Lead-Lag Taker (IOC)

Two books price the same contract, but they don't move at the same instant. When the leader jumps, the laggard's resting quotes are briefly wrong — take them with IOC orders that can never rest, and you've bought at yesterday's price after the news.

The idea & why the edge exists

The BTC spot/index feed moves first; native CLOB quotes update when makers re-quote. After a sharp BTC move, the feed has repriced while stale resting quotes can still sit in the book for a second or two. Whoever takes those quotes first collects the difference — this is the oldest HFT trade there is, runnable here at human speeds because the venue's natural latency is seconds, not microseconds.

IOC (immediate-or-cancel) is the entire risk model: the order fills against whatever stale size exists and cancels the rest. You never leave a resting order that someone faster can pick off in turn.

Market setup

1. Stream both books of a pair (native + mirror) over WebSocket and timestamp every mid change locally.
2. Determine the leader empirically: log which book's mid moves first across 50+ moves. (Typically the mirror leads during external-venue-driven moves, the native leads during RAETH-flow moves — your signal can use whichever moved FIRST this time.)
3. The BTC spot feed itself (feed.btc channel) is a third, even faster leader: a $50 spot jump predicts BOTH binary books re-pricing.

The core loop

1. On every leader-mid change, compute move = leader_mid − laggard_mid.
2. If |move| ≥ TRIGGER and the laggard's touch hasn't re-priced within STALE_MS: fire one IOC at the laggard's stale touch in the direction of the move.
3. Exit when the laggard re-prices (usually the next quote refresh): unwind with a reduce-only IOC at the new touch, or hold as one leg against an opposite position on the leader book.
4. Cooldown after each shot — the second IOC into the same stale book mostly buys the level you just moved.

Starter logic

TRIGGER  = 4        # cents of divergence between the two mids
STALE_MS = 800      # laggard touch unchanged for this long = stale
COOLDOWN = 5        # seconds between shots
SIZE     = 20

on leader book update:
    move = mid(leader) - mid(laggard)
    if abs(move) < TRIGGER: return
    if ms_since_last_change(laggard.touch) < STALE_MS: return
    if cooling_down(): return

    if move > 0:   # leader repriced UP, laggard asks are stale-cheap
        px, side = best_ask(laggard), "BUY"
    else:          # leader repriced DOWN, laggard bids are stale-rich
        px, side = best_bid(laggard), "SELL"

    POST /v1/orders {market_id: laggard, side: side, type: "LIMIT",
                     tif: "IOC", price: px, qty: SIZE}
    start_cooldown()

on laggard reprice (mid within 1c of leader):
    unwind any open position with reduce-only IOC at the touch

Parameters

Risks & failure modes

1. The leader can be wrong: a one-print spike that mean-reverts leaves you positioned with the move already faded. The reduce-only exit on reconvergence caps this.
2. You ARE the adverse selection: this strategy is exactly what makers defend against, so expect quote-pulling — fill rates drop as makers get faster. Edge decays; re-measure weekly.
3. Near expiry the binary's fair moves violently with tiny spot moves — divergences are real but so is the risk; either widen TRIGGER 3x inside the last minute or stand down.
4. Don't run this against your own maker quotes from another strategy on the same account — same-account self-trade prevention will cancel your resting orders instead of filling.

Related: Lead-Lag Maker