From Regime-Conditioned Option Surfaces to Spread Shape Mispricing

Raw implied volatility does not tell the whole story

Options are not priced only by direction. Option prices reflect expected volatility, time to expiration, skew, convexity, liquidity, demand for protection, supply-demand pressure, and the current market environment.

That is why the simple question, “Is implied volatility high or low?”, is usually incomplete. High relative to what? A long-term average? Recent realized volatility? A Black-Scholes baseline? Those comparisons can be useful, but they often miss the market regime.

The same implied volatility level can mean different things in different environments. A 25% implied volatility may be expensive in a calm growth regime, normal in a transition regime, and cheap in a defensive or stress regime. The same logic applies to put skew and call wing pricing.

The regime grid behind the option surface

MARFIN was originally designed as a drawdown-aware market regime and exposure framework. Its purpose is not to predict every market move. Its purpose is to organize market conditions into a structured state: what kind of environment the market is in, how aggressive or defensive exposure should be, and how strong the internal score is.

For the option-surface research, we use a grid based on 4 regimes, 4 allocation states, 4 score buckets, and 5 DTE buckets. This creates 320 unique state/DTE nodes.

Market regimes

Allocation states

Score buckets

The MARFIN score is bucketed instead of treated as a fragile point estimate. This reduces noise and lets us compare options in broader, more stable state groups.

DTE buckets and moneyness nodes

The surface is built across 7D, 14D, 30D, 45D, and 60D expiration buckets. The moneyness nodes cover the main parts of the volatility smile: Put 90M, Put 95M, Put 98M, ATM, Call 102M, Call 105M, and Call 110M.

In simple terms, this means the research does not look at one option. It looks at a full structured grid: market state, allocation state, score bucket, expiration window, and option moneyness.

MARFIN Fair Surface: a theoretical fair-risk layer

The first layer is the MARFIN Fair Surface. It estimates what the option surface could look like if pricing were based on regime-conditioned forward behavior, realized volatility, VXN-related information, tail behavior, and drift-neutralized payoff logic.

It answers a fair-risk question:

What level of implied volatility looks theoretically fair for this MARFIN state, DTE, and moneyness node?

The Fair Surface is not designed to replicate the market perfectly. It is a theoretical layer. It gives us a state-aware estimate of risk before we ask how the market historically priced that risk.

MARFIN Market-Expected Surface: the historical market norm

The second layer is the MARFIN Market-Expected Surface. This is not theoretical fair value. It is the implied-volatility surface the market historically assigned to similar MARFIN states.

That distinction matters. Markets do not have to trade at theoretical fair value. They can price a premium for fear, liquidity, protection demand, crash risk, positioning, or structural supply-demand imbalances.

This gives us a cleaner baseline. Instead of asking whether today’s IV is high compared with a broad average, we ask whether today’s IV is high compared with what the market usually paid in similar MARFIN conditions.

What the surface research showed

The first research step was structural: can the Fair Surface and the Market-Expected Surface be joined cleanly into one consistent analytical framework?

This is a basic but important result. Without structural consistency, any later analysis of premium, skew, shape, or spread mispricing would be unreliable.

State specificity and fallback levels

Not every market state appears equally often in history. Some states have enough observations for a very specific match; others require a broader grouping. We do not hide that. The Market-Expected Surface uses fallback levels to describe how specific the historical comparison is.

This matters because it allows us to separate the signal from its confidence. A future dashboard can show not only the deviation, but also how state-specific the historical baseline is.

Market-Expected minus Fair: the normal market risk premium

Once both surfaces exist, we can compare them:

Market-Expected IV − MARFIN Fair IV

This is not automatically a market error. It can be interpreted as the historical market risk premium over the fair-risk layer.

The important finding is that this premium is not constant. It changes by DTE and by moneyness node.

Example: Put 90M median premium

Example: ATM median premium

The conclusion is simple: fair value and market expectation should not be collapsed into the same number. Fair value estimates model-implied risk. Market expectation estimates how the options market historically priced similar conditions.

From option surfaces to a more stationary deviation chart

After building the Market-Expected Surface, the next step is to compare the current market with that state-conditioned historical norm:

Current Market IV − MARFIN Market-Expected IV

This is different from raw IV. It asks how far today’s market is from what the market usually paid in similar MARFIN states.

Raw IV is often not stationary. It can stay high in stress regimes and low in calm regimes. But a deviation from a regime-conditioned norm can be more stable because part of the regime effect has already been removed.

Shape Mispricing: measuring distortion in the volatility smile

A full option surface is not only about ATM IV. The shape of the smile matters: put wings, call wings, and the relative premium of each node versus ATM.

First, we calculate the current market skew for each node:

Market Skew(node) = Current Market IV(node) − Current Market IV(ATM)

Then we calculate the normal historical skew for similar MARFIN states:

MarketExpected Skew(node) = MarketExpected IV(node) − MarketExpected IV(ATM)

Shape Mispricing is the difference between the two:

Shape Mispricing(node) = Market Skew(node) − MarketExpected Skew(node)

In plain English, Shape Mispricing shows whether a specific point of the volatility smile is rich or cheap relative to its normal regime-conditioned shape.

