Backtesting journal for honest manual testing

A manual backtest is only as good as its records: skipped trades, rules changed halfway and a sample of twenty are how a losing idea starts to look profitable. Here is how to run a manual backtest step by step, how many trades you need, and how Haunted keeps the results in a free backtesting journal.

What a backtesting journal is, and what it is not

A manual backtest means scrolling a historical chart bar by bar, usually with a replay tool, taking every trade your rules allow and writing it down. The chart tool shows the past; the journal is where the decisions and results are stored and measured. Haunted is that journal. It does not run or automate backtests, simulate fills or test code: you replay the chart in your platform and log each test trade in the backtests section with outcome, R:R, risk, session, model, tags, notes and a pasted screenshot. Haunted then builds the equity curve and calculates win rate, expectancy and profit factor for the test, so the result is a set of numbers rather than a feeling that the idea works.

How to run a manual backtest step by step

Start by writing the rules before you look at the chart: what defines the entry, where the stop goes, where the target or exit is, which filters apply and which sessions you trade. Then choose the market and a period that includes different conditions, not only the last strong trend. Hide the future with bar replay and move forward one candle at a time. Take every signal that fits the rules, including the ones you do not like, and log each trade immediately with a screenshot of the entry. Do not review the statistics until the batch is finished, because looking at the running result invites quiet changes to the rules. Only after the batch do you analyse and decide what to adjust.

How many trades a backtest needs

Small samples lie. With a true win rate of 50 percent, a test of 30 trades can easily show anything from roughly a third to two thirds wins purely by chance; at 100 trades the typical range narrows to about plus or minus ten percentage points. That is why a common working minimum is around 100 trades per model, spread over different market conditions: trend and range, quiet and volatile periods, several months rather than a single week. If a model takes a trade only once a week, test it on more history or on several instruments with similar behaviour. Haunted keeps each test's trades together, so you can see how the equity curve and expectancy settle as the sample grows.

Record rules consistently, or the test means nothing

A backtest compares trades with each other, so every trade has to be described the same way. Decide in advance what counts as a win, how partial exits are recorded, whether the result is measured in R and how sessions are defined, then stick to it. If you change a rule halfway, do not keep writing into the same test: start a new test or mark the new version with a tag, otherwise you are averaging two different strategies. In Haunted each backtest trade has the same fields as a live trade, including model, tags and notes, so filtering by rule version, session or condition later takes one click instead of rereading hundreds of rows in a spreadsheet.

Backtest, forward test and live trading

A backtest shows how rules behaved on history. A forward test runs the same rules in real time, usually on a demo or minimal size, and checks whether you can actually execute them when the right side of the chart is hidden. Live trading adds slippage, emotions and the cost of real money. Expect each step to be somewhat worse than the previous one; if live expectancy is far below the backtest, the usual culprits are hindsight in the test or execution errors. Because backtests and live trades in Haunted use the same metrics, expectancy, profit factor, win rate and the equity curve, you can compare the three stages directly and see where the edge leaks.

Common backtesting mistakes

The most frequent is hindsight: seeing where price went and unconsciously taking only the obvious winners. Close to it is cherry-picking, testing only the period where the idea looked good. Another is curve fitting, adding filter after filter until the past looks perfect while the model becomes too specific to work in the future. Many tests also ignore costs: spread, commission and slippage can turn a small positive expectancy into a loss, so record results net. Finally, traders often stop at twenty trades because the result already looks convincing. A journal with fixed fields, screenshots for every trade and notes on why a signal was taken or skipped makes each of these mistakes visible on review.

FAQ

How do I backtest a strategy manually?

Write the rules first, pick a market and a varied period, hide future candles with bar replay and step forward one bar at a time. Take every valid signal, log each trade right away with a screenshot, and analyse the statistics only after the whole batch is complete.

How many trades are enough for a backtest?

A common working minimum is about 100 trades per model, collected across different market conditions. Thirty trades leave too much room for chance: a strategy with a 50 percent true win rate can show anywhere from roughly a third to two thirds wins on such a sample.

Does Haunted run backtests automatically?

No. Haunted is a journal, not a backtesting engine. You replay the chart in your trading platform and log each test trade in Haunted, which then calculates win rate, expectancy, profit factor and the equity curve for the test and lets you compare it with live results.

What is the difference between a backtest and a forward test?

A backtest applies the rules to historical data you replay. A forward test applies the same rules in real time, usually on a demo or small size, without knowing what comes next. The forward test checks whether you can execute the rules, not just whether they worked in the past.

Is the backtesting journal free?

Yes. Haunted is entirely free, with no paid tier and no trial. The backtests section, live trade journal, analytics, prop account tracking, ideas, documents and share links are all included. It runs in the browser on a computer and on a phone.