Does Gamma Pinning Actually Exist? We Tested It on 2,288 Strike Touches

If you have spent any time around gamma exposure charts, you know the catechism. Where dealers are long gamma, their hedging leans against the market: they sell as price rises into a strike and buy as it falls away. Price is said to pin: stall, stick, gravitate. Where dealers are short gamma, the mechanics invert: hedging chases the move, and price is said to slip through those strikes and accelerate. It is a beautiful, mechanical story. It is quoted in newsletters, drawn on charts as "gamma walls," and sold in subscription products at hundreds of dollars a month.

What it mostly is not, in the retail and even semi-professional tooling world, is tested. The academic literature has documented expiration-day pinning at option strikes, but the day-to-day intraday claim (that a strike's current gamma exposure changes how price behaves when it gets there) is usually asserted, not measured. We run a platform built around gamma exposure, so this question is not academic for us. If GEX levels do not actually do anything to price, we would rather know than keep drawing them.

So we measured it. This article describes the test, including the part where our first analysis fooled us. How this kind of study goes wrong is half of what makes the result trustworthy.

The Experiment: First Touch of a Strike

The design is a classic event study. For each trading day we compute the gamma exposure at every SPX strike: open interest times gamma, signed by option type under the standard dealer convention (call-heavy strikes count as positive dealer gamma, put-heavy strikes as negative). Then we watch the index price minute by minute through the session. Every time price touches a strike for the first time that day, we record an event: how fast price was moving in the three minutes before the touch and how much it moved in the three minutes after.

If the pinning story is true, touches at strongly positive-GEX strikes should be followed by less price movement: the move decays into the strike. If the acceleration story is true, touches at negative-GEX strikes should be followed by more movement. Both are directly measurable.

Getting the data right

Two engineering details matter. First, resolution: strike interactions play out over minutes, so a 15-minute price series is useless: it misses the touch entirely whenever price crosses a strike and comes back between snapshots. We reconstructed a 1-minute SPX price path from index futures, calibrated bar by bar to the cash index level through the session, accurate to well inside our touch tolerance of a fraction of the strike spacing.

Second, timing: a strike's gamma changes through the day as spot and volatility move. Measuring a strike's GEX at 10:00 and testing a touch at 14:00 quietly turns the test into "how far did price travel today": a contaminated measure. Each touch in our study reads the strike's gamma exposure from the most recent option-chain snapshot before that touch. The exposure is measured minutes (not hours) before the event it is supposed to explain.

The result: 2,288 first-touch events across 200 trading days of SPX options data, split into a 156-day sample used to establish the effect and a fully held-out 44-day sample used to confirm it out-of-sample.

The Seductive First Answer and Why We Did Not Trust It

The raw result looks spectacular. Sort all touch events into quartiles by the strike's signed GEX and ask a simple question: after touching the strike, did price decelerate?

From 33% at the most negative-gamma strikes to 80% at the most positive: a textbook confirmation, monotonic, and nearly identical in both time periods. This is the kind of chart that sells subscriptions. It is also, taken at face value, substantially an illusion. If we had stopped here we would be publishing the same overclaim we are criticising.

Two problems hide inside it. The first is arithmetic: "deceleration" compares speed after the touch to speed before it, and any outcome defined as after minus before is mechanically biased against whatever came before: fast approaches "decelerate" by construction. The second is selection: which strikes get touched, and how fast price is moving when it reaches them, is determined by the same market conditions that set the GEX sign in the first place. Calm, drifting sessions produce both positive-gamma configurations and slow approaches; volatile sessions produce the opposite. The GEX quartiles were partly taking credit for the character of the day.

There is also a channel most tests destroy

Here is the subtlety that makes this study genuinely hard: you cannot simply regress the confound away either. In our data, signed GEX correlates meaningfully with the approach speed itself: price approaches high-positive-gamma strikes more slowly. Under the theory, that is not a nuisance: it is the same dampening effect operating before the touch. Statistically removing everything correlated with approach speed removes a large part of the thing being tested. Our own first pass made exactly this mistake and briefly concluded there was no effect at all.

The clean design does neither. It asks the question the theory actually poses: given the same incoming speed, on the same day, does price move less after touching a high-gamma strike than a low-gamma one? Approach speed is held fixed by comparing only touches with similar incoming velocity. Day conditions are held fixed by comparing strikes against other strikes touched the same day: news, volatility regime, and trend identical by construction. The outcome is the magnitude of price movement after the touch, which is what "pinning" actually claims to reduce.

The Verdict: Pinning Is Real and One-Sided

Finding one: positive-gamma pinning survives every control we threw at it. With approach speed matched, higher signed GEX predicts smaller post-touch price movement in nine of ten buckets across both samples. The effect holds when strikes are compared only against same-day strikes (statistically decisive in the main sample), it appears in every hour of the trading session rather than clustering at the open or close, and it holds separately for touches from below and touches from above. The dampening is strongest exactly where the theory says it should be: at the most call-heavy, dealer-long-gamma strikes.

