Reading Skew and Term Structure: A Trader's Field Guide

Option prices contain the market's view of how uncertain the future is, which direction it fears, and when it expects trouble. SOAP2's VSkew (across strikes, one expiry) and HSkew (across expiries) tabs measure these. This guide shows how to read them, how to tell what changed since the previous run, and how to turn them into ranges and scenarios. It needs no mathematics and no knowledge of the Greeks.

For the definitions behind RR25 and BF25, the user-level page is OA Risk Reversal & Skew. This page is the practical companion, in the same family as Reading the Hedging Picture (GEX / VEX). A section for an AI assistant (§8) uses the real field names of the get_skew_read call.

Read this first. Skew describes how options are priced, which reflects demand for protection and upside as well as expectations. It is not a forecast, and a "steep" or "flat" skew has no fixed threshold. SOAP2 reports facts and sign words; judging whether a change is meaningful is yours. Nothing here is a trade signal.

1. The four dials

Every option has an implied volatility (IV): the uncertainty its price implies, quoted as an annual percentage. Different strikes and expiries carry different IVs. Four simple readings summarise the pattern:

On equity indices the normal picture is: puts cost more than calls (negative RR: investors pay for protection), wings above the middle (positive BF), and far expiries above near ones (contango). Trading skew means watching for departures from "normal" and for changes in how far it is from normal.

2. The vocabulary

Names in code are those the AI bridge returns. IVs are decimals (0.152 = 15.2%); changes are in vol points when multiplied by 100 (0.003 = 0.3 vol points).

3. The five-step reading routine

Step 1: Is it comparable and fresh?

• Which expiry? Skew differs by expiry. VSkew is stored per expiry; HSkew spans several.
• Age and session. Read the run's age and whether it crossed a session boundary. After a boundary, time to expiry shortened and spot moved on.
• Quality. Check the data-quality note (feed type, number of inputs received) and, for VSkew, delta_used, not_bracketed and the wings' headroom. A wing near the edge of the ladder is the least robust number on the page.

Step 2: Level (ATM IV)

Is uncertainty high or low for this market? Compare with the history columns and with your own experience of the symbol. Convert it into a plain range with the expected move (§5). A level on its own says nothing about direction.

Step 3: Tilt (RR) and wings (BF)

1. Sign and size of RR: how much extra is the market paying for downside over upside?
2. Divide by ATM IV (rr_to_atm_iv) before comparing with other days: the same RR means more when ATM IV is low.
3. BF: are tails priced richly? A rising BF with a stable ATM IV means more is being paid for the extremes specifically.

Step 4: Term structure

• Slope sign: contango (calm) or backwardation (stress, or an event in the front).
• Look at the shape, not only the slope. A single elevated point (an event date) shows as a bump. A same-day (0DTE) point at the front makes the slope swing with the time of day, so read the curve.

Step 5: Context, then invalidation

• Combine with GEX: richer put skew means more customers holding downside protection, which under the usual convention means dealers short gamma on the put side. Compare with the GEX regime.
• Check the calendar: a scheduled event explains a front-end bump.
• Write down what would change your mind: skew normalising, the front dropping back into contango, or ATM IV collapsing after the event.

4. Reading change: previous run and history columns

There are two kinds of comparison, and they must not be mixed:

Separate the three kinds of move

1. The level moved (ATM IV): everything repriced together.
2. The shape moved: RR or BF changed relative to ATM. Use the ratios to ATM to tell.
3. Time passed: shorter time to expiry changes the numbers even with no market move. Over a session boundary this is always part of the change.

Patterns and what they may suggest

Like for like. For HSkew, the slope change is taken on the expiries both runs share (common_term); what was left out is listed. RR and BF changes are left empty when the two runs solved at different deltas. Always read changes_in_evidence and columns (including its advice about which history offsets to keep ticked so the next comparison is like for like).

5. From the picture to ranges and scenarios

Skew does not give support and resistance. It gives ranges and how the market weights each side.

