A backtest trading strategy exercise runs a strategy’s rules on past prices to show what would have happened, before any real money is at stake. It is one of the cheapest ways to find out that an idea is weak, but a good result is not proof that the strategy will work from now on. A backtest answers one narrow question: did these rules fit the past? Whether they fit the future is a different question.
In brief
- A backtest applies a fixed set of entry and exit rules to historical prices and records the trades, the gains, the losses and the largest fall along the way.
- Its main value is to reject weak ideas early, and to show how deep and how long the losing periods of a strategy could be.
- Its main danger is overfitting: the more variations you try on the same data, the more likely one of them looks good by pure chance.
- Fees, slippage and unrealistic data can make a backtest look better than real trading would have been.
- Treat the result as a hypothesis to check with a small amount or with paper trading, never as a forecast.
In this article · 10 min read
- Backtest trading strategy: what it is, in one sentence
- How to backtest trading strategy rules, step by step
- The number that shows how easy it is to fool a backtest
- Common mistakes that make a backtest too good
- Advantages and limits of backtesting
- Where a backtest trading strategy check fits with Crypto Go Bot
- How to start, step by step
- Checklist before trusting a backtest
- Who it is for
- Key terms
- Frequently asked questions
- Sources
Backtest trading strategy: what it is, in one sentence
A backtest is a simulation that applies a trading strategy to historical market data to estimate how it would have performed. The Commodity Futures Trading Commission (CFTC) has a rule for this kind of result in its advertising regulation: simulated or hypothetical results must carry a warning that they have inherent limitations, that unlike an actual performance record they do not represent actual trading, and that they are designed with the benefit of hindsight. That warning is a good summary of how to read any backtest.
A few words you will meet again below. Historical data is the record of past prices (and sometimes volumes). In-sample data is the part of that history you used to build and tune the rules; out-of-sample data is a part you kept aside and did not touch while designing them. Slippage is the gap between the price you expected and the price at which your order was really filled.
How to backtest trading strategy rules, step by step
The method is the same whether you code it, use a platform that has a backtesting tool, or even do it by hand on a chart:
- Write the rules down. When do you enter, when do you exit, how much do you risk per trade, and where is the stop loss (the order that closes a losing trade at a set price)? A rule you cannot write precisely cannot be tested.
- Choose the data. Use a period long enough to include rising, falling and sideways markets. Make sure the data is clean and that it covers the coins you would really trade.
- Run the simulation. The software applies the rules candle by candle and records every hypothetical trade, its fees and the balance over time.
- Read the results. Look beyond the final profit. Check the drawdown (the largest fall from a peak to a low), the number of trades, the share of winning trades and the longest losing streak.
- Test on data the rules have not seen. Keep a period aside, or move forward in time step by step (walk-forward testing), and see whether the behaviour holds.
For crypto there is a practical point: markets trade all day and all year, and the history of many coins is short. A coin that has existed for two years gives you far less to learn from than a stock index with decades of prices. A backtest on a short, one-directional stretch can say very little about a different phase of the market.
The number that shows how easy it is to fool a backtest
The best-known study of this problem is "Pseudo-Mathematics and Financial Charlatanism", by David Bailey, Jonathan Borwein, Marcos López de Prado and Qiji Zhu, published in 2014. They show that if a researcher tries enough variations of a strategy on the same data, a backtest can be made to look good almost whatever the strategy's real merit. In their example, after trying only 7 independent configurations, the best one is expected to show an annualised Sharpe ratio of 1 on a 2-year backtest, while the expected result on new data is zero. (The Sharpe ratio is a measure of return per unit of risk; 1 looks respectable.) With 5 years of data, the limit rises to 45 configurations.
In plain words: the shorter your history, the fewer things you are allowed to try before the winner is probably luck. And the authors note that most backtests do not report how many configurations were tried, so a reader cannot judge the risk. That is why, in our experience, the most useful habit is to keep a written count of every version of a strategy you test.
A second data point comes from a 2016 study of 888 trading algorithms built on the Quantopian platform. As summarised by Quantpedia, the researchers found that commonly reported backtest metrics such as the Sharpe ratio offered little value in predicting performance on new data (R² below 0.025, meaning they explained almost none of the variation).
Common mistakes that make a backtest too good
| Mistake | What it means | Why it flatters the result |
|---|---|---|
| Overfitting | Tuning the rules until they fit the past perfectly | The rules memorise noise that will not come back |
| Look-ahead bias | Using information that was not available at the time of the trade | The simulation "knows" the future |
| Survivorship bias | Testing only the coins that still exist today | Failed coins are missing, so average results look better |
| Ignoring costs | Leaving out fees, spreads and slippage | Small gains per trade can disappear once costs are counted |
| Too few trades | Drawing conclusions from a handful of trades | A short lucky run looks like skill |
Costs deserve special attention for any strategy that trades often. A gain of a fraction of a percent per trade can be wiped out by fees and slippage, and a simulator that fills every order at the exact price shown on the chart cannot see this. Kraken, for example, describes its grid tool as letting users backtest an exact configuration against historical price data before a single order goes live, and Finance Magnates reports that the exchange itself says traders remain responsible for ranges, monitoring and risk settings. Even the tool's own maker does not treat the backtest as a promise.
