What Is Backtesting?

What is backtesting?

Backtesting takes a defined strategy (which securities to hold, how to weight them, how often to rebalance) and applies those rules to historical prices and fundamentals as if you had been trading them the whole time. The output is a simulated track record: an equity curve showing what a starting balance would have grown to, period by period, along with a battery of statistics describing the ride you would have taken to get there.

In a screening platform, backtesting is the step that comes after you define a strategy. You build a screen (for example, profitable companies with low debt and improving margins), decide how the portfolio is constructed and how often it resets, and the backtest then walks that recipe forward through history. At each rebalance date it rescreens the universe, rebuilds the portfolio from the names that qualified at that moment, and records the result before moving to the next date.

A well-run backtest tells you two things: how a rule behaved across different market environments, and how much risk you would have carried to earn its returns. What it does not tell you is what will happen next. Backtesting measures the past behavior of a rule, and that measurement is only as trustworthy as the process that produced it.

Why Backtesting Matters in Investing

Backtesting lets you compare strategies on equal footing and reject ideas that sound convincing but fall apart under scrutiny. A rule that beats the market in one dramatic year might trail it over a full cycle. A screen that posts a high average return might have earned all of it in a single window and drifted the rest of the time. Simulating the strategy across years of data, including the bad years, is the only way to separate a repeatable edge from a lucky headline.

The failure mode is treating a backtest as a promise. A simulation that ignores trading costs, includes only companies that survived to today, or uses information that was not public at the time will produce returns no investor could ever have captured. The gap between paper and reality is large and well documented. When McLean and Pontiff studied 97 published market anomalies, they found that returns fell by roughly half after the strategies were published, and a meaningful share of the original edge appears to have been an artifact of the backtest itself rather than a durable pattern.

The lesson is not that backtesting is useless. It is that a backtest is a hypothesis test, and the discipline built into the simulation is what determines whether the result means anything. A clean process gives you a defensible estimate of a strategy's character. A sloppy one gives you a flattering fiction.

How SledgeKey Implements Backtesting

In SledgeKey the backtest is the destination of a workflow that starts with a screen. You define the rules that pick your stocks, set how the portfolio is built (the weighting method, the maximum number of holdings, any cap on a single position), choose how often it rebalances (quarterly by default, with monthly, semi-annual, and annual also available), and pick a benchmark to measure against (SPY by default). Running the backtest then simulates that strategy across your chosen window and returns a simulated equity curve plotted against the benchmark.

The result is not a single number but a full report. You get headline performance (total return and annualized return), risk-adjusted measures (Sharpe and Sortino ratios), a max drawdown reading of the worst peak-to-trough decline, a distribution of monthly outcomes, and a year-by-year breakdown of how the strategy did in each calendar year. Every trade in the simulation is charged a transaction cost so the returns reflect friction rather than an idealized frictionless world.

The part that matters most sits underneath all of this: the backtest is built on point-in-time data. When it evaluates a company on a historical date, it uses the fundamental figures that were actually public on that date, identified by when the filing became available rather than the period the numbers describe. This is the discipline that keeps a backtest from quietly using information the strategy could not have known at the time, and it is what separates an honest simulation from a flattering one.

Common Pitfalls

The first and largest pitfall is treating the backtest as a prediction. A simulated 15% annual return is a statement about how a rule would have behaved in a specific past, not a rate you are entitled to going forward. Markets change regimes, popular strategies get crowded, and the exact conditions that produced a backtest rarely repeat.

The second pitfall is overfitting: tuning a strategy until the historical curve looks beautiful. If you try enough combinations of thresholds and filters, some of them will fit the past by accident. The more parameters you adjusted to make the equity curve smooth, the less the result tells you about the future, because you have partly memorized the history rather than discovered a pattern in it.

The third pitfall is trusting a backtest without asking how the data was assembled. Survivorship bias (testing only on companies that still exist), look-ahead bias (using data before it was public), and unmodeled trading costs each inflate results in ways that are invisible unless you go looking. A backtest is only as good as the answers to those questions.

Related Concepts