The simplest way to compute Greeks for a calendar spread option is to compute them for each leg separately and add. Front-month delta minus back-month delta gives spread delta. Front-month gamma minus back-month gamma gives spread gamma. This approach is taught in derivatives courses, implemented in most trading systems, and wrong for commodity markets.
Why Leg-by-Leg Aggregation Fails
The arithmetic works if the two legs are independent — if knowing the front-month price tells you nothing about the back-month price. For commodity futures, the assumption fails almost completely. Front-month and back-month crude oil futures have daily return correlations above 0.95. Natural gas correlations drop to 0.70 for contracts separated by a winter delivery window, but they are still far from zero.
When two underlyings are correlated, the Greeks of a position that depends on both are not separable. The cross-gamma — the second derivative of option value with respect to both the front-month and back-month price simultaneously — is non-zero. Ignoring it means your gamma is wrong. And if your gamma is wrong, your delta hedge breaks down as soon as spot moves significantly.
The error is not hypothetical. In March 2023, WTI front-back spreads moved 85 cents per barrel in a single session when US inventory data surprised to the upside. Desks that computed spread Greeks independently missed the cross-gamma contribution, which added approximately 0.07 barrels of unexpected gamma exposure per option contract. On a book with 50,000 contracts, that was 3,500 barrels of unhedged gamma that showed up as unexplained P&L.
The Math: Cross-Gamma and Its Magnitude
For a spread option with payoff max(F1 - F2 - K, 0), where F1 is the near-month forward and F2 is the deferred forward, the spread delta with respect to F1 is straightforward: it equals the probability that the spread expires in the money, weighted by the local vol at that strike. So is the delta with respect to F2.
But the gamma calculation involves three terms: the self-gamma with respect to F1, the self-gamma with respect to F2, and the cross-gamma dV/dF1dF2. The cross-gamma is negative for a long call spread: when F1 rises and F2 rises together (which happens when they are highly correlated), the spread stays near its current level and the option behaves less like a call. The cross-gamma reduces the total gamma you think you have.
Using realistic correlations — 0.93 for nearby WTI calendar spreads — the cross-gamma correction reduces effective gamma by approximately 15% for at-the-money spread options with a 30-day tenor. For longer tenors where correlation drops to 0.80, the correction grows to 22%. These are not small numbers.
How Most Risk Systems Get This Wrong
Standard risk systems — including several widely-used prime brokerage systems that Allasso customers migrated away from — compute spread Greeks using a single-factor model. They treat the spread S = F1 - F2 as the underlying, fit an implied vol to spread option prices, and run Black-76 against that spread. The spread becomes the single underlying, and cross-gamma disappears by construction.
This sounds reasonable until you need to hedge. You cannot trade the spread F1 - F2 directly. You trade F1 and F2 separately. The hedge in F1 and the hedge in F2 must be computed using the correct decomposition of spread option sensitivities into the two underlyings. If you compute that decomposition using the single-factor spread model, you get wrong hedge ratios in each leg. The sum of the leg hedges is approximately correct (the spread hedge), but the individual leg allocations are systematically off.
This matters when F1 and F2 do not move together perfectly — when they decorrelate. That happens during delivery squeezes, storage limit events, and geopolitical supply shocks. Exactly when you are most exposed and most need accurate Greeks.
The Correct Approach: Two-Factor Models
A two-factor model for spread options treats F1 and F2 as correlated log-normal (or normal, in low-price environments) processes and computes option value using numerical methods — typically Monte Carlo or PDE on a two-dimensional grid. The output is a full set of Greeks including cross-gamma.
Margrabe's formula handles the K=0 case (exchange options) analytically. For non-zero strike spread options — which are the practical case in commodity markets — there is no clean closed form, but Bjerksund-Stensland provides a good approximation that captures cross-gamma explicitly.
Allasso implements the Bjerksund-Stensland spread option model with time-varying correlation inputs. Users specify correlation per commodity pair and tenor bucket, and the system computes the full cross-Greek matrix. For spread positions, the platform displays front-leg Greeks, back-leg Greeks, and cross-Greeks separately, so the risk manager can see exactly how much of the position's gamma is correlated vs. idiosyncratic.
Practical Impact on Delta Hedging
Consider a trader long 100 lots of the M1-M2 WTI spread call struck at $0.50 with 30 days to expiry. The spread is currently at $0.60. Under a single-factor model, the spread delta is approximately 0.72, which the trader translates into +72 front-month lots and -72 back-month lots.
Under the two-factor model with 0.93 correlation, the correct hedges are +70 front-month lots and -74 back-month lots. The difference looks small — 2-4 lots — until you realize the discrepancy compounds across a book with hundreds of spread positions across different tenors and commodities.
More importantly, when correlations temporarily break — which happens in natural gas during polar vortex events and in crude during storage limit events — the two-factor model signals a hedge adjustment because the cross-gamma changes. The single-factor model is blind to the correlation change and signals no adjustment. The delta hedge falls further behind on exactly the days when it needs to be most accurate.
Calendar Spread Options vs. Outright Options: What Matters More
For outright options on a single commodity, the error from treating cross-Greeks as zero is exactly zero. There is one underlying. The standard Greeks framework applies cleanly. The problem is exclusive to multi-leg spread options — calendar spreads, crack spreads, crush spreads — and to portfolio-level aggregation where positions in correlated underlyings interact.
Portfolio-level delta aggregation is where even desks that price individual spread options correctly often make errors. If you hold a long M1-M2 WTI spread call and a short M1 WTI put, the combined position's gamma is not simply the sum of the two individual position gammas. The M1 exposure in both positions interacts. A proper portfolio Greek must track the M1 and M2 exposures separately and compute the cross-Greeks at the portfolio level.
Allasso's portfolio Greek engine does exactly this. Every position is decomposed into its underlying factor exposures — by commodity, contract month, and pricing model — and the cross-Greeks are computed across the full matrix of correlated underlyings. The result is a portfolio-level Greek report that survives correlation breakdowns and gives an accurate picture of risk under stress scenarios.
What to Check in Your Current System
To test whether your current system handles spread Greeks correctly, run this check: price an at-the-money M1-M2 spread call and a synthetic replication using a long M1 call and a short M2 call at equivalent strikes and vols. The two prices should differ by the value of cross-gamma and covariance terms. If your system shows them as equal, it is using single-factor spread pricing and your cross-Greeks are zero by assumption.
A second check: perturb M1 by $1 and M2 by $1 simultaneously. The spread option value should barely change. Perturb M1 by $1 and M2 by -$1 (a spread widening). The spread call value should increase sharply. If your risk system shows symmetric response to both perturbations, the correlation structure is not being modeled. Book a session with our team to walk through your specific spread option portfolio — we can demonstrate exactly where the discrepancy appears in your current numbers.