Futures & Market Research Glossary

Rate of Change (ROC) measures how much a market value has changed relative to its value a fixed number of periods earlier. In technical market analysis, a common percentage convention is ROC = 100 × (X_t / X_{t-n} − 1), where n is the lookback. For an ordinary positive-valued price series, a positive ROC means the current value is above the lagged value and a negative ROC means it is below it. Despite the name, ROC is not a derivative or a change per unit of clock time: the measurement horizon is defined by the chosen lag.

The lookback defines the question

A 5-day ROC and a 20-day ROC can point in opposite directions because they compare the same current observation with different starting points. If a price is 110 today, 100 twenty sessions ago and 120 five sessions ago, the 20-session ROC is +10%, while the 5-session ROC is about −8.3%. Neither reading overrides the other; they describe different horizons. For a mathematically defined observation with a nonzero lagged value, ROC = 0 means only that the current value equals the value n periods earlier. It does not by itself establish a trend reversal or a directional trading signal.

For a price series, percentage ROC under this convention is mathematically the simple price return over the same lookback. Implementation still matters. One function may report the fractional change multiplied by 100, while another may return the fractional change itself or a price ratio under a related ROC label. The input series matters as well: settlement versus another price field, a price index versus a total-return index, and the construction of a continuous futures series can all change the result. In futures research, an unadjusted contract splice can introduce a non-traded jump at a roll, while adjustment methods alter how the historical series is represented. If the lagged value is zero, the percentage formula is undefined because it requires division by zero. Software must therefore decide how to handle that observation, such as by returning a missing value, suppressing the result or substituting an implementation-specific value; such a substitute is not the mathematical ROC. Values close to zero can also produce extreme percentage changes, while negative denominators make the usual directional interpretation of percentage change less intuitive.

Overlapping windows are not independent observations

A less obvious issue appears when a long-horizon ROC is recalculated more frequently than its lookback. Daily observations of a 12-month ROC use heavily overlapping measurement windows; adjacent readings reuse almost all of the same underlying return history. Serial dependence can therefore arise partly from the construction rather than from independent new information. This matters in backtests and statistical inference because a dense sequence of overlapping ROC observations does not provide the same amount of independent evidence as an equal number of non-overlapping observations. ROC is most useful when its horizon, scaling, input series and sampling convention are explicit, and when its magnitude is interpreted as measured past change rather than as a standalone forecast of what price must do next.

Real yield is the yield on a fixed-income investment expressed after accounting for inflation, so that the return is measured in purchasing-power rather than purely nominal terms. Conceptually, it is the yield that remains after removing the effect of inflation over the relevant horizon. In forward-looking market analysis, the relevant comparison is expected future inflation rather than the latest reported inflation rate. A common approximation is real yield ≈ nominal yield − expected inflation; more exactly, nominal rates, real rates and inflation are related multiplicatively rather than by simple subtraction.

Observed real yields are market prices, not inflation forecasts

In U.S. markets, real yields are commonly observed through Treasury Inflation-Protected Securities (TIPS). TIPS principal is adjusted with the Consumer Price Index, while the coupon rate is fixed and applied to the inflation-adjusted principal. Their market yields are therefore commonly interpreted as yields on inflation-linked Treasury cash flows in real terms. Real yields can be positive or negative. A negative real yield means that the market price implies a return below zero in purchasing-power terms before taxes, even though the security can still make positive nominal coupon payments and its principal can rise with inflation.

The difference between a nominal Treasury yield and a comparable TIPS real yield is commonly called breakeven inflation or inflation compensation. It is not a pure measure of expected inflation. Inflation-risk premiums, differences in liquidity between nominal Treasuries and TIPS, and other market effects can create a wedge between inflation compensation and investors' underlying inflation expectations. A change in real yields likewise does not identify a single economic cause: required real rates, expectations for monetary policy, term premiums and market conditions can all contribute.

The quoted number may not belong to any single bond

Researchers should also distinguish the yield of a particular TIPS from an official constant-maturity real-yield series. The U.S. Treasury's Real Constant Maturity Treasury rates are interpolated par real yields read from a fitted TIPS par real yield curve constructed from indicative secondary-market quotations. A quoted 10-year real constant-maturity rate can therefore exist even when no outstanding TIPS has exactly 10 years remaining to maturity. It is a constructed curve point, not necessarily the observed yield of a security at that precise maturity. This distinction matters when comparing datasets, backtesting rate relationships or interpreting small historical changes in “the” 10-year real yield.

