brainGo

One Stone Flips the Answer: brainGO's Original Go Training

⏱ Read ~5 min · Updated July 2026 · Part of: Learn Go

Same position, add or remove one stone, the answer flips.

That is what brainGO's "one-stone variations" feature lets you see. Most Go practice asks "what is the best move?"; one-stone variations ask a different question: in this position, which stones actually decide who wins?

It does not answer by feel. It answers with a contrast you can see for yourself — in the same position, removing this stone flips the result entirely; removing that one changes nothing.

This is a brainGO original

Take one position, add a stone at every empty point — or remove every stone already there — work out the result for each, and lay them out side by side. Nobody else offers this. It is only possible on a small board where every position can be solved exactly — 4×4 is small enough. Every comparison is exactly computed, not guessed.

One stone, opposite answer

Take this position (Black to move, to capture the White stones in the top-left). The brainGO algorithmic engine's verdict: Black wins in 4 moves.

Now add one White stone in the bottom-right corner at (2,3) — two ranks away from the target White group, not even connected to it. Yet the brainGO algorithmic engine's verdict flips: draw — Black runs out of moves.

Same shape, add one completely unrelated stone, and Black goes from a win to a draw. Why does this far-away stone save the top-left White? See the cap-099 one-stone variation page for the full comparison — reason it out yourself.

Ko positions are even more fragile

When a position involves a ko, one stone's impact gets extreme. Take the ko position cap-125: Black wins in 7 moves. But change a few stones — remove a Black stone, remove a White stone, or add a White stone in the center — and the position collapses into a draw. See the cap-125 variation breakdown.

How this helps you learn Go

Its value is not in telling you "the correct move" — that is the job of tsumego (life-and-death problems). It does something else: it turns "this stone matters" from a feeling into a comparison you can check.

You may have assumed a certain stone was critical, only to open the comparison and find removing it does nothing — while a stone you overlooked turns out to be the real "key stone" carrying the result. That correction, "I had it wrong," cuts deeper than solving another puzzle with a known answer, because it shakes your actual judgment of the position.

As for why it is that stone — whether it changed a liberty, sealed an eye, or cut a connection — the feature gives you the contrast, and the mechanism is yours to reason out. The fact that you have to reason it out is exactly what trains judgment, rather than having you memorize answers.

Honestly: what carries over, and what doesn't

The shapes on a 4×4 board almost never appear on a 19×19 board. So what this training carries over is not shapes, but a thinking habit — before you play, you ask, "which stone is the key stone in this position?" That habit of asking follows you to the big board.

It is also not a replacement for tsumego. Tsumego trains reading deeply in one position (can this group be killed?); one-stone variation trains reading broadly across near-positions (which stone decides the outcome?). One is depth, the other is breadth — both are legs of Go calculation, and you need both.

For brain training: counterfactual reasoning

People who don't play Go benefit too. One-stone variation is, at heart, counterfactual reasoning — "if there were one more stone here, what would happen?" The ability to change one condition in your mind and work out the consequence is central to working memory and decision-making. Exercises that require holding multiple board states in mind at once are among the few training types cognitive science considers to have transfer effects.

We say it "trains thinking," not that it "prevents dementia" — the latter is not scientifically settled.

See the full map

brainGO's 198 one-stone variation sets each come with a full ±1-stone comparison grid — the base position next to every variation that flips the result or shifts the difficulty, with exact win/draw values and move counts.

→ See the full one-stone variation map

The takeaway

A game is not just about "the best move" — every stone carries its own weight, and one-stone variations let you see that weight directly. Play the capture puzzles to feel it, then explore the full variation map to grasp the weight of every move.

FAQ

Why can one-stone variations only be done on 4×4, not 9×9 or 19×19?

Because the computation explodes exponentially. 4×4 has only 16 points, roughly 43 million reachable states — a modern computer exhausts it in seconds, so "what happens if one stone is added or removed" can be computed exactly for every position. 9×9 (~10^38 states) and 19×19 (far more) cannot be exhausted even with supercomputers running to the end of the universe. This is why brainGO "distills" Go to 4×4 — not because 4×4 is "simpler" than a large board, but because only 4×4 can be fully solved, giving you provable answers like "this stone truly decides the outcome" instead of an AI guess. On large boards, asking "which stone matters" can only rely on intuition or neural-net approximation; on 4×4, you see the truth.

I don't play Go — what's the use of one-stone variations for me?

Even if you don't play Go at all, this feature trains a general capability: counterfactual reasoning — "if condition A changes a little, what happens to the result?" This thinking style is central to medical diagnosis ("what if we switch the drug"), investment decisions ("what if that order disappears"), and engineering debugging ("what if I remove this line of code"). One-stone variation turns this thinking into a visual exercise: you literally see "remove one stone, the result flips" or "remove one stone, nothing changes," training you to distinguish "the variable that truly matters" from "one that looks important but is irrelevant." Most people don't lack logic — they lack the working memory to hold multiple scenarios in mind and compare them simultaneously, and that's exactly the few training types cognitive science considers to have transfer effects.

Related guides