Advantages and limits of backtesting
Advantages. It costs nothing in capital, it forces you to write precise rules, and it can reject a bad idea in an afternoon. It also shows what the bad periods would have felt like: a strategy with a good average but a 50% drawdown is a very different thing to live with than one with a 15% drawdown. And it gives you a baseline to compare with later, real results.
Limits. The past is one path among many that could have happened. The market may behave differently next year, other traders may have adapted, and your own behaviour under real losses is not in the simulation. A backtest also says nothing about events that were not in the data. In our experience, the strategies that disappoint most in real life are those that looked best after the most tuning.
John Bax's view
In Lazy Investor, Smart Investor, John Bax explains that anyone can technically build a trading bot, but one that works needs a deep understanding of the markets and the ability to adapt the strategy to ever-changing conditions. This is the practical lesson behind backtesting: the conditions in your test period will not stay the same, so a rule set is never "finished". The book is also clear that no system can guarantee profits, which is the right attitude to bring to any simulated result.
Where a backtest trading strategy check fits with Crypto Go Bot
Crypto Go Bot does not run a rule set that you tune on a price chart. It is a signal bot: it follows the Smart Money, the on-chain flows of wallets with a track record of profitable trades, read as Low, Medium or Strong, and it opens trades on your own Kraken account, with take profit and stop loss as real orders on Kraken. Our overview of how a crypto trading bot works explains the method. The lesson of this article still applies to any bot or signal service: ask how its results were measured, how many versions were tried, and what the numbers look like once costs are included.
The same discipline applies to the exit rules you choose. Our guides on stop loss and take profit and on how much to risk per trade cover the settings you would test, and our article on grid trading shows a strategy whose backtest depends heavily on the range you choose. Past results do not guarantee future ones, and losing trades are a normal part of trading.
Dangers of Backtest Overfitting (Marcos López de Prado) (in English)
How to start, step by step
- Pick one simple idea and write its rules in a few lines, including the exit.
- Decide the number of versions in advance. A short list is better than endless tweaking.
- Split the history. Build on the older part, check on the newer part, and do not go back to adjust after you have seen the check.
- Include costs. Use realistic fees and add slippage for fast markets.
- Look at the worst case, not only the average: the drawdown, the longest losing streak and the number of trades.
- Run it on paper or with a very small amount for a while before trusting it.
- Keep a journal of every version, the data period and the result.
Checklist before trusting a backtest
- I know how many versions of the rules I tried.
- I tested on data the rules had not seen.
- Fees and slippage are in the numbers.
- The sample has enough trades to mean something.
- I can accept the drawdown shown, in real money, without changing the plan.
Who it is for
Backtesting is a skill for anyone who wants to examine a trading idea with evidence before putting money behind it.
- Beginner with a first idea: writing the rules down and testing them on past prices shows quickly whether the idea is worth any further effort, as long as you treat the result as a starting point.
- Spot trader who follows a method: a backtest helps compare variations of an entry or an exit and see the typical size of losing streaks before a real one arrives.
- Person who uses a bot or a signal service: knowing how backtests mislead helps you ask better questions about any tool's claimed results.
- Cautious investor with a small budget: testing first, then trading small, keeps the cost of a wrong idea low, as long as the money can be lost without hurting your plans.
Key terms
- Backtest: a simulation of a strategy on past market data.
- Overfitting: tuning a strategy so closely to past data that it fails on new data.
- Out-of-sample: data kept aside and not used to build the rules.
- Drawdown: the fall from a peak to the following low, in percent.
- Slippage: the difference between the expected price and the price actually obtained.
- Paper trading: trading with virtual money in real time to see how the rules behave live.
Frequently asked questions
What is a backtest in trading?
It is a test of a trading strategy on historical prices. The rules are applied to the past data to see the trades they would have produced and the result.
How do I backtest trading strategy rules?
Write precise entry, exit and risk rules, choose a long and varied period of data, run the simulation with realistic costs, and then check the rules on data they have not seen.
Can a backtest predict future profits?
No. It only shows how the rules fit the past. Regulators require simulated results to be labelled as such because they are designed with hindsight and do not represent actual trading.
What is overfitting in a backtest?
It means tuning the rules until they match the past data too closely. The more versions you try on the same data, the more likely a good-looking result is luck.
How much data do I need to backtest a crypto strategy?
There is no fixed number. A longer period that includes rising, falling and sideways phases is more informative, and a short history allows only very few variations to be tried.
Is backtesting enough before using a bot?
No. It is a first filter. Follow it with a small amount or paper trading, and keep checking the results against costs and real conditions.
Does Crypto Go Bot use backtesting?
Crypto Go Bot follows Smart Money signals read from on-chain flows rather than a rule set tuned on a price chart. The questions in this article still help you judge any trading tool, including how its results are measured.
Sources
- Wikipedia, Backtesting
- Bailey, Borwein, López de Prado and Zhu, Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance, Notices of the American Mathematical Society, May 2014
- Wiecki, Campbell, Lent and Stauth, All that Glitters Is Not Gold: Comparing Backtest and Out-of-Sample Performance on a Large Cohort of Trading Algorithms, summarised by Quantpedia, 4 May 2016
- CFTC, 17 CFR 4.41, Advertising by commodity pool operators, commodity trading advisors, and the principals thereof
- Kraken, Introducing Smart Grid, 9 September 2026
- Finance Magnates, Kraken Adds Editable Grid Bot with Backtesting to Desktop App, 9 September 2026