Real yield is most useful as a measure of the market price of inflation-protected real interest-rate exposure. It is not, by itself, a forecast of economic growth, monetary policy or asset prices, and an investor's realized purchasing-power return can differ because of purchase and sale prices, holding period, taxes and differences between the inflation index used by the security and the investor's own cost of living.

Realized volatility is a backward-looking measure of how much an asset’s returns varied over a completed observation interval. It is calculated from a past return series rather than inferred as the volatility input consistent with current option prices. That separates it from implied volatility: implied volatility is extracted from option prices for a future horizon, whereas realized volatility summarizes variability that has already occurred.

The number depends on the measurement rule

A common close-to-close construction starts with periodic returns, often logarithmic returns such as r = ln(Pt/Pt−1), calculates their standard deviation over a chosen lookback window, and scales the result to an annualized figure using the square root of a stated number of periods per year. But “realized volatility” does not identify one universal estimator. Researchers can choose simple or log returns, demean returns or set the mean to zero, use different lookback lengths, sample at different frequencies, include or exclude overnight moves, and apply different annualization conventions. Two defensible calculations can therefore produce different values from the same market history.

This distinction becomes especially important with intraday data. A 20-day close-to-close measure answers a different question from a one-day measure built from five-minute returns. In high-frequency work, a standard realized-variance estimator is the sum of squared intraday returns; taking its square root gives a realized-volatility measure for that interval, which may then be annualized for comparison. The naming and scaling conventions used by data providers and researchers are not perfectly uniform.

Finer sampling can add noise as well as information

More frequent observations can capture price movement that daily closes miss, but extremely fine sampling can also admit market-microstructure effects such as bid–ask bounce and other transaction-price noise. The useful sampling interval therefore depends partly on the market’s liquidity and trading structure. Session boundaries matter as well: an intraday calculation confined to regular-session observations excludes price changes outside that sampled session unless they are incorporated separately.

For traders and researchers, a realized-volatility percentage is therefore incomplete without its observation window and estimator conventions. Meaningful comparison requires the return definition, sampling frequency, session coverage, lookback horizon, and annualization basis to be aligned. This is particularly important when realized volatility is compared with implied volatility or used in work on the volatility risk premium. Realized volatility measures past variability; it does not by itself reveal price direction, explain why volatility occurred, or specify the volatility that will be realized next.

Relative value is the assessment of whether one financial instrument, contract or market appears cheap or expensive compared with another economically related instrument after accounting for the exposures that make the comparison meaningful. A relative-value position commonly expresses that view through opposing exposures—long the relatively cheap side and short the relatively rich side—so that the research question concerns their performance difference more than the outright direction of the broader market.

The difficult part is deciding what “relative” should mean. Two prices cannot necessarily be compared one-for-one simply because their markets are related. Contract sizes, currencies, duration, volatility, beta, maturity, quality, location or other economic sensitivities may differ. Researchers therefore often compare a spread, ratio or normalized residual and choose a hedge ratio intended to reduce the dominant common exposure. Equal notionals are not automatically equal risks. In fixed income, for example, positions with the same notional value can retain substantial directional interest-rate exposure if their durations differ.

A simple valuation example

Suppose a researcher adjusts two related instruments for the exposures judged economically relevant and estimates that Instrument A should have a comparable value of 103 given the current value of Instrument B. If A actually trades at 100, the resulting residual is −3, so A appears relatively cheap by 3 units against that benchmark. A relative-value position might therefore be long A and short an appropriately hedged amount of B. The thesis is not that 100 is intrinsically low or that B is intrinsically expensive; it is that A is cheap relative to the specified relationship. If the residual later narrows from −3 to −1, the relative valuation has moved toward the benchmark even if both instruments rose or both fell in absolute terms.

The hedge ratio defines the comparison

This construction choice can determine whether an apparent valuation gap is informative or largely an artifact of mismatched exposures. A practical fixed-income example is duration matching: Treasury and interest-rate-derivative exposures may be converted into common duration-based units before their opposing legs are compared. The broader lesson is transferable—the economically relevant unit of comparison may be risk exposure rather than raw price or notional amount.

Relative does not mean convergent

A relative-value observation is not proof that the cheaper instrument will rise, the richer one will fall, or the gap will close. The relationship itself can change because of fundamentals, funding conditions, liquidity, contract mechanics or a structural break. Correlation may help identify common movement, and cointegration may provide evidence of a more persistent statistical relationship, but neither is synonymous with relative value. The valuation thesis still requires an economic reason for the comparison and a defensible method for estimating the deviation.

This distinction becomes especially important when small valuation differences are financed with substantial leverage. Some relative-value strategies depend on short-term funding or derivatives margin, so an adverse spread move or tighter financing conditions can force positions to be reduced before any expected convergence occurs. Relative value therefore describes a comparative valuation framework, not a low-risk or market-neutral outcome by definition.

