How to Read a Backtest Without Being Fooled

To read a backtest without being fooled, ignore the headline return at first and check five things: the benchmark over the same period, the worst fall and how long it took to recover, whether costs were included, how the rules and assets were chosen, and whether the result depends on one asset or one stretch of history. A backtest that survives all five questions is worth a closer look. One that hides any of them is a marketing chart.

Below is a practical checklist, with real backtest figures from AlgolaneBots's strategies used as worked examples. All of them are historical backtests: computer tests on past prices, not money anyone earned, and future results will differ.

Step 1: Find the period and the data source

Start with the boring details. What dates does the test cover, and what data was used?

• Length. A test covering five calm years tells you far less than one covering twenty years and several crashes.
• Start and end dates. Starting a test right after a crash makes almost anything look good. Ending it at a market peak does the same.
• Real or simulated history. Some funds did not exist for the whole period. If earlier years were filled in with a simulated version, that should be labelled clearly.

AlgolaneBots's published stress tests, for instance, mark some crash periods as a "simulated three-fund version" because the actual leveraged funds had not launched yet. That label tells you to give those years less weight than the years with real fund prices.

Step 2: Put the benchmark next to every number

A return on its own is meaningless. The first question for any figure is "compared with what?"

For a US stock strategy, sensible benchmarks are the S&P 500 and the Nasdaq 100. For a gold strategy, it is simply holding gold. The benchmark must cover exactly the same dates. A strategy tested from 2011 should not be compared with an index measured from 2000.

Our article on picking a fair benchmark explains why the choice matters.

Step 3: Read the worst fall before the return

Max drawdown, the largest drop from a peak to a low, is the number that decides whether a real person would have stayed with the strategy. Read it before you read the return, and read it in dollars as well as percent.

Then look for the recovery time: how long it took to get back to the previous high. This is often missing from marketing material, and it is often the most sobering number.

Worked example: the same strategy at different risk levels

Gold Breakout lets the user set risk per trade at 1%, 2%, 3% or 5%. Its historical backtest from June 2004 to November 2026 shows how the trade-off between return and fall plays out. The benchmark is gold bought and held over the same period.

Notice three things. First, the 1% setting returned less than simply holding gold, but with about half the worst fall. Second, higher risk settings raised both the return and the fall; nothing came for free. Third, a 74% fall means losing three quarters of the account on paper. Very few people hold through that.

At 2% risk, the deepest fall ran from May 2018 to October 2019 and took 41 months after the bottom to recover. The strategy's caveat: doubling trading costs cuts its yearly result roughly in half. This is a historical simulation; nobody earned these numbers. For more on choosing a risk level, see risk per trade in gold trading.

Step 4: Check the crash periods individually

A long-run average can hide how a strategy behaved in specific bad markets. Look for results in named crashes such as the 2008 financial crisis, the 2020 Covid crash and the 2022 rate-hike bear market, each compared with the benchmark.

Worked example: when crash tests disagree with the average

Market Switch holds SPY in normal markets, switches to UVXY (a volatility fund) when many markets look overheated, and buys leveraged funds after sharp falls. In its historical backtest from November 2011 to November 2026, it returned +71.9% a year with a worst fall of 42%, against +15.0% a year and a 34% worst fall for the S&P 500, and +19.2% a year and a 35% worst fall for the Nasdaq 100. Its deepest fall, in March 2020, recovered about 3 months after the bottom.

The crash tests tell a very different story for earlier years. Using a simulated three-fund version:

A strong average from 2011 onwards sits alongside simulated results that would have wiped out the account in earlier crashes. The caveat explains part of this: nearly half of its tested profit came from UVXY, and it failed outside the US. UVXY itself has lost most of its value over long periods, so a strategy that leans on it depends heavily on timing. Future results will differ. Our article on how strategies held up in 2008, 2020 and 2022 goes through all five strategies.

Step 5: Ask whether costs were included

Look for a stated cost per trade or a note on slippage. If there is none, assume the results are optimistic. Then think about how often the strategy trades. A monthly strategy is less sensitive to costs than an hourly one. The honest sign is when a provider tells you how sensitive the result is, as with the Gold Breakout caveat above.

Step 6: Ask how the assets and rules were chosen

This is where most impressive backtests quietly fall apart.

Were the assets picked with hindsight? If a strategy holds a fixed list of stocks, ask when and how that list was chosen. Picking today's winners and testing them over the past fifteen years is a guaranteed way to produce a good-looking chart.

Does most of the profit come from a few names? Concentration means the result depends on a small number of outcomes that may not repeat. Big Tech Trend discloses that its stock list was chosen with hindsight and that three stocks made 81% of the profit.

How many versions were tried? If a researcher tested hundreds of rule variations and published the best one, the best one is likely to be partly luck. See overfitting in trading strategies and hindsight and survivorship bias.

Step 7: Translate large numbers into plain terms

Long backtests compound into figures that are hard to grasp. $10,000 becoming millions sounds like a different universe from $10,000 becoming $80,000, but both can come from yearly percentages that look reasonable on their own. Compare yearly rates, not end balances, and remember that an end balance is the product of one specific path that will not repeat. Our guide to CAGR vs total return covers this.

Step 8: Look for the stated weakness

Every strategy has conditions in which it struggles. A trustworthy backtest says what they are. An untrustworthy one shows only a rising line and words like "proven" or "consistent profits".

The weakness should be specific. "Past performance does not guarantee future results" is a legal line, not a disclosure. "Three stocks made 81% of the profit" or "doubling trading costs cuts its yearly result roughly in half" is a disclosure: it tells you exactly where the result is fragile and what to watch.

When you find the stated weakness, use it to plan your own checks. If a strategy depends heavily on one fund, watch how often it holds that fund during a paper test. If it is sensitive to costs, compare paper fills with real prices. If its stock list was chosen with hindsight, give its long-run figure less weight and pay more attention to how it behaves on new prices.

If you cannot find any stated weakness at all, treat that as the weakness. It usually means either the provider has not looked closely or does not want you to.

A quick checklist to keep

Key takeaways

• Read the benchmark, the worst fall and the recovery time before the headline return.
• Compare the same strategy at different settings to see the trade-off between return and loss.
• Check crash periods individually; long-run averages can hide catastrophic simulated results.
• Ask how assets and rules were chosen; hindsight and concentration flatter many backtests.
• Every backtest is a simulation on past prices. Nobody earned the figures, and future results will differ.

Watch the rules on live prices

A backtest is one piece of evidence. You can add a second by watching AlgolaneBots's five strategies on real, current prices: start a free 90-day paper trial with $10,000 of pretend money and no card. Full backtests and caveats are on the strategies overview, and our risk page covers the risks.