A Backtest Checklist: Costs, Walk Forward Tests and Honest Results

22 October 2026

A backtest replays a rule over past prices and reports what would have happened. It is a wonderful learning tool, and it also makes it easy to fool yourself. The same rule can look brilliant or ordinary depending on how carefully the test was built. This checklist helps you build tests whose results mean something.

First, label results accurately

A backtest is a simulation on history. A paper portfolio is a simulation running forward in time with assumed fills. A live account holds real orders. These are three different things, and a good report names which one it shows. On BitBank, the /coins page shows a backtest of the best strategy over the last seven days of candles, and the backtester runs simulated fills. Each result is a simulation, and reading it that way sets the right expectations.

Step 1: write the rule before you look

Decide your entry, exit, fee level and position size first, then run it once. Every tweak after seeing a result is another experiment on the same history, and enough experiments will produce a winner by chance alone. Writing the rule first keeps your test a real question.

Step 2: include costs

Fees and slippage apply to every trade. A rule that trades hourly pays them many times, so small costs add up. The backtester lets you set the fee rate and slippage in basis points. Try doubling both. A rule that stays healthy with doubled costs shows resilience, and a rule that needs perfect fills shows fragility.

Step 3: split time into train and test

Shape your rule on an earlier period, then judge it on a later period it has never seen. Walk forward validation repeats this many times. Fit on months one to six and test on month seven. Fit on months two to seven and test on month eight. Continue until the data ends. Count how many test windows did well, not only the average. Steady results across windows build confidence, and a single lucky window adds little.

Step 4: count the trades

Twenty trades can reflect luck. A few hundred trades across different market moods tell a richer story. Read the trade count first, and the return second. Also look at results by coin and by month. If one coin or one week supplies most of the gain, the rule may rely on that single event.

Step 5: study the worst stretch

Maximum drawdown is the largest fall from a peak to a later low. It answers a practical question: could you hold this rule through its hardest period? A strong total return with a deep drawdown may test your resolve more than a modest return with a shallow one. Decide your tolerance before you test.

Step 6: compare against simple alternatives

Place the result beside holding the coin and beside doing nothing. If a complex rule matches a simple hold after costs, the simple hold wins on effort. The goal of a test is to learn what adds value.

Step 7: watch it forward

After a rule survives the earlier steps, observe it on fresh data as it arrives. A forward paper record cannot be tuned after the fact, so it is the cleanest evidence. The trading bot page shows simulated results with their assumptions, which makes a good example of how to present a forward record.

A quick checklist to keep

• Rule written before testing.
• Fees and slippage included, then doubled as a stress check.
• Train period and test period kept separate, with rolling windows.
• Trade count, per coin results and drawdown reviewed.
• Comparison with a simple hold.
• Forward observation before any real decision.

Start practising

Open the BitBank backtester, choose one pair and one rule, and follow the checklist from top to bottom. For deeper reading, see honest backtesting and reading forecasts and bots honestly.

This article is educational and general information only. Simulated and historical results offer no guarantee of future outcomes, and real fills will differ.