Reportable Positions are the long and short open-interest amounts in the CFTC’s Commitments of Traders (COT) reports attributable to traders captured by the Commission’s large-trader reporting system. The term describes a reporting-status aggregate, not a trader category. Reportable open interest is divided according to the classification framework of the applicable COT report family. Nonreportable Positions are then derived separately for each side: Nonreportable Long equals total open interest minus Total Reportable Long, while Nonreportable Short equals total open interest minus Total Reportable Short. The two residuals therefore need not be equal.

A reporting level brings the broader position into view

The crucial detail is what happens when a reporting level is reached. If, at the daily market close, a trader has a position at or above the applicable CFTC reporting level in any single futures or option expiration month, the reporting firm reports that trader’s entire position across all futures and options expiration months in that commodity, regardless of size. The threshold is therefore an inclusion trigger, not a cutoff under which every smaller position remains unreported.

This creates an important interpretation effect. A relatively small change that pushes a trader across the reporting level can cause pre-existing positions in other expirations to enter the reportable aggregate at the same time. Reportable Positions can therefore change by more than the position change that triggered reportability. Such a movement should not automatically be read as an equivalent amount of newly created directional exposure. The reporting level determines when a trader enters the large-trader framework; Reportable Positions measure the positions captured once that reporting status applies. COT publication adds a separate market-level condition: the reports cover markets in which 20 or more traders hold positions at or above the applicable reporting levels.

Reportable does not mean one trader type

The composition of Reportable Positions depends on the COT report family. In the Legacy COT report, reportable open interest is classified as Commercial or Non-Commercial. The Disaggregated COT report uses Producer/Merchant/Processor/User, Swap Dealer, Managed Money and Other Reportables. Traders in Financial Futures uses Dealer/Intermediary, Asset Manager/Institutional, Leveraged Funds and Other Reportables. The Supplemental COT report uses Non-Commercial, Commercial and Index Trader classifications. These frameworks are not interchangeable simply because every category belongs to the broader reportable population, and classification does not establish the motive behind every position held by a trader.

Report format also changes what is measured. A Futures Only COT report contains futures positions, whereas a Futures and Options Combined COT report incorporates options on futures on a delta-adjusted futures-equivalent basis. Two datasets can therefore describe the same market and observation date yet show different reportable totals because their coverage differs. For research, changes in Reportable Positions should be compared within a consistent report family and format. Reportable status identifies which positions are inside the large-trader reporting aggregate; by itself, it does not establish whether those positions represent speculation, hedging, bullishness or bearishness.

Return is the gain or loss on an asset, position or portfolio over a specified interval, expressed relative to a defined starting value. A one-period simple holding-period return can be written as R = (V1 - V0 + C) / V0, where V0 is the beginning value, V1 is the ending value and C represents cash income or distributions generated by the investment during the period. The denominator and the treatment of cash flows are part of the measurement, not bookkeeping details.

Price return and total return are different measurements

A price return captures only the change in quoted price or market value. A total return also incorporates economically relevant income or distributions, such as dividends or interest. For multi-period total-return series, the treatment and reinvestment of those distributions must be defined consistently. Settlement versus closing price, local versus base currency, adjusted versus unadjusted data, fees and financing can also change the result, while investor contributions and withdrawals must be separated appropriately from investment performance.

For example, suppose an asset begins at 100, ends the period at 105 and pays a cash distribution of 2 during the interval. Its price return is 5%, because the price increased by 5 relative to the initial value of 100. Its simple total holding-period return is 7% before costs and taxes: (105 - 100 + 2) / 100. Both percentages are valid, but they measure different things. Reporting only that “the return was 5%” would therefore be incomplete if the intended measure included the distribution.

For a consistently defined wealth or total-return series, multi-period simple returns compound rather than add: cumulative return is the product of the individual growth factors, (1 + R), minus one. The common log-price return, ln(P1) - ln(P0), is additive across time and convenient in statistical work, but it requires positive price observations. That restriction matters in derivatives research because futures prices can become non-positive. The May 2020 WTI crude oil futures contract, for example, settled below zero, making a conventional log-price return undefined whenever a required price observation is zero or negative.

A market move is not always the investor’s return on capital

Futures make the denominator especially important. Futures margin is a performance bond rather than the purchase price of the contract, and positions are marked to market as gains and losses are credited or debited. If a futures quotation moves from 100 to 102, the quoted price has risen 2%, but the trader has not necessarily earned 2% on the cash committed to the strategy. Contract profit or loss depends on the price change and the contract multiplier; percentage return then depends on the capital base chosen for measurement, including how margin, collateral, financing and cash reserves are treated.

