Crypto Backtesting: What to Test and What to Ignore

Every trader has a strategy that looks perfect on the chart. The question is whether it worked across hundreds of trades and different markets, or only on the five examples you remember. Backtesting answers that before the market does. Done badly, it gives you false confidence, which costs more than no test at all.

• Backtesting means testing your trading rules on historical data to see how they would have performed

• Test at least 100 trades, across bull, bear, and sideways markets

• Focus on expectancy, max drawdown, and the longest losing streak, not total return

• Include fees, slippage, and funding, or your crypto backtest will look better than reality

• Avoid overfitting: fewer rules, fewer tweaks, and always test on data you did not optimise on

• Forward test next. A strategy is only ready when it holds up in paper trading too

Backtesting is the process of applying a set of trading rules to historical price data to see what would have happened if you had traded them. You record every entry, exit, win, and loss, then measure the results.

A backtest does not predict the future. It filters out strategies that never worked, and shows you what the bad periods look like before you meet them with real money.

1. Write the rules down first. Entry, stop loss, take profit, position size, and when not to trade, specific enough to apply the same way every time. Our trading plan template is a good starting point

2. Pick your market and timeframe. Test the pair and timeframe you will actually trade

3. Choose a data period that includes different conditions. For crypto, that means at least one bull run, one bear market, and one long sideways range. A strategy tested only on 2021 has only been tested on a bull market

4. Log every trade. Date, direction, entry, stop, exit, result in R (multiples of your risk), and market context. Do not skip the trades you would have "obviously avoided"

5. Calculate the metrics. Expectancy, win rate, average win and loss, max drawdown, and the longest losing streak (see below)

6. Test on data you have not seen. Hold back the most recent 20 to 30% of your data, and only test on it once your rules are final

Manual Backtesting

Manual backtesting means stepping through historical charts candle by candle and recording trades as if they were live. It is slower than automated testing, but it suits discretionary strategies like breakout trading and trains your eye at the same time. The key rule: hide the future, or your entries will be better than they ever are live.

TradingView Backtesting and the Strategy Tester

TradingView covers both:

• Bar Replay hides future candles and plays the chart forward one bar at a time, which makes manual backtesting much more honest. Intraday history depends on your plan. On Essential, for example, you get 6 months of 1 minute data, and Plus extends that to a year (TradingView)

• The Strategy Tester runs automated backtests on Pine Script strategies and reports net profit, drawdown, win rate, and profit factor. Set realistic commission and slippage in the properties before trusting any result

Expectancy is the number that matters most:

Expectancy = (win rate × average win) - (loss rate × average loss)

A 40% win rate with 2R winners and 1R losers gives (0.4 × 2) - (0.6 × 1) = +0.2R per trade, roughly +20% over 100 trades at 1% risk, before costs.

The losing streak is the number most traders underestimate. At a 40% win rate, a run of 7 or more losses in a row within 100 trades is normal, not bad luck. Your backtest should show it, and your risk to reward and position size need to survive it.

• Total return on its own. Without the drawdown that came with it, it means nothing

• Small samples. Twenty trades can make any strategy look brilliant or useless

• One great period. If most of the profit came from one move, you found an event, not an edge

• Perfect fills. Exact lows and exact highs do not happen live

• Endless tweaking. If it only works with a 17 period moving average and not 15 or 20, it does not really work

Curve Fitting and Overfitting

Curve fitting, or overfitting, happens when you keep adjusting rules until they fit past data perfectly. The backtest looks excellent, then fails live, because it learned the noise rather than the pattern. Research by Bailey, Borwein, López de Prado and Zhu shows that high backtested performance is easy to achieve after trying a relatively small number of strategy variations, and the more variations you try, the more likely the result is overfit (Notices of the AMS, 2014).

How to protect yourself:

• Keep the rules simple, with as few parameters as possible

• Decide the rules before you look at results, and limit how many versions you test

• Check that nearby settings still work, not just the best one

Walk Forward Analysis

Walk forward analysis is a stricter form of out-of-sample testing. You optimise on one window of data, for example 12 months, test on the next unseen 3 months, then roll both windows forward and repeat. If the strategy keeps performing on unseen data, the edge is more likely real.

• Fees add up fast. Taker fees on every entry and exit can turn a small edge negative. Model them using real rates, and see our guide to maker vs taker fees

• Slippage in fast markets. Breakouts and news spikes rarely fill at your level. Add slippage to every trade, especially on smaller coins

• Funding on perpetuals. Holding a perpetual position for days costs or pays funding every few hours. Most simple backtests ignore it

• 24/7 markets. There is no close, and weekends trade differently from weekdays

• Survivorship bias. Testing only on coins that still exist today ignores the ones that collapsed. Stick to liquid majors or include delisted coins

• Regime changes. Crypto swings between violent trends and long ranges. Check results for bull and bear markets separately

A backtest tells you a strategy worked on past data. Forward testing tells you whether it works now, and whether you can actually execute it.

Paper trading crypto on a demo account is the simplest form of forward testing. Trade your rules on live prices for 30 to 50 trades and compare with your backtest. If live results are much worse, the cause is usually unmodelled costs, rules you are not following, or overfitting. A trade journal will show you which.

A challenge adds hard limits, so test against them, not just against profit.

• Check the worst day. At Mubite, the daily drawdown is 5%, calculated from equity at 00:00 UTC, and floating losses count. Find the worst day in your backtest and make sure it stays inside that limit

• Check the worst losing streak. On a $100,000 2-Step account the maximum drawdown is 8%. If your backtest shows 8 losses in a row, risking 1% per trade takes you to the edge of that limit on a single bad streak. Risking 0.5% would not. See drawdown explained and position sizing for prop traders

• Forward test before you pay. The Mubite free trial lets you run your strategy on live market data under challenge rules, with no money at risk

Tip from the author: Size your risk from the longest losing streak in your backtest, then add 50%. If your backtest shows 8 losses in a row, plan for 12. Backtests are always a little kinder than the live market, and the losing streak is where that difference hurts most.

Need a strategy to test? Try trend following, mean reversion, or the opening range breakout.