> That might looks suspicious sure, but there may be other factors (not reflected in the book) influencing a change of desire on my part
I think your example would easily satisfy regulators. But you have to have the logging with real time stamps to prove it or you need to be able to reconstruct the internal state by replaying the market data against that version of the code.
Same goes for other pairs trades or other non-book signals. For any automated trading on regulated exchanges, you need to be able to explain to regulators why you chose to send an order. Otherwise you’re basically admitting that you don’t understand / have control over your algorithm, which they will obviously object to.
> I've never come across a strategy that responds to volume changes far from the BBO.
I agree it’s unlikely that a deep quote will provoke an immediate action. But sustained changes in the book will over time get aggregated into moving averages etc and influence behaviors in the long term (for whatever timescale may be appropriate).
> Quantity change far away from the action is mostly noise.
I agree that individual quotes are noisy, and it’s hard to extract signal from the noise. A limit order book by definition is supposed to quantify an aggregated demand to buy/sell, and thus it should be meaningful to aggregate over it to construct a distribution. Spoofing distorts that.
No one gets investigated for spoofing for sending a single order. It’s a persistent pattern that stands out from the noise.