For research, “return” is reproducible only when the horizon, price field, cash-flow treatment, compounding convention and denominator are specified. Two analysts can begin with the same market history and obtain different return series without either making an arithmetic error; the difference may come from construction choices. The practical question is therefore not only “what was the return?” but “return on what, measured how, over which interval?”

Roll yield is the component of a futures return associated with how the contract price evolves relative to the underlying or reference market as the contract approaches expiry, within an exposure that is maintained by moving into later maturities. It is often discussed as the return effect of “rolling” futures, but the roll transaction itself is not what creates the yield. Depending on the analytical methodology, curve slope may be used as an ex-ante indication of potential roll yield, while realized roll yield is measured from the return actually accumulated through the futures position relative to an appropriate reference.

Curve shape matters, but conditionally

For a long position, contango generally creates a negative roll-yield tendency if the term structure and underlying market otherwise remain broadly unchanged: a deferred contract trading above the nearby or reference price has room to decline as it ages toward expiry. Backwardation generally creates the opposite tendency because a discounted deferred contract can rise relative to the nearby market as maturity approaches. The sign is reversed for an otherwise comparable short exposure. These relationships are not guarantees. A changing curve, a different contract tenor or a different roll schedule can materially alter the realized result.

The roll transaction itself is not the yield

The familiar description that a long investor in contango must “sell low and buy high” is useful shorthand for the curve structure but misleading as an account of profit and loss. Closing one futures contract at 100 and opening a different maturity at 105 does not create an immediate loss of 5 simply because the quoted prices differ: the contracts are separate instruments, and futures gains and losses arise from subsequent changes in the prices of the positions held. The economically relevant roll effect accumulates as those contracts move through the term structure.

This distinction also separates Roll Yield from Cost of Carry and Convenience Yield. Carry costs, income and non-cash benefits help explain why futures of different maturities trade where they do; roll yield describes a return attributed to holding and renewing futures exposure. It is therefore methodology-dependent. Two strategies referencing the same underlying market can realize different roll yield because they hold different maturities, roll on different dates or use different rules for constructing the return series.

Rolling correlation is a correlation coefficient recalculated repeatedly over a moving window of observations, producing a time series of how the measured relationship between two variables changes. The rolling procedure can be applied to different correlation estimators; the discussion here uses Pearson correlation, which measures linear association. In market research it is often applied to synchronized returns rather than raw price levels when the objective is to study short-horizon co-movement. The result is descriptive: it measures the relationship present inside each window, not a causal link or a forecast of what the relationship will be next.

The window is part of the statistic

A rolling correlation cannot be interpreted without its construction choices. A 20-observation window responds quickly to new data but is statistically noisier; a 250-observation window is usually smoother but can hide shorter-lived changes. Sampling frequency matters as well: daily, weekly and intraday returns can produce materially different correlations because they capture different horizons and microstructure effects. Data alignment, missing-value treatment and the choice between arithmetic and logarithmic returns can also alter the observations entering each estimate.

The moving-window design creates an important practical trap. If a 60-day correlation is advanced one day at a time, two adjacent estimates share 59 of their 60 observations. The resulting correlation series can therefore look smooth and persistent partly because the windows overlap so heavily. A gradual-looking change in the statistic is not, by itself, evidence that the underlying economic relationship is changing gradually.

What the sign and magnitude mean

For Pearson correlation, the sign gives the direction of the measured linear relationship and the absolute value indicates its strength. A positive value means the two series have tended to move in the same direction within the window; a negative value means they have tended to move in opposite directions. A value near zero indicates little consistent linear co-movement. Values closer to +1 indicate stronger positive linear association, while values closer to -1 indicate stronger negative linear association; +1 and -1 represent perfect positive and perfect negative linear relationships in the sample. Near-zero correlation does not prove that the variables are unrelated, because nonlinear relationships can still exist.

What a changing correlation can and cannot say

Changes in rolling correlation can reflect genuine shifts in market behavior, but they can also be driven by extreme observations, volatility changes, asynchronous trading, contract-roll effects or changes in sample composition. Missing data deserve particular care in multi-asset work: pairwise-complete calculations can cause different asset pairs to use different subsets of dates, complicating comparisons and potentially producing a covariance or correlation structure that is not positive semidefinite. Rolling correlation is therefore most useful as a regime-sensitive measurement tool when its estimator, input transformation, window length, observation set and economic hypothesis are specified in advance. It should not be confused with cointegration, which addresses a different question about longer-run relationships between time series.