KILLED · unreliable + risky
Hypothesis
Large fleeting orders that cancel before execution (spoofs) signal the spoofer’s true intent; fading the displayed pressure captures the real move.
Math
$$ \text{spoof score} = \frac{\text{displayed size cancelled before trade}}{\text{total displayed size}} $$
Method
Flag rapid place-then-cancel clusters; trade opposite the fake pressure.
Results
| Spoof detection | noisy on public data |
| False positives (genuine cancels) | high |
| Edge | inconsistent |
Distinguishing manipulative spoofs from ordinary order management on public data is unreliable (most cancels are benign), and building a strategy around reading manipulators is fragile and adversarial. Killed on both robustness and principle.
Trying to out-read manipulators is a game where the other side controls the signal. Most cancelled orders are not spoofs, and the ones that are can flip on you.