Why gamma walls do not produce edges

The observation is straightforward and correct. Open interest clusters at round strikes, those clusters are visible in public data, and prices do sometimes stall near them. From this an inference is routinely drawn: the strike acts as a barrier, and a strategy can be built around it.

The inference does not survive being stated precisely, and the reasons are instructive beyond this particular case.

The hypothesis is usually unfalsifiable

"Price will be attracted to the wall" and "the wall will act as resistance" are the same claim about a level, made in opposite directions, and in practice both are invoked after the fact. If price approaches and stalls, the wall held. If it approaches and passes, the wall broke and the move accelerated. If it never approaches, the wall repelled it from a distance.

Every outcome confirms the hypothesis, which means it forecasts nothing. This is not a technicality. A rule that can only be evaluated retrospectively cannot be traded, because the trader has to act before knowing which of the three descriptions will apply.

Making it testable requires committing in advance: at what distance the position is taken, in which direction, with what invalidation, and over what horizon. Almost all published analysis skips this step, and most of the apparent evidence disappears once it is taken.

The level is not exogenous

The barrier argument treats the strike as an external feature acting on price. The causation runs at least partly the other way.

Open interest concentrates at round numbers because round numbers are where people choose to transact, and round numbers are also where price behaviour clusters for reasons entirely unrelated to options: they are natural reference levels, they attract resting orders, and they anchor stop placement. A study finding that price stalls near a large strike has found that price stalls near round numbers, unless it has controlled for roundness.

That control is rarely present, and it is the single most important one. Comparing behaviour at large-open-interest round strikes against behaviour at round levels with little open interest is the minimum test that distinguishes the options explanation from the far simpler alternative. When that comparison is run, the difference between the two groups is usually much smaller than the raw effect and frequently within the noise.

The mechanism is real but does not scale to a rule

The underlying hedging mechanism is genuine. A dealer short options near a strike, hedging delta as spot moves, does trade against the move as gamma peaks near expiry. That flow exists.

Three things stop it becoming a rule.

The sign is uncertain. The mechanism depends on which side the dealer holds, which is not observable. As covered in gamma exposure: signal or description, the sign convention is a heuristic, and it is least reliable at exactly the large institutional strikes where the walls are largest.

The magnitude is small relative to flow. Hedging flow around a strike is meaningful compared to a quiet hour and negligible compared to a macro release or an index rebalance. A strategy built on it is taking a position that the dominant flow that day will be the hedging flow, which is a claim about everything else in the market, not about the option chain.

It is concentrated in a narrow window. Gamma peaks near expiry and near the money. The conditions where the mechanism is strong are a small fraction of observations, and a backtest run over all observations averages a strong effect in a rare state with no effect in a common one. The result is a small positive number that looks like a weak edge and is an artefact of pooling.

What a real test requires

The construction that would settle it is not complicated, only laborious.

Define the event precisely: a strike above some percentile of open interest for that expiry, spot within a stated distance, a stated number of days to expiry. Fix the entry rule and the horizon in advance. Then build two control groups, not one: round levels with low open interest, and high open interest strikes at non-round levels. The effect attributable to options is what remains after both controls.

Condition the results rather than pooling them. By days to expiry, by realised volatility regime, by whether the level was approached from above or below. The hypothesis, if it has content, predicts a specific pattern across these conditions, and a strategy that works in all of them equally is more likely to be picking up something else.

Report the whole surface of results, not the best cell. With this many conditioning dimensions, some cell will look strong by construction, and the number of cells examined is the number that has to be corrected for.

The general lesson

The failure here is not about options. It is that a visible feature of the data, a plausible mechanism, and an after-the-fact narrative combine into something that feels like evidence and contains none.

Each component is individually reasonable. Open interest does cluster. Hedging does happen. Price does sometimes stall. The error is treating co-occurrence of three true statements as support for a fourth that was never tested, and it is the most common way that a descriptive market observation becomes a strategy that does not work.

The discipline that catches it is stating in advance what would count as the hypothesis failing. Under that discipline the barrier claim is difficult to formulate at all, and that difficulty is the finding.