Spread Shape Mispricing: the cleaner spread-level signal

Many option strategies depend on the relative price of two legs, not on the absolute IV of one option. That is why the next step is to analyze the spread itself.

For a vertical spread, we compare the Shape Mispricing of the short leg and the long leg:

Spread Shape Mispricing = Shape Mispricing(short leg) − Shape Mispricing(long leg)

This asks a practical question: is the leg we would conceptually sell richer than the leg we would conceptually buy, relative to the normal smile shape for similar MARFIN states?

For put credit spreads, this is especially intuitive. If the closer-to-ATM put leg becomes unusually rich versus the farther OTM put leg, the spread shape may be distorted in a way that can later normalize.

Z-score: identifying unusual spread distortion

The raw Spread Shape Mispricing value is useful, but we also need to know whether the value is unusual for that specific spread. We therefore calculate a rolling mean, rolling standard deviation, and z-score:

Z = (Spread Shape Mispricing − Rolling Mean) / Rolling Std

In the convergence study, the test logic was simple:

• Enter a research signal when |z-score| >= 2.
• Expect the spread mispricing to compress toward normal.
• Exit when |z-score| <= 0.5 or after a maximum of 20 trading days.

What the Spread Shape Mispricing research showed

We tested several vertical spread structures on historical QQQ option-chain data across 7D, 14D, 30D, 45D, and 60D buckets.

For each spread, we measured Spread Shape Mispricing, converted it into a z-score, and then checked what happened after extreme deviations.

Overall entry/exit simulation

This is the main evidence that Spread Shape Mispricing behaves like a mean-reverting spread. After extreme z-score readings, the mispricing usually compressed and moved toward zero.

Put-side vs call-side

Both sides showed mean reversion, but put-side spreads were stronger. That makes sense: put skew is often more directly connected to fear, protection demand, drawdown risk, and defensive market regimes.

The strongest put-spread candidates

The most interesting research candidates were put spreads in the 14D, 30D, and 45D zones. We treat 60D more carefully when the number of signals is smaller, but it is still useful to monitor.

Forward convergence after extreme z-scores

We also checked what happened after extreme |z| >= 2 readings over fixed forward windows. For put spreads, the results were consistent.

This supports the same idea from a different angle: after extreme spread-shape deviations, the mispricing most often compressed and moved back toward zero.

The complete research logic

The full chain is important because it shows that this is not a random indicator. It starts with market regime classification and ends with spread-level mean reversion.

1. Classify the market state. Use MARFIN regime, allocation state, and score bucket.
2. Build the Fair Surface. Estimate theoretical fair risk by state, DTE, and moneyness.
3. Build the Market-Expected Surface. Estimate how the market historically priced similar MARFIN states.
4. Measure normal risk premium. Compare Market-Expected IV with Fair IV.
5. Measure current deviation. Compare today’s IV with the Market-Expected norm.
6. Move from level to shape. Compare each smile node with ATM and calculate Shape Mispricing.
7. Move from node to spread. Calculate Spread Shape Mispricing between the short and long legs.
8. Normalize with z-score. Identify unusually large distortions.
9. Test mean reversion. Check whether the distortion compresses and returns toward zero.

What this means for MARFIN option analytics

The research suggests that MARFIN can be extended from a drawdown-aware market regime framework into a deeper option analytics framework.

The most important shift is this: we are not only asking whether implied volatility is high or low. We are asking whether a specific part of the option surface is high or low relative to the market’s historical behavior in similar MARFIN states.

That creates a more disciplined way to study option risk. It can support future dashboards for regime-conditioned option surfaces, Shape Mispricing maps, spread mispricing rankings, and stationary deviation charts.

What we are not claiming

We are not claiming that MARFIN predicts every option price movement. We are not claiming that every z-score event is directly tradable. We are not claiming that compression in volatility points automatically equals realized dollar profit.

The current research is best understood as a structured volatility-spread convergence study. It shows that regime-conditioned spread-shape deviations can be measured, normalized, and tested for mean reversion.

That is a strong research foundation, but it is not the same as a complete executable trading system. A live trading system would still need execution modeling, bid/ask handling, risk limits, margin logic, assignment controls, liquidity filters, and portfolio-level risk management.

From market regimes to option-surface intelligence

For us, the main value of MARFIN is not that it tries to forecast every market move. Its value is that it organizes market conditions into a useful regime framework.

That same framework can be applied to options. First, we build a fair-risk surface. Then, we build a historical market-expected surface. Then, we measure the current market’s deviation from that norm. Finally, we move from raw IV to smile shape and from smile shape to specific spread-level mispricing.

The research results are encouraging. The surface structure is internally consistent, the historical market premium over fair value is measurable, and Spread Shape Mispricing shows a clear tendency to compress and move toward zero after extreme z-score readings.

This does not make MARFIN a signal machine. It makes it a more precise lens for understanding option risk.

The final idea is simple:

Not just: “Is IV high or low?”

But: “Has the shape of this option spread moved away from its MARFIN regime-conditioned norm, and has this type of deviation historically tended to return?”