Game The short answer
Across 400 endgame positions (10 empty squares, 2617 moves solved exactly), taking a corner was the best move 50.8% of the time. In roughly half of the 246 positions where a corner was available, something else was better.
How much did it cost to take it anyway? 0 discs at the median (half the time the corner was simply best) and 5.11 on average — but 16 or more in the worst tenth, up to 38. In 22 positions taking the corner threw away a win — 8.9% of the 246 where a corner was playable, or 13.3% of the 165 that best play actually wins. A small average does not mean the bad cases are rare enough to ignore.
What we counted
Our earlier articles each took a single position and compared the candidate moves. The trouble with that is the obvious objection: maybe that position was just unusual.
So this time we solved positions in bulk. Othello endgames can be read to the very last move, which means we can state how often a piece of common advice was actually right as a measured number rather than an impression.
Method
We played games out with random legal moves and collected positions with 10 empty squares and Black to move. Duplicates were discarded, and we kept only positions with at least three legal moves, since a position with no real choice tells us nothing.
In each position we solved every legal move exactly, to the end of the game, and recorded the final disc margin. A move counts as "best" if it matched the highest final margin in that position (ties included).
The counting, precisely: one count per position. "Cost" is the best move's final margin minus this move's final margin, so it is 0 whenever the move was (tied) best. Where two corners were playable (84 of the 246 positions) we count the better corner, which gives the rule the most favourable reading. The denominator is the number of positions where the move existed: 400 for the flip counts, 246 for corners, 153 for X-squares.
Results
"Positions" is how many of the sampled positions offered that kind of move; "was best" is how often it matched the top final margin. Cost is the gap in final discs to the best move: median is the middle value, max the worst that actually occurred in these 400 positions.
“Lost a win (of all)” is counted against every position where the move could be played: positions that best play wins for Black but this move does not, divided by the positions where the move existed. A draw does not count as a win — turning a win into a draw counts as losing it. Where more than one corner was playable, the win/loss call uses the better corner, the same one the cost is measured on.
The other reading — counted only against the positions that were winnable — is available too. For corners: of the 246 positions where a corner could be played, 165 were wins for Black with best play, and in 22 of those (13.3%) taking the corner gave the win away. The table shows the first reading.
| Rule of thumb | Positions | Was best | Avg cost | Median cost | Max cost | Lost a win (of all) |
|---|---|---|---|---|---|---|
| Taking a corner | 246 | 50.8% | 5.11 | 0 | 38 | 8.9% |
| Flipping the most discs | 400 | 19% | 13.99 | 12 | 60 | 23.8% |
| Flipping the fewest discs | 400 | 22.5% | 11.13 | 8 | 48 | 20.5% |
| Playing an X-square | 153 | 9.2% | 16.77 | 16 | 60 | 31.4% |
A percentage is hard to picture, so here is one position out of the 400 — the one where taking the corner cost the most. Black can play the corner h1, but it finishes −22, a loss. The best move, a2, finishes +16: a difference of 38 discs.
BOARDBBBBB..WWB.B.BWWWWBWBWW.WWWWWWWBWBWWBWWBWBWWWBW..WWWWWW.WWWWW.B. black
| Move | Flips | Final margin | Note |
|---|---|---|---|
| a2 | 5 | +16 | — |
| h3 | 3 | −12 | — |
| f1 | 1 | −22 | — |
| h1 | 1 | −22 | Corner |
| e7 | 4 | −26 | — |
| h6 | 2 | −32 | — |
| g8 | 6 | −38 | — |
What the numbers say
Corners really are in a class of their own. They were best 50.8% of the time, more than double any other rule, with a median cost of 0 and an average of 5.11 — both clearly lower than anything else here.
That is not the same as "you can't go wrong". The worst tenth of those positions cost 16 discs or more, the worst of all 38, and of the 165 positions best play wins, 22 (13.3%) stopped being wins once the corner was taken. The board above is that 38-disc case, drawn from this very sample.
In about half of the 246 positions where a corner was on offer, another move did better. As Two corners, two very different values showed, the endgame is decided by reading it out, not by counting corners.
The number of discs flipped tells you almost nothing. The move flipping the most discs was best 19% of the time, costing 12 discs at the median and 13.99 on average, and of the positions best play wins it gave the win away 41.1% of the time; the move flipping the fewest was best 22.5% of the time, costing 8 at the median and 11.13 on average. Neither "take more" nor "take fewer" works as an endgame guide. The take-fewer instinct that serves you well earlier in the game does not carry over.
X-squares were the worst by every measure. Best only 9.2% of the time, at a median cost of 16 discs and an average of 16.77 — and of the 95 positions best play wins, an X-square gave the win away in 48 of them (50.5%) — more often than any other rule here. The traditional warning holds up.
Limits of these numbers
To be straight about it: these positions come from games played with random moves, so their distribution differs from what strong players reach. In real games, the shapes that force you into an X-square are often avoided long beforehand.
The sample is also specifically the endgame, at 10 empty squares. Nothing here applies to the opening or middlegame. It speaks only to how to choose moves once the endgame has started.
The numbers themselves can be checked. Every figure in this article comes from othello-stats.json, written by tools/generate-othello-stats.mjs. Positions are drawn with a fixed random seed, so running the script again produces the same 400 positions and the same figures. Because the solve is exact, nothing depends on engine strength or search depth.
What to remember at the board
- In the endgame, always consider a corner first — it has the best hit rate
- But it is best only about half the time, and the worst case here cost 38 discs — read it out before taking it
- How many discs a move flips is not a useful endgame guide
- X-squares are rarely best and cost the most when they are not
Check it in the Othello Dojo
Replay the endgame and pick again, with the evaluation in view. Free, no account.
⚫ Play Othello