🪜 Trading Desk · Guide

Grid Strike — directional grids with an order-flow brain

A classic grid bot buys every dip forever and hopes. Grid Strike is the disciplined version: directional grids with per-level take-profits, entries gated by live order-flow evidence, a faithful triple-barrier exit, auto-follow re-centering — and the same 🛟 rescue governor that guards the market-maker desk, on each side independently.

How a strike grid differs from a classic grid

A classic grid is symmetric and unconditional: N buy levels below, N sell levels above, filled blindly. It prints in ranges and gets annihilated by trends — the strategy has no opinion, so the trend supplies one. A strike grid is directional and conditional. You (or the optimizer) choose a side; levels ladder into pullbacks of that side; each filled level carries its own take-profit; and a level only fires when the order book agrees that the pullback is being absorbed rather than continuing. The grid stops being a hope machine and becomes a structured way to scale into a thesis.

Levels, spacing & per-level TP

• Levels & spacing (bps) — how many rungs and how far apart. Spacing is your averaging distance: match it to the pair's typical pullback depth, not to a wish. Volatile movers want 2–4× the spacing of majors.
• Per-level TP (bps) — each rung exits on its own profit target, so a partial recovery pays the deepest rungs first. The grid harvests oscillation instead of waiting for full round-trips.
• Lot sizing — flat or scaled per rung; live orders are clamped to the exchange minimum notional exactly like the MM desk.

Order-flow entry gates

Before a rung is allowed to fill its entry, the desk consults the depth-analysis stack — the same detectors documented in the Order Flow guide. Optional gates include book-imbalance in your favor, absorption at the level, CVD agreement, and a spoof/stop-hunt veto. The philosophy is the platform doctrine: a missing signal skips the entry, it never guesses. A gated grid takes fewer, better rungs.

The triple-barrier exit

Every position lives inside three barriers, first touch wins: the profit barrier (per-level TP), the loss barrier (stop, in bps or structural), and the time barrier (max holding). The time barrier is the one retail bots always skip and always regret: a thesis that hasn't paid within its window is stale evidence, and stale evidence is exited, not nursed. The implementation is faithful — barriers are evaluated on the tape, not on candle closes, so a wick that touches the stop is a stop.

Auto-follow re-centering

When price walks away from a grid, a classic bot sits abandoned. With auto-follow enabled the desk re-centers the ladder around the new price once it has drifted a configured distance — flat rungs are moved, held rungs keep their barriers. You stay in business where the market actually trades, without the re-center ever averaging a losing rung deeper.

🛟 The rescue governor, per side

Adopted from the ⚖ dual-book MM desk (full write-up in the MM guide) and applied per side:

• A side's unrealized loss hits the rescue trigger (USDT) → that side stops opening new rungs. Fail-closed: the pause is checked before every entry gate.
• Realized profits from the healthy side fund reductions of the trapped side, dollar for dollar — reduce-only, position-side-aware on hedge accounts.
• Recovery to half the trigger releases the side into a cool-off before it may ladder again.

Running long and short together

On a hedge-mode account you can run a long grid and a short grid on the same symbol simultaneously — a two-sided oscillation harvester. Each side carries its own rescue trigger, its own pause and its own cool-off, so a trending tape traps at most one side while the other both earns and pays the rescue. In effect the pair of grids behaves like the MM desk's dual-book, with rungs instead of quotes.

FAQ

Related guides