What should I check first in a backtest?

The window and the assumptions: what date range, whether fees are included, and whether dividends are reinvested. Then read total return, annualized return and maximum drawdown together, not the single prettiest number.

Why look at calendar years and not just the total?

A smooth average can hide a terrible single year. Calendar years show whether returns came steadily or in a few lucky bursts — and how a strategy behaves when markets are bad.

How to read a backtest honestly

Q: Can a backtest predict the future?

No. A backtest only describes how a strategy or fund would have behaved in the past. It can't promise future results — but it can reveal how a strategy handles different market conditions, which is genuinely useful.

SeeFund · educational series · 2026-09-06

A backtest is a useful tool and an easy one to fool yourself with. A smooth line and a big “total return” at the end feel convincing — but what a backtest can really tell you is narrower than it looks, and just as valuable. This is a plain-language checklist for reading one without lying to yourself.

1. Check the window and the assumptions first

Before any number means anything, know what you’re looking at:

What date range? A 10-year backtest and a 3-year backtest are different stories. More years usually means more honesty — and more different markets lived through.
Are fees included? “Net” (after fees) is what you’d actually get. Gross numbers flatter.
Are dividends reinvested? See Dividends reinvested vs paid out — both are legitimate, but they must be compared on the same convention.
Which start and end dates? Starting a backtest just after a crash, or ending it at a peak, quietly changes the whole picture. Prefer windows that include full cycles.

2. Read the numbers together, never alone

The total return at the end is the headline, but it hides how you got there:

Annualized return — what the growth really averaged per year (see Total vs annualized return).
Maximum drawdown — the worst fall you’d have lived through (see Maximum drawdown). Can you actually sit through that?
Risk-adjusted measures like Sharpe — how much bumpiness came with the return (see Sharpe ratio).

3. Look at the calendar years, not just the average

A smooth average can hide one terrible year — or one lucky one that carried the whole result. Calendar years show you:

Did the return arrive steadily, or in a few lucky bursts?
How did it behave when the market was bad — not just over the whole window?
Was the recent period the source of the returns (a clue that you’re seeing a recent style in fashion)?

4. Remember what a backtest can and can’t say

Can: describe how a strategy or fund behaved in the past, and show how it handled different conditions.
Can’t: promise future results. Past performance is not a forecast — a strategy that backtests beautifully can still underperform going forward, and the more a backtest was “tuned” to look good, the less you should trust it.

Honesty rule of thumb: a backtest is evidence about the past and a question about the future — never a guarantee. If a number looks too good, ask which window, which fees, and which start date produced it.

FAQ

Can a backtest predict the future?

No. A backtest only describes how a strategy would have behaved in the past. What it can reveal is how a strategy handles different market conditions — which is genuinely useful, as long as you don’t mistake it for a promise.

Why do two tools show different results for the same fund?

Almost always because of different assumptions — date range, fees included or not, dividends reinvested or not, or a different benchmark. Always check the labels before comparing.

Is a longer backtest always better?

Usually, because it covers more market conditions — but only if it’s the same fund or strategy the whole way, and the same assumptions throughout.

Run a backtest with the window, fees and reinvestment assumptions labeled — and see the calendar years behind the headline: open SeeFund. History is shown as history, never sold as a forecast.