How We Estimate Dealer Positioning

Every gamma figure on TradeAI News is an estimate built from published inputs and one stated assumption. A fair objection to any gamma exposure (GEX) number is that the assumption can be wrong. It can. This page sets out what we use, what we assume, what we leave out and how to read the label on our charts, so you can judge the numbers for yourself.

Where the open interest comes from

Open interest is the number of option contracts still open at the end of a session. The Options Clearing Corporation (OCC) publishes it after settlement, the morning after the session it describes. Our options data provider republishes those figures later, so they reach us before the open of the second trading session after the one they describe. Measured over 24 sessions from Sep 25 to Oct 28, 2026, each new set arrived between 7:35 and 8:22 AM ET, with no gaps.

Fetching more often would not make it fresher, because the delay sits at the source. Instead, every gamma view names the session its open interest describes. Quotes and spot are newer than that, and combining them does not make them contemporaneous: contracts opened or closed since that session are not in the count.

The positioning assumption, and its limits

Open interest says how many contracts are open at each strike. It does not say who is long and who is short. To turn a count into an exposure we need a sign, and we use the usual published convention: call gamma counts as positive and put gamma as negative, which assumes dealers are long the open calls and short the open puts. It is an assumption, not an observed position, and every sign on our charts follows from it.

Where it is weakest:

• Customers also open long calls and short puts. Where that dominates at a strike, dealers there are short calls or long puts, and the real sign is the opposite of ours.
• Spreads share one chain. A multi-leg position adds open interest on both sides, and open interest does not link the legs.
• Not every contract sits with a dealer. Two customers can hold opposite sides of the same contract.
• The count is two sessions old, as described above.

We keep the convention fixed instead of adjusting it strike by strike, because no published data tells us when to reverse it. A fixed, stated rule can be checked; a rule tuned by judgement cannot. The size of open interest at each strike is observed. The sign is assumed.

How the six greeks are computed

For each contract with a two-sided quote, we take the bid-ask midpoint and back-solve implied volatility with Black-Scholes, using a 4.5% risk-free rate and no dividend yield. Each greek is multiplied by open interest and the contract multiplier of 100, signed by the convention above, and summed by strike and by expiration. Contracts without open interest or without a usable quote are left out, and each view says how many contracts were used.

• Delta: delta × open interest × 100 × spot, the dollar-equivalent exposure to the underlying. Under the convention both legs add positive delta, so its size carries the information, not its sign.
• Gamma: gamma × open interest × 100 × spot² × 1%, the change in delta exposure for a 1% move.
• Vega: the change in the value of the open options for a 1-point change in implied volatility.
• Vanna: the change in delta exposure for a 1-point change in implied volatility, spot held fixed.
• Charm: the change in delta exposure per calendar day as expiration approaches.
• Vomma: the change in vega exposure for a 1-point change in implied volatility.

Black-Scholes assumes European exercise, while US equity and ETF options are American, and dividends are not modelled; both add model error. Implied volatility from a wide quote is less reliable, so illiquid strikes carry more error than liquid ones. The Estimated Gamma Exposure methodology shows the formula next to the public SPY view; the gamma flip guide covers the zero crossing.

What we do not do

• No direction. None of these figures forecasts where price goes, and none of them is a trade recommendation. Why we do not predict direction.
• No signed flow. We do not infer which side started each trade, and we do not track whether today's trades opened or closed positions. Exposure comes from open interest alone.
• No same-day expirations. Contracts expiring on the current New York date are left out during their own session. Estimating them would need intraday trade data that this method does not use.

How to read “Estimated, not observed”

You will see Estimated, not observed on the Gamma Heatmap, on the By expiration block and on the gamma cards in Today and My Exposure, and each greek's note in the dashboard opens with “Estimate, not an observation.” Read them literally; they apply to every gamma figure we publish:

• Estimated: computed by us from quotes, open interest and the sign convention. Change the convention and the number changes.
• Not observed: nobody outside a dealer sees its book, and that includes us. No chart shows actual dealer inventory.
• Size, not direction: a large negative net gamma describes how hedging would behave if the assumption held. It says nothing about the next move.

Each view in the Gamma Suite also dates two things: when the figure was calculated and which session the open interest describes. A wall or a flip is a description of where open interest concentrates under a stated rule, as of that session.

How this compares with methods built on trades

Another family of methods starts from individual option trades instead of open interest. Each trade is assigned to the customer's side or the dealer's, usually by comparing its price with the quote at that moment, and the results are accumulated through the day into signed positions.

That approach has a real advantage: it can see positions opened today, including in same-day expirations, which open interest cannot. It also rests on inferences of its own. A trade at the midpoint has no clear side. The trade record does not say whether a trade opened or closed a position. Multi-leg orders print as separate trades. And knowing the customer's side says nothing about what the dealer did next. It also needs a full intraday trade feed.

Our method gives up the same-day view in exchange for inputs anyone can check: published open interest, quoted prices and one stated rule. Neither family observes dealer inventory. Both make an inference about it, and each should say which one. The options market structure guide covers the inputs in more detail.

Frequently asked questions

We measure significance and magnitude, never direction.