Taking a crypto vol surface to your model committee

Fifteen questions a risk or model validation team should ask before an external volatility surface goes into marks, limits or client pricing, and what a good answer looks like.

Empirical observations from VolaLabs surface data. Not investment advice or a recommendation to transact.

An implied vol surface is a model output, not market data. Between the order book and the number in your system sit choices about which quotes count, how a smile is parameterised, how gaps are filled and what happens when a venue goes quiet. A committee does not need to re-derive those choices. It needs to know they were made deliberately, that they are written down, and that they will be visible when they matter.

Below is the list, with how VolaLabs answers each point. Scroll through and watch it fill up.

01Data and lineage

• Which venues and instruments feed the surface?You need to know whose prices you are marking against, and what is out of scope.Deribit BTC and ETH options, every listed expiry from same-day to two years.
• How are quotes filtered before fitting?Crossed markets, stale quotes and wide wings distort a fit long before anyone notices in the output.Filtering rules are documented in the methodology pack.
• Is every snapshot timestamped and reproducible?An auditor will ask what the surface looked like at a given time. "Roughly then" is not an answer.Every snapshot carries its timestamp. History and live data share one schema, so any past surface can be pulled back as it was.

02Model and fit

• Which benchmark model is used, and why?An interpretable parameterisation makes it possible to explain a move, not just observe it.SVI per expiry as standard, with parameter vectors published. Alternative models on request.
• How are quotes weighted in the fit?Fitting to mids treats a wide wing like a tight at-the-money quote. Weighting by bid-ask keeps the surface where you can trade.Fits are weighted by liquidity, so tight markets count for more.
• What fit-quality metrics are published, and how often?Look for residuals in vol points and the share of fitted points inside the bid-ask, per expiry.Calibration residuals and fit metrics are available for each snapshot.
• How stable are parameters between updates?If a refit moves parameters more than the market did, your greeks and P&L move with it.

03No-arbitrage

• Is butterfly arbitrage ruled out per expiry?A negative density means some strike spread pays out for free. Prices downstream inherit it.Enforced in every fit.
• Is calendar arbitrage ruled out across expiries?Short-dated stress, especially around quarterly expiries, is where this breaks first.Enforced across all listed expiries.
• Are violations flagged when they occur in the raw data?You want to know when the market itself was inconsistent and how the surface handled it.Arbitrage flags are available for every snapshot.

04Interpolation and extrapolation

• How are points between listed expiries filled?Interpolating in vol creates kinks when expiries roll off. Interpolating in total variance does not.Constant maturities are interpolated in total variance.

05Operations

• What is the update cadence, and the latency from quote to surface?Intraday risk and margining need surfaces that keep up with the market.Every five minutes, sub-minute on request.
• Is uptime and incident history published?Operational risk is assessed from a record of real incidents and conveyed transparently.Uptime and incident history are documented.

06Governance

• Is the methodology written down and versioned?The committee approves a specific version. Changes should come with notice and a change log.Methodology pack available, covering construction, constraints and robustness reporting.
• Can you benchmark the surface independently?Compare against exchange mark vols and your own fits on the same timestamps before going live.History and live data in one schema, so you can line our surface up against exchange marks on matching timestamps.