights and any en passant square. The same position appearing three times is an automatic draw. The computer recognises repeated positions and the fifty-move rule in its search and scores them as a draw, so when it is worse it will happily steer into a repetition. Since a draw ends your streak, it is you who must vary the moves when you are winning.
  3. The fifty-move rule. After one hundred consecutive half-moves with no capture and no pawn move, the game is declared drawn automatically. Long, aimless king hunts are the usual culprit. Have a mating method ready: with king and rook, use the rook to cut the enemy king off on a rank or file and bring your own king to face it.

A draw ends your streak just as a loss does, so there is no safe draw to take when a game turns unclear. Keep playing for the win, and be most careful exactly when the computer starts repeating moves: that usually means it is worse and looking for a way out. Only checkmate moves the counter up.

Controls and special moves

Tap one of your pieces to select it. Empty squares it can legally reach show a small dot, and pieces it can capture show a ring. Tap a highlighted square to move, tap the same piece again to put it down, or tap another of your pieces to switch. The squares of the last move are tinted, and your king's square glows red when it is in check. Pinned pieces simply do not offer illegal destinations, so a missing highlight always means the move really is not allowed.

To castle, move the king two squares towards the rook; the rook jumps over automatically. En passant appears as an ordinary diagonal move onto the empty square behind the enemy pawn. When a pawn reaches the last rank, a small picker offers queen, rook, bishop or knight. Pick the queen unless you have checked that it would leave the enemy king with no legal move and no check, in which case a rook or knight can save the win.

Every result card shows your current streak and best. Aim first for three or four wins in a row against the lighter levels, then treat each game from five onward as a separate endgame exercise: get a safe material lead, simplify, and mate cleanly.

▶ Play Chess Take white against the computer and see how long your winning streak can grow.