Additive adjustment preserves differences within an equally shifted segment but changes percentage changes. Positive ratio adjustment preserves within-segment ratios but rescales differences. Neither adjusted level is automatically an actual historical trade price.
Ask what the adjustment preserves
Back-adjustment changes earlier values when joining futures contracts. Its purpose is to impose a continuity convention, not to recover a missing execution history. Additive and multiplicative methods preserve different mathematical relationships. The academic paper Demystifying Time-Series Momentum Strategies explicitly discusses both ratio adjustment and difference adjustment. Its methodological distinction is relevant here without importing its portfolio assumptions or results.
The important question is not whether adjusted data are good or bad in general. It is whether the transformation preserves the property used by a particular calculation. Point differences, percentage changes, absolute thresholds and cross-window comparisons do not all behave identically. A method description should identify the preserved quantity before assigning economic meaning to an adjusted chart.
Work through an additive adjustment
Take hypothetical historical levels 40, 45 and 50 in one consistent quote unit. Suppose the outgoing contract is 50 and the incoming contract is 60 at the chosen splice reference. Under an illustrative latest-level-anchored additive convention, add 10 to the earlier segment. Its displayed history becomes 50, 55 and 60, aligning the old segment with the incoming level.
The first within-segment difference remains five, but its fractional change changes. Originally, 45 divided by 40 minus one is 12.5%. After adjustment, 55 divided by 50 minus one is 10%. The same point movement now has a different denominator. Reporting the adjusted percentage as though it were the original quote's percentage would therefore misdescribe the transformation.
Contrast a positive ratio adjustment
Using the same splice levels, a ratio convention multiplies the earlier segment by 60 divided by 50, or 1.2. The levels become 48, 54 and 60. The first difference is now six, but 54 divided by 48 minus one remains 12.5%. Positive rescaling preserves this within-segment proportional change while altering the number of quote units separating adjacent observations.
The pandas fractional-change documentation also clarifies a display trap: its calculation produces fractional change, which must be multiplied by 100 to express a percentage. A result of 0.125 means 12.5%, not 0.125%. Correctly formatting that number still does not turn an adjusted reference change into an account return.
| Hypothetical representation | First two levels | Difference | Percentage change |
|---|---|---|---|
| Original | 40 to 45 | 5 | 12.5% |
| Add 10 | 50 to 55 | 5 | 10% |
| Multiply by 1.2 | 48 to 54 | 6 | 12.5% |
State the limits of invariance
If every observation inside a comparison window receives the same additive constant, a latest-price-minus-mean difference is unchanged. The added constant cancels. Likewise, applying the same positive multiplier to the entire window preserves the sign of that difference. These are algebraic properties, not evidence that every historical adjustment leaves every directional signal unchanged under all possible construction conventions.
A window crossing a splice may contain segments transformed by different constants or factors. The cancellation argument then no longer applies uniformly. A fixed absolute threshold can also change meaning after rescaling. Before claiming that an adjustment cannot affect a rule, specify the exact window, transformation and threshold; a broad statement about all trend signals is not justified.
Do not mistake adjusted negatives for traded negatives
An additive transformation can create negative historical levels even when the original quotes were positive. Subtracting 60 from hypothetical levels 40 and 42 yields -20 and -18. Their difference remains +2, but the fractional calculation gives -10%. That negative ratio result does not mean the original quote fell. It illustrates why the denominator matters when interpreting transformed data.
A negative adjusted value and a negative actual settlement are separate facts. Establishing the latter requires the original contract-level source, not a back-adjusted chart. Ratio methods also require special care around zero: a zero denominator prevents calculation, while a negative factor can reverse ordering. These cases need explicit handling rather than silent absolute values, clipping or replacement with convenient positive numbers.
Track the version of history
Some back-adjustment conventions revise earlier displayed levels when another roll occurs. A chart downloaded later may therefore show different historical numbers from an earlier download even when the underlying contract settlements are unchanged. That is a consequence of the construction, not necessarily a source correction. Reproducibility requires the adjustment rule, anchor and data version alongside the displayed dates.
This does not imply that every changed historical level creates lookahead in every statistic. A calculation invariant to a common transformation can retain its result. Other calculations may change. The careful approach is to examine the specific transformation and information available at the decision time, preserving original constituent observations wherever possible instead of assuming either universal safety or universal contamination.
Use the right record for the question
An adjusted series can answer a clearly bounded descriptive question about its own transformed path. A statement about a historical executable quote requires the original contract and timestamp. A statement about account performance requires still more: position history, fills, cash flows, costs and a capital convention. Those evidential requirements are not supplied by choosing a smoother adjustment method.
When reading results, look for the series construction before interpreting percentages. Check whether arithmetic crosses a splice, whether the denominator can reach zero and whether historical levels were subsequently changed. A transparent explanation can acknowledge these limits while still using adjusted data for an appropriate purpose. What it must not do is present synthetic continuity as proof of an actual trading path.
Research checkpoints
- Additive and positive ratio adjustments preserve different quantities.
- Invariance within a uniformly adjusted segment does not automatically extend across splices.
- An adjusted negative level does not prove a negative historical trade.
Source record
Related methods
Continuous Futures Series: A Construction, Not a Contract
Learn how contract selection, switching and blending shape a continuous reference series and limit what it can establish.
Negative Futures Prices: Contract Context and Arithmetic Limits
Understand the documented May 2020 WTI event and why zero crossings require explicit handling of ratios, logarithms and contract sensitivity.
Reproducible Method Comparisons Without Performance Claims
Freeze inputs, timing and tie rules to compare three deterministic directional methods with a two-observation lag and a turnover proxy.
Published 16 October 2026. Illustrations and reference diagnostics do not establish strategy profitability or execution quality.