How the numbers are made.

What the site computes, where the data comes from, how much history sits behind each measurement, and what none of it can tell you.

1. Dealer gamma, in plain words

When you buy an option, the other side is usually a market maker (a “dealer”). Dealers hedge by trading the underlying, and how much they have to trade when the price moves depends on the option’s gamma.

We assume dealers are long the calls and short the puts: investors mostly sell calls (covered calls, overwriting) and mostly buy puts (protection), so the dealer ends up on the other side of each. Under that assumption, call open interest counts as positive gamma for dealers and put open interest as negative.

This is a convention, not an observation. Nobody outside the dealer knows its true book. When customers are actually buying calls (for example in a speculative rally), the real sign at those strikes can be the opposite of what we show.

For each strike we compute open interest × 100 × gamma × spot: the dollars of stock a dealer would have to buy or sell for a one-point move. Calls are added and puts subtracted to give the net figure. Labels on the heatmap are in that unit (“$ per 1-point move”).

• Positive net gamma means dealers hedge against the move (sell rallies, buy dips), which tends to damp it.
• Negative net gamma means they hedge with the move, which tends to amplify it.
• Gamma flip is the price where the summed net figure changes sign. The king node is the strike with the largest absolute exposure.

2. Where the data comes from

• Option chains (strikes, open interest, implied volatility, greeks) come from Massive. On the current plan they are 15-minute delayed. When greeks are missing we compute gamma with Black-Scholes from the implied volatility.
• Open interest is published once a day, after the close. Intraday, the levels move with price and implied volatility, not with new positions opened that day. The “OI + volume” view adds today’s volume as a rough estimate and is labelled as an estimate.
• Strikes within ±15% of spot are included. The display widens or narrows from there, and always keeps the largest nodes.
• 0DTE means the nearest expiry on or after the active session. At 4:00 PM ET it rolls to the next session’s expiry.
• Index prices (SPX, NDX, RUT) come from Massive’s index feed. One background job refreshes them about every 10 seconds and every page reads that shared copy. Each page shows whether the reading is pre-market, live, closed or delayed, and when it was taken.
• Daily price history for the measurements comes from public daily bars (Yahoo Finance).

3. How much history there really is

Read from the measurement files the site ships with, not typed in by hand. Dates are trading days.

Measurement	Data	Span	Size	In / out of sample
Quantitative Exhaustion levels (PB1, PB2, buy zones)run Sep 1, 2026	Daily price bars	Aug 22, 2011 – Aug 31, 2026 (15.0 years)	432,133 bars, 116 tickers	In-sample for the rules: the detector was designed with this history available. The part before Aug 3, 2020 is out-of-sample for the claim that was originally published about it, and is reported separately.
Scanner signals (direction after a signal)run Sep 1, 2026	Daily price bars	Aug 22, 2011 – Aug 28, 2026 (15.0 years)	435,907 bars, 117 tickers	In-sample. The signal rules existed before the test and were not re-tuned on it, but the same history was visible while they were written.
Confluence score (rebuilt from 6 of its 10 inputs)run Sep 1, 2026	Daily price, volume and VIX bars	Nov 3, 2016 – Aug 31, 2026 (9.8 years)	2,468 scored S&P 500 sessions	In-sample. The four option-derived inputs (38% of the score's weight) have no history and are absent from the test.
Level re-test decay (price-only stand-in for gamma nodes)run Sep 22, 2026	Daily price bars, SPY, QQQ, IWM, DIA	not recorded – not recorded	11,792 touches	In-sample. It measures swing-pivot levels, not dealer gamma nodes, because there is no option history to test nodes on.

Option-derived readings have no long history. Historical open interest and implied volatility per contract are not available on our data plan and cannot be backfilled. The dealer gamma levels, vanna, skew and the VIX term structure therefore have only as much history as the site has recorded itself: one end-of-day snapshot of the nodes per ticker per session, kept since recording began. That is months, not years. No historical result on this site is a test of the gamma levels themselves.

4. How the measurements are run

• Walk-forward, no look-ahead. Each signal is computed using only bars up to that day, by the same code the product runs. A test perturbs every bar after a random cut and checks that nothing before the cut changes.
• Against a base rate. A hit rate means nothing alone, so every rate is compared with the same outcome on signal-free bars of the same tickers, in the same calendar months.
• With intervals. Each result carries its sample size, a Wilson interval and a month-block bootstrap interval. If the interval includes the base rate, the result is reported as does not separate, which is the most common outcome.
• In-sample vs out-of-sample. Rules were not re-fitted to these tests, but they were written by people who could see the same history, so we treat the results as in-sample. Out-of-sample evidence only comes from sessions after a measurement was run. Those are tracked separately on the results page, where a signal with too few trades says “no verdict yet”.

5. What this is not

None of the levels or signals on this site is presented as predictive. They describe where hedging pressure sits under a stated assumption, and the historical measurements describe what happened in the past, mostly without a measurable edge. This is analytics for education and information, not investment advice.