Bluffing, Signalling and Reading Intent

A trader places an unusually large order. Is it a genuine, high-conviction bet, or a bluff meant to move the market before a smaller, opposite order slips through? You can't ask, and the size alone doesn't settle it — a bluffer can place a large order too. What actually distinguishes a real signal from noise isn't the message itself, it's whether sending that message is cheap enough for a bluffer to fake convincingly, or costly enough that only someone genuine would bother.

Cheap talk versus costly signals

Cheap talk is a message that costs the sender nothing to send regardless of whether it's true — a trader simply saying "I think this is going higher" costs the same whether they believe it or not, so in a game where their interests aren't aligned with yours, cheap talk carries little to no reliable information; a rational listener should mostly ignore it. A costly signal, by contrast, is only worth sending if you're telling the truth — because sending it when you're bluffing costs you more than the bluff could gain. The classic non-trading example is a large, non-refundable order deposit: only a buyer who genuinely intends to complete the purchase would risk forfeiting it, so posting the deposit itself, not the accompanying promise, is what's actually informative.

Worked example

A market maker is deciding whether to trust a large sell order as a genuine signal that the seller expects the price to fall, versus dismissing it as an attempt to spook the price down before the seller buys back cheaper (a bluff). Suppose the seller's true cost of placing a real sell order (locking in a loss if wrong) is $2 per share if the bet fails, while a bluffer who has no real position faces slippage and re-entry costs of only $0.20 per share if caught (they can unwind quickly since they never had real exposure). If the potential gain from successfully spooking the price is $0.50 per share, a bluffer happily pays $0.20 to try for $0.50 — bluffing is profitable for them, so the signal is not costly enough to be credible, and the market maker is right to discount it. Now suppose exchange rules or position limits push the bluffer's unwind cost to $0.80 per share (say, wider spreads for rapid round-trips) — now paying $0.80 to chase a $0.50 gain is a losing bet for the bluffer, so a rational bluffer stops trying, and any large sell order that does appear can be trusted as more likely genuine. The exact same order size becomes informative or not, purely based on whether the cost of faking it exceeds the payoff from faking it.

What this means in practice

Traders read order flow, quote behavior, and even body language in voice-broker markets through this lens without necessarily naming it: the question is never "what did they say or do," it's "would someone with the opposite intent have found it worthwhile to do the same thing." Size alone rarely settles it, because size is often cheap to fake; what actually carries information is size combined with a cost structure (capital committed, reputational risk, exchange fees on rapid reversal) that a bluffer wouldn't willingly eat.