The expected-move range

SOAP2 defines one standard deviation to expiry as sd = spot × ATM IV ÷ 16 × √(trading days left): the one-day move (the Risk Graph's 1-sigma line) scaled by the square root of the sessions remaining. Example: spot 5,000, ATM IV 16%, 21 trading days: sd = 5,000 × 0.16 ÷ 16 × √21 ≈ 229 points. The market's pricing implies roughly a two-in-three chance of finishing within ±1 sd (risk-neutral pricing, not a forecast).

• Use ±1 sd as a scenario boundary: moves beyond it are what the market considers unusual for that horizon.
• For a shorter horizon, scale by the square root of the sessions. One session is roughly a fifth of the 21-session figure (√(1/21) ≈ 0.22).
• This is the same figure as the expected_move block in get_market_read, and the "sd" distances of the wings use it too (see Turning volatility into ranges). The one exception is get_trade_probs, whose sigma is that of the distribution its odds are computed from (ATM IV × √(days ÷ 360)), so it differs a little.

The 25-delta strikes

In VSkew, wings gives the strikes where the 25-delta call and put sit, with their distance from spot in points, percent and standard deviations. They mark, roughly, the "one-in-four" ends of the market's pricing. Use them to see how asymmetric the market is: if the put strike is 1.0 sd away and the call strike 0.8 sd, the market treats the downside as farther and costlier to reach.

Weighing the two sides

• Steeper put skew: the market is paying more for downside. Combined with a GEX accelerator regime below the Flip, it supports a "downside is where the fragility is" reading. It does not say the market will fall.
• Flatter skew or call skew: upside demand is relatively strong (call skew can appear in squeezes or in single stocks). Check what is driving it.
• High BF: both tails are priced richly. If you hold short premium, you are paid more, because the market fears more.

6. Five situations

A. Calm contango, mild put skew

The normal state. Ranges implied by ATM IV are narrow; shocks are priced modestly. The question is what would break it: an event, or a sudden steepening of RR.

B. Steepening put skew into an event

RR becomes more negative and the front end lifts (a bump on the curve). Participants are buying protection ahead of the date. Expect IV to fall after the event unless it surprises; the drop is the "volatility crush", whose hedging effect is the Vanna story in the theory primer.

C. Backwardation

Front IV above back IV: the market is paying for uncertainty now. Typically a stress state. Read the term curve to see whether it is a single event date or the whole front end.

D. After an event (the crush)

ATM IV, BF and RR all drop. Compare with the history columns: are you back to the pre-event level, or merely lower than the peak? Remember that comparing Latest with Previous across the event includes a session boundary or the event itself.

E. 0DTE day

The front point is a same-day expiry and its IV varies with the clock. Slope can swing during the session; read the curve, and compare like times only.

7. A worked example (illustrative numbers)

Invented numbers to show the method. Not a forecast, not a real session.

1. Comparable? Same session, both solved at 25 delta, same expiry set: a clean comparison.
2. Level. ATM IV +0.4 vol points while spot fell 12 points: a modest rise, typical for a down move.
3. Tilt. RR25 moved 0.7 vol points more negative, and the ratio to ATM moved from −0.21 to −0.24: not just the level; the shape tilted toward put skew. Verdict by SOAP2: more_put_skew.
4. Wings. BF barely changed: no sign of more tail premium.
5. Term. Slope down 0.3: slightly flatter, still contango.
6. Range. sd ≈ 5,000 × 0.16 ÷ 16 × √21 ≈ 229 points: the market treats 4,771–5,229 as the ±1 sd band for the month.
7. Reading (derived). "Demand for downside protection edged up as spot slipped; the overall level and tail premium barely changed and the term structure is still normal. This is a mild risk-off tilt, not stress. Invalidation: RR reverting, or the slope turning negative (which would mean stress)."

8. For the AI assistant: how to answer from get_skew_read

The call is get_skew_read with optional symbol (default: the symbol selected in SOAP2) and expiry (yyyy-MM-dd, a VSkew tab). The response has two blocks: hskew (term structure) and vskew (one entry per stored expiry, nearest few by default). It reads stored results built by Options Analysis windows in HSkew and VSkew mode; if nothing is stored, the block's hint says which OA mode to open. Premium (or trial) and the AI bridge are required; a denied call answers NOT_LICENSED.