Finding two: the effect follows the sign, not the size. When signed GEX and raw unsigned size (total OI × gamma, calls and puts lumped together) compete head-to-head in the same statistical test, the signed measure wins in both periods; the size measure does not survive out-of-sample. A big options wall is not, by itself, a pin candidate. A big call-heavy wall is. Two strikes with identical total open interest can behave completely differently depending on their call/put composition. That is why a gamma chart that shows only magnitude, without the net sign per strike, is throwing away the predictive half of the information.

Finding three: the negative-gamma half of the folklore did not show up. Touches at put-heavy, negative-GEX strikes showed no measurable acceleration: not in the main sample, not out-of-sample, not in any sub-analysis. Price does not detectably "slip through" negative-gamma strikes faster than comparable strikes. In our data the folklore's symmetric story (stabilising on one side, destabilising on the other) is half right.

Why might the negative side be missing?

Two candidate explanations, both plausible. The economic one: long-gamma hedging is passive and mechanical (resting orders that lean against the market simply by existing) while short-gamma hedging is discretionary, spread over time, and often internalised. Its footprint at the individual strike level is diffuse as a result. The measurement one: the signed convention itself (dealers long the calls, short the puts) is the industry's standard assumption about who holds what. It is weakest precisely on the put side, where institutional hedging patterns are most varied. If the sign attribution is noisy for puts, a real effect could be invisible to this test. Distinguishing these explanations requires trade-level opening/closing data. That is where our research goes next.

Every GEX vendor tells you positive gamma stabilises and negative gamma accelerates. We tested both halves. One is measurable. The other, at minimum, cannot be seen at the strike level with the data everyone in this industry is actually using.

What This Means When You Read a Gamma Chart

Treat large positive net-GEX strikes as tested pin candidates. When price approaches a strongly call-heavy strike, the measured tendency is for the move to lose energy there: given the same incoming speed, post-touch movement contracts. These are the levels worth watching as potential stall and reversal zones, and this is now a claim backed by a controlled event study rather than by mechanism-story alone.

Treat negative net-GEX strikes as positioning information, not as trade signals. Put-heavy strikes tell you where downside protection and hedging pressure are concentrated. Genuinely useful context. What they are not, on our evidence, is acceleration zones. A strategy built on "price will slip through the red bars faster" is built on the untested half of the folklore.

Insist on the signed view. Since the call/put composition of a strike (not its total size) carries the predictive information, a per-strike net gamma display (positive and negative bars, not one lumped magnitude) is the minimum standard for a chart that reflects what is actually measurable. This is how our own dashboard has always displayed it, and this study is the reason we will keep it that way.

Keep the effect size honest. This is a statistically robust tendency across thousands of events. It is not a law of nature at any single strike on any single day. Pinning shifts the distribution of outcomes; it does not guarantee one. Anyone who tells you a gamma level will hold is selling something stronger than the evidence.

Limitations, Openly

One instrument, one year. The study covers SPX over roughly 200 trading days spanning multiple volatility regimes. SPX is where gamma exposure analysis is most complete and most heavily used, but extension to other indices and single names is untested and on the roadmap.

The gamma exposure measure is the industry-standard estimate. Open interest times gamma, signed by the conventional dealer assumption: the same construction underlying essentially every commercial GEX product. That makes the result directly relevant to the charts people actually use. But it also inherits the convention's blind spots, especially on the put side, as discussed above.

Moderate effect strength. The correlations are consistent and highly significant in aggregate, and modest per event. This is exactly what you should expect from a real structural effect operating inside a market full of other forces. It is worth stating because raw uncorrected analyses (like our own first pass, and like most vendor charts) make the effect look several times stronger than it is.

What comes next. The single biggest upgrade to this line of research is trade-level opening/closing classification: knowing which open interest actually sits on dealer books rather than assuming it. That data exists, it is on our integration roadmap, and it converts the sign assumption into a measurement. If the negative-gamma effect is real but hidden by attribution noise, that is what will reveal it.

Conclusion

Gamma pinning is not folklore. But the folklore version of it is wrong in an important way. Price measurably calms when it reaches strikes where dealers are net long gamma; the effect survives controls for approach speed, day regime, time of day, and direction. It replicates out-of-sample. The mirror-image claim about negative gamma does not appear in the data at all. And the information that predicts pinning is the net sign of a strike's gamma exposure, not the raw size of the wall.

For us, this study did double duty. It validated the foundation our platform is built on: that a continuously updated, signed, per-strike gamma exposure surface describes something real about how price behaves. And it drew a bright, evidence-based line between what a gamma chart can honestly promise and what it cannot. We think that line should be industry standard. Until it is, we will keep publishing where our own numbers land on it.

This study was built on the intraday open interest infrastructure described in our companion article, Real-Time Gamma Exposure: How We Estimate Intraday Options Open Interest with Machine Learning: the model that keeps the OI surface behind these gamma levels current through the session instead of frozen at yesterday's close.