Some OverGrid decisions reward calculation. Others do not give you enough time or certainty for a full reply tree. The useful skill is knowing which kind of proof the board can support.

Use two lanes:

  • Reply-backed: when two constructions are close and the opponent has a legible best answer, compare what survives that answer.
  • Board proof: when the replies are too broad, require one visible result—banked height, a removed hinge, a contained swing, or safe closure.

Both lanes share one rule: the clever move must remain a competent move.

Figure 2One shared search, two kinds of proof.Public board only · same 0.005 regret rail
••
Shared candidate census
↓
Lane A

Reply-backed

When both moves have a complete public reply receipt, compare projected margin and terminal safety.

Lane B

Threat proof

When receipts are sparse, accept only a visible denial, contained swing, or safe closure.

No hidden hands, labels, future deals, RNG cursor, or extra root search enter either lane.

Board sequenceA reply-backed proof needs a before, a defense, and a surviving gain.

Follow the marked cells from position to reply to result.

Your territoryRival territoryObjectiveYour lineTheir reply
  1. State 1Create the threat
    Create the threatBlue prepares a capture chain that will harden the marked pocket next turn.
    If ignored
    +30 next turn

    Blue prepares a capture chain that will harden the marked pocket next turn.

  2. State 2Reply is forced
    Reply is forcedRed abandons its preferred height route to occupy the only closing cell.
    Red’s preferred route
    delayed

    Red abandons its preferred height route to occupy the only closing cell.

  3. State 3Spend the tempo
    Spend the tempoBlue uses the stolen turn to advance a separate objective while red remains tied to defense.
    Initiative
    blue keeps it

    Blue uses the stolen turn to advance a separate objective while red remains tied to defense.

Read this asThe proof is not the arrow in state one. It is the opponent's diverted turn in state two and the separate objective still available in state three.

Lane one: calculate a reply when the choice is genuinely close

Start with your two best complete constructions. Do not manufacture a dramatic alternative just to make the exercise interesting.

For each move, choose the opponent’s best public answer: the reply that most improves their own score, route, height, or endgame—not a spite move that only damages you. Then compare the positions that remain.

The winning construction is often the one whose value is boringly intact after that exchange. A route that scores 30 and gives back 25 may be worse than one that scores 20 and gives back 5. Captures make this especially important because the swing touches both totals.

Lane two: demand a small board proof

Sometimes there is no honest single reply to calculate. Several hands could attack the frontier, or several players will act before the board returns. In that case, prove less.

A useful board proof names one thing the move accomplishes immediately:

  • it banks exact printed height;
  • it removes a shared route hinge;
  • it contains a visible capture swing;
  • it reduces neutral territory without opening a larger reversal; or
  • it preserves two independent ways to score.

“Creates pressure” is not yet a proof. Pressure becomes legible when you can name who must answer, what they must protect, and what remains if they choose the other problem.

Keep a competence rail

Characterful play is fun; self-sabotage is not. Give every speculative construction a maximum acceptable sacrifice against your plain best move.

You do not need a decimal evaluator. Use a simple rail:

  1. What guaranteed score or territory am I giving up?
  2. What concrete obligation do I receive in exchange?
  3. Can one ordinary reply erase that obligation?
  4. If the idea fails, do I retain mobility or a safe exit?

If the sacrifice is clear and the compensation is adjectives, stay with the competent move.

Change the proof when the score changes

When ahead, “safe closure” is a stronger proof than “more possibilities.” Reduce neutral territory, protect the final hinge, and make the remaining swing smaller.

When behind, closure helps the leader. Your proof should preserve volatility: two obligations, an accessible height bank, or a route that forces the leader away from simplification.

The same construction can therefore be correct at one score and wrong at another. Strategy is not a list of favorite shapes; it is a claim about what the current board needs.

The twenty-second version is simple: calculate when the reply is legible; otherwise prove one public consequence. If you can do neither, you have an idea—not yet a move.

Read the paired developer note: why Ender’s first policy never activated →