The answering routine

1. Primer, if the user needs one (one or two sentences): RR25 is the call IV minus the put IV at 25 delta, normally negative on indices (more negative means more put skew); BF25 is the wings' premium over ATM; the term slope is long-dated minus short-dated ATM IV (negative means backwardation).
2. Freshness and comparability: computed_at/age, crossed_session, the quality note, delta_used, not_bracketed and wings headroom.
3. Level now: atm_iv, risk_reversal with skew_side, butterfly, the ratios rr_to_atm_iv and bf_to_atm_iv, and for HSkew term_shape and the term_curve.
4. Change since Previous, only with elapsed_s and same_session, and every changes_in_evidence item. Separate a move in skew from a move in the vol level (use the RR/ATM ratio) and from the passage of time (DTE). Quote the sign words but judge materiality yourself.
5. History comparison (vs_history): compare Current with each history column of one run, never a history column between two runs. State the reference time (reference.as_of) and as_of_session/bar_lag_min when it rests on the previous close.
6. Relay columns.advice when present (which History offsets to keep or re-tick so the next comparison is like for like).
7. Tie to other reads: ATM IV term, VIX term and expected move in get_market_read; GEX (get_cboe_gex); sticky skew (get_sticky_skew); and outside information such as the event calendar.
8. Derived reading, labelled as derived, with an invalidation; no trade recommendation.

Rules that must not be broken

• IVs are decimals: 0.003 is 0.3 vol points. Convert before speaking.
• An absent or null field has no value; never read it as 0.
• RR and BF at different deltas are not comparable; their change is left null for that reason.
• SOAP2's move words have no materiality threshold. Do not call a change "significant" because the word is present.
• History columns come from sparse 30-minute bars: a few tenths of a vol point in them is not necessarily signal.
• Do not infer direction from skew alone. It is a price of protection, not a forecast.

A response template

Data: SPX, expiry 2026-11-20, VSkew stored at 15:40 UTC (3 min ago), same session; 25-delta solved on both runs.
Level now: ATM IV 16.0%; RR25 −3.9 vol points (put skew), ratio −0.24; BF25 +0.8.
Change vs previous (60 min): ATM IV +0.4, RR25 −0.7 (more put skew), BF25 +0.1; spot −12. The shape tilted, not just the level.
Versus recent days: compared with yesterday's same time, RR25 is 0.5 more negative.
Term: contango, slope +0.8 (common expiries), slightly flatter than before.
Range: ±1 sd to expiry ≈ 229 points (4,771 to 5,229).
Reading (derived): mild risk-off tilt with a normal term structure. Invalidation: RR reverting or the slope turning negative.
Caveats: skew is a price, not a forecast; no materiality threshold; history columns come from sparse bars.

9. What you cannot conclude

• Direction. Put skew means protection is expensive, not that the market will fall.
• Significance. There are no calibrated thresholds for "steep", "flat" or "large change".
• Comparability across deltas, expiries or sessions unless the response says it is like for like.
• Precision at the edge of the strike ladder. A wing with little headroom is the least robust figure.
• Exact levels. The expected-move band is a pricing statement (risk-neutral), not a probability forecast.
• Positioning. Skew reflects demand and supply of options, but not who is long or short.

10. One-page checklist

1. Right expiry, fresh, same session, same delta in both runs? Any headroom warning?
2. Level: ATM IV, converted into an expected-move range.
3. Tilt: RR sign and size; use the ratio to ATM to compare.
4. Wings: BF, and whether the 25-delta strikes are asymmetric.
5. Term: slope sign and the shape of the curve; any event bump; 0DTE at the front?
6. Change: level, shape or time? Read changes_in_evidence; never compare a history column across runs.
7. Cross-check with GEX, VIX term and the calendar. Write the invalidation.

Go back to the main help page: SOAP2 Help Guide

⬅ Back