How GammaFlux Estimates Intraday Gamma

Honest walkthrough of our proprietary model. What we use, what we infer, what we don’t know.

The honest framing first

We want to be upfront: nobody outside of the dealer desks themselves has perfect visibility into net dealer gamma positioning at any given moment. The data simply isn’t public. Every third-party gamma tool in the world — ours included — is producing an estimate based on models that infer positioning from the data that is public.

What separates good estimates from bad ones is:

1. The quality and timeliness of the input data
2. The soundness of the modeling assumptions
3. How honestly the tool communicates its limitations

The data streams we use

What the model does with those inputs

At the highest level: the open-interest profile gives us the standing structural map, and the live directional order-flow engine builds a continuously-updated net-gamma surface on top of it. We don’t lean on a once-a-day OI snapshot for the parts of the read that move — the flow surface is recomputed each cycle from how options are actually trading right now.

From those two layers we derive the levels you see in the dashboard. The Call Wall, Put Wall, and OI Flip are read off open interest. The magnets (P1, N1, AG1, AG2) and the Flow Flip are read off live flow — the Flow Flip starts the day at the OI Flip and migrates as flow rewrites the regime. Each level also carries a convergence score telling you how many independent signals from the chain agree with its placement. See the convergence docfor how that’s calculated.

The levels track the near-term (front) expiration intraday, plus a structural aggregate of the front ~5 expirations, so you get both today’s 0DTE gravity and the broader standing structure on one chart.

The model runs server-side, updates continuously during market hours, and the output is pushed to your TradingView chart via the Chrome extension and to the web dashboard in real time.

What we deliberately don’t share

We’re not going to publish the exact mathematical formulation of the model. The weighting scheme, smoothing parameters, and feature engineering choices are what took time to build and what make the output reliable. If we published them, a well-resourced competitor could clone the tool in a week.

What we WILL tell you — and what this doc covers — is every category of input we use, the reasoning behind the approach, and where the approach has limitations. We want you to understand what the numbers on your chart represent, even if the specific implementation stays proprietary.

The limitations (honest list)

Every model has tradeoffs. Here are ours:

• info
  It’s an estimate, not a measurement.We don’t have God-mode access to dealer books. We’re inferring positioning from public data. The estimate is good — we’ve validated it against end-of-day OI prints and it converges to reality overnight — but it is an estimate.
• info
  Order-flow direction is inferred, not certified.We classify each trade as buyer- or seller-initiated from price and quote context, which is far better than assuming 50/50, but it isn’t perfect. Spread trades, complex multi-leg structures, and unusual prints can be misattributed, and not every contract that trades is a dealer-hedging flow.
• info
  Unusual conditions degrade quality.FOMC announcements, major gap days, fed minutes releases, and holiday sessions can create conditions where the model’s assumptions break down. The levels are still computed, but treat them with extra skepticism during those moments.
• info
  We refresh on a cycle, not tick-by-tick. The flow surface is recomputed continuously through the session, but on a polling cadence rather than as a true sub-second stream. Most moves in the underlying complete within our refresh window, so you see them. Very fast moves can get partially aliased.
• info
  End-of-day OI is still ground truth.When tonight’s OI prints, it settles whatever was happening during the day. If our intraday estimate disagreed with where actual positioning ended up, we accept the OI and start fresh tomorrow. We don’t pretend our estimate overrides reality.

Academic research backing this approach

The premise underlying GammaFlux — that options market maker gamma hedging creates measurable, predictable intraday price impacts that retail traders can usefully observe — isn’t a pet theory. It’s well-supported in top-tier peer-reviewed finance research.

The most directly relevant published work is Baltussen, Da, Lammers & Martens (2021), “Hedging Demand and Market Intraday Momentum” (Journal of Financial Economics, Vol 142, pp. 377-403). Using 45 years of data across equities, bonds, commodities, and currencies, they find that returns in the last 30 minutes before the market close are positively predicted by returns earlier in the day — and they trace this directly to gamma hedging activity by options market makers and leveraged ETF managers. Crucially, the effect reverses over subsequent days, indicating transitory hedging-driven price pressure rather than fundamental information being incorporated. This is the empirical foundation for the entire intraday-gamma category that GammaFlux operates in.

Recent extending work includes Baig, Strong & Zaynutdinova (2025), “Seeking Gamma: Lessons from the Meme Frenzy” (accepted to the 2026 AEA conference), which extends the Kyle (1985) microstructure model to include inventory constraints and identifies 669 gamma squeeze events across US equities from 2019-2023 with sustained abnormal returns averaging roughly 5% in the month following initiation. They explicitly distinguish ephemeral “Net Delta Volume” (short-term speculative pressure that dissipates quickly) from persistent “Net Delta Open Interest”, which they describe as “mandatory buying power” that dealers are contractually obligated to maintain and rebalance daily — a distinction that maps directly onto our two-layer design: the live order-flow engine tracks the fast, directional flow, while the open-interest map holds the standing structural positioning.

GammaFlux’s implementation differs from both papers — we read public options open interest and live directional order flow on indices and ETFs, recomputed continuously, rather than the academic-grade Consolidated Audit Trail data they have access to — but the theoretical framework is the same. Additional context comes from Johnson, Liang & Liu (2018), “What Drives Index Options Exposures?” (Review of Finance), and the broader demand-based option pricing literature dating back to Garleanu, Pedersen & Poteshman (2009).

To be clear about what this body of work does and doesn’t prove: the academic literature establishes that gamma hedging creates measurable price effects. It does NOT prove that any specific GammaFlux level (Call Wall, Put Wall, OI Flip, Flow Flip, etc.) will hold or break in any specific instance. Our levels are model outputs that reflect estimated dealer positioning at a point in time — informed by the mechanisms these papers describe, but not guaranteed predictions.

How to use this responsibly

GammaFlux gives you a structural read on where dealer hedging pressure is concentrating. That is useful context. It is not a trading signal. Levels can hold, break, shift, or prove irrelevant depending on broader market conditions and news.

The right way to think about it:

• checkUse high-convergence levels as context for planning trades, not as triggers to enter them blindly
• checkWatch how levels shift during the session — level migration is often more informative than the static levels themselves
• checkPay attention to the gamma regime (positive vs. negative) for volatility context
• checkCombine GammaFlux with your own chart read — we’re a structural layer, not a complete system

Why we built this

The original GammaFlux came out of frustration with paying for expensive static-OI gamma tools that were obviously missing 0DTE flow. Watching a $150/month dashboard sit frozen on yesterday’s levels while actual dealer positioning was clearly shifting in real-time didn’t feel like a premium product — it felt like the tool was just charging more for worse information.

We built this because we wanted something that actually responded to the market as it unfolded. The result is a tool designed for traders who care about what’s happening at 11:47 AM, not what happened at 4:00 PM yesterday.