A backtest is a simulation: take a set of rules, run them over past prices, and see what would have happened. Done honestly, it is the best tool there is for understanding a strategy before risking anything on it. Done to impress, it is the easiest number in finance to manufacture. The same word covers both, so you need a way to tell them apart.
The seven checks
1. Is it labeled as a backtest? This sounds trivial. It is the whole game. A figure that is not labeled backtest, paper or live is a figure you cannot evaluate, because you do not know what produced it. The desk labels every number and never blends the three; a page that mixes them has already decided how you should feel.
2. What is the window, and is the worst year in it? A backtest that starts after the last crash is not a backtest of the strategy; it is a backtest of the good times. Ask for the full window and look for the down year. CopyOrders' six-sleeve backtest is quoted over its full window, and the worst calendar year in it, 2022, is quoted on the same page as the return: −2.2%, next to 9.5% compound annual growth and a −6.3% maximum drawdown.
3. Were the rules fixed before the test, or fitted to it? Any rule set with enough parameters can be tuned until the past looks perfect. The sign is a strategy with many precise thresholds and a beautiful curve. The cure is walk-forward testing: fix the rules on one stretch of history, test them on the next, and repeat, so the test never sees the data the rules were built on.
4. How many strategies were tried before this one? If a desk tested a hundred variations and shows you the best, the best is largely luck: the luckiest tail of a hundred draws will look excellent whether or not any skill is involved. Ask how many were tried. An honest answer is a number; a dishonest one is silence.
5. Does it include costs and slippage? Fees, spreads and the gap between the price in the data and the price you would actually have got. A backtest with zero costs is a backtest of a world that does not exist. Over many trades, costs alone can turn a positive curve negative.
6. Does the drawdown match the return? A high return with a tiny drawdown over a long window is the single most common fingerprint of an overfitted or mislabeled test. Real strategies pay for their returns in drawdown. If the price looks too low, the return is probably not real.
7. Is there a separate paper or live record? A backtest is a hypothesis. Paper trading, executed in real time on a simulated account, is the first evidence. Live trading with real capital is the second. A desk that publishes all three, separately and labeled, is showing you its homework; a desk that publishes only the backtest is showing you the hypothesis and calling it a result.
What a backtest can tell you
Read honestly, a backtest answers useful questions. Roughly how deep and how long are the drawdowns, so you know what the strategy would ask you to sit through? How does it behave in a regime like the one you are in? Which sleeves carry it and which stand aside? Does it depend on one market or one year?
What it cannot tell you
It cannot tell you what you will earn. It knew the rules in advance, never slipped on a fill, never hesitated at the bottom of a drawdown, and ran over a past that will not repeat. That is why the desk's method pages call the backtest hypothetical every time it is quoted, and why the paper and live records are kept shorter and separate rather than appended to it to make the curve longer.
When to walk away
Walk away when any of these are true: the figure has no label; the window excludes the bad years; the drawdown is too small for the return; the number of strategies tried is a secret; costs are absent; or there is no paper or live record at all after the backtest has been "proven" for years. None of these mean the strategy is bad. They mean you cannot know, and a position you cannot evaluate is a position you should not take.
Backtested figures are hypothetical, do not represent actual trading, and have inherent limitations. Past performance of any kind does not predict future results. Nothing here is personalised investment advice.