Rolling 12-Month Returns

What is Rolling 12-Month Returns?

A backtest's total return tells you what one investor would have earned by holding from the first day to the last. Almost no one invests on exactly those two dates. Rolling 12-Month Returns answers a more realistic question: if you had bought this strategy at any month in its history and held for one year, what would you have made? It computes the trailing twelve-month return at every point in the test and plots the whole series, turning one lucky or unlucky start date into the entire range of one-year experiences.

Each point is the percentage change in the portfolio's value over the preceding twelve months. As the window rolls forward, you get a continuous record of how the strategy treated one-year holders who entered at different times. The spread of that record is the real story. The best rolling year, the worst rolling year, and how often the trailing year finished negative together describe the strategy far more honestly than a single compounded average can.

In the workflow, rolling returns sit between the headline return numbers and the risk metrics. Total return and CAGR tell you the destination; rolling returns tell you how bumpy the ride was for investors who arrived at different gates. It is the most intuitive consistency check available, because everyone understands "what would I have made in a year" without needing a ratio or a statistical definition first.

Formula

At each month with at least twelve months of history behind it, the portfolio value is divided by its value twelve months earlier and one is subtracted, giving the simple return an investor would have earned over that exact year. Equivalently, it is the compounded product of the twelve intervening monthly returns minus one, written (1 + r₁)(1 + r₂) … (1 + r₁₂) − 1. Because the period is already exactly one year, no annualization is applied: the number is the literal one-year gain or loss. Values are measured net of the transaction costs charged at each rebalance, so the figure reflects what was actually keepable rather than a gross paper return.

Two conventions matter. First, this is a simple (arithmetic) holding-period return, not a logarithmic return, so a sequence of these figures should not be summed to recover total return. Second, it is computed on a calendar-spaced monthly value series, so a twelve-month window always spans a true year rather than a fixed count of trading days. The windows overlap: each one shares eleven of its twelve months with its neighbor, which is harmless for plotting but matters a great deal for interpretation, as the pitfalls below explain.

Why Rolling 12-Month Returns Matters in Backtesting

The decision this informs is expectation-setting. A strategy with a 12 percent CAGR sounds like it returns about 12 percent a year, but rolling returns usually reveal that the typical year was nothing like 12 percent. There were +35 percent years, −20 percent years, and very few that landed near the average. Seeing that spread before investing is what keeps you from abandoning a sound strategy during an ordinary bad year you should have expected all along.

The failure mode of ignoring it is anchoring on the average and being blindsided by the variance. Two strategies with identical CAGRs can have completely different rolling-return distributions: one clustered tightly around the mean, the other swinging from large gains to large losses. The tight one is far easier to hold through a full cycle. Rolling returns also surface the worst twelve-month stretch the strategy ever produced, which is often a more useful planning number than maximum drawdown because it maps directly onto a holding period investors actually use, the one-year mark where many people check their statements and decide whether to stay.

How SledgeKey Implements Rolling 12-Month Returns

Rolling 12-Month Returns appears on the results page as a line plotted across the backtest timeline, starting one year in, since a trailing-year return needs a full year of prior data to exist. Each point is the trailing twelve-month percentage change in the simulated portfolio value, measured after the transaction costs applied at each rebalance. The same calculation is applied to the benchmark, so the strategy's trailing-year line and the benchmark's trailing-year line can be read side by side.

Reading them together is the point. When the strategy's rolling return sits above the benchmark's, the strategy was beating its reference on a one-year basis at that moment; when it dips below, it was lagging, even if its cumulative equity curve was still higher overall. That separates the question "is it ahead in total" from "is it winning right now," which is exactly the distinction a holder feels month to month. Pair the chart with rolling Sharpe to see whether strong trailing-year returns came efficiently or were bought with extra volatility.

Common Pitfalls

The first pitfall is treating overlapping windows as independent data points. Consecutive rolling-year readings share eleven of twelve months, so a single strong or weak month echoes through twelve consecutive points. A cluster of great rolling years can reflect one exceptional quarter that happens to sit inside many overlapping windows, not twelve separate years of success. Count distinct, non-overlapping years before concluding the strategy "usually" returns a given amount.

A second pitfall is confusing a rolling 12-month return with annual return. CAGR is one compounded growth rate for the whole test; a rolling 12-month return is one specific year's simple gain. The average of all rolling 12-month returns is generally higher than the CAGR, because averaging arithmetic returns ignores the compounding drag of volatility. They are different statistics and should never be quoted interchangeably.

A third pitfall is reading the best and worst rolling years as the strategy's hard limits. They are the extremes that happened to occur inside this particular test window, not a ceiling and a floor. A longer or differently timed backtest would almost certainly produce a worse worst-year, because the sample simply had not yet met every kind of market. Treat the worst rolling year as a lower bound the sample reached, never as the actual worst that can happen.

Related Concepts