Backtesting
A backtest is a kind of time-machine game. You pick one rule — 'buy when this condition hits, sell when that happens' — then go back through the old chart, pretend you traded that rule like a machine, and add up every result. Because you're counting with data instead of gut feel or memory, you avoid the selective-memory trap of 'I would have made money if I'd bought back then.'
Say you backtest the rule 'buy when RSI drops below 30, sell 24 hours later.' Over two years of hourly candles the signal fired 500 times, and 260 of those times price was higher 24 hours later — a 52% win rate. A coin flip is 50%, so now you know this rule is only a tiny bit better than a coin.
Win rate alone isn't enough. You also need how much you make when you're right versus how much you lose when you're wrong. A 52% win rate that makes 1% on wins and loses 2% on losses still shrinks your account. Subtract the fees paid on every trade, and a thin edge can vanish entirely.
Backtests usually break in one of three ways. First, lookahead leakage — sneaking in information you couldn't have known at the time (like a day's closing price before that close was final) makes results look fake-good. Second, ignoring fees and fills — in real life you pay fees and don't always get filled at your price; leave that out and the numbers come out optimistic. Third, too few samples — if a signal only fired 10 times in history, even a 70% win rate can be pure luck.
So the honest way to use a backtest is not as proof that 'this rule makes money,' but as a tool for checking folk wisdom against the numbers. When someone says 'oversold means a bounce,' you count what actually happened and look at the past distribution. As long as you remember it's a record of the past, not a prediction of the future, it's the best practice a beginner can get at separating folklore from data.
What the data actually shows
Barobara's setup catalog and the /odds pages are exactly these backtest results. We publish how many times each popular Bitcoin chart signal actually fired in the past and what percentage of the time price rose afterward — and spoiler: for most signals the outcome is close to a coin flip. For example, check the RSI oversold (1h) results to see how far the 'oversold means bounce' folklore sits from the actual numbers. How much fees eat into the results is on the fees page.Common misconceptions
'A high backtest win rate means you'll make money' is a common misconception. Win rate only counts how often you were right — it says nothing about how much you make when right versus how much you lose when wrong. Even a 90%-win-rate rule can bleed your account if one loss in ten is a big one.
'It worked in the past, so it will work in the future' is dangerous thinking too. When the market's mood changes, rules that used to work often stop working. A backtest is a historical record, not a guarantee.
FAQ
Q. Do I need to know how to code to backtest?
Doing it yourself takes programming, but starting by reading results someone already computed is perfectly fine. Look first at resources that publish per-signal historical results, like Barobara's /odds pages. When you do try it yourself, pay special attention to including fees and getting enough samples.
Q. Is a 52% backtest win rate good?
It's barely different from a coin flip (50%), and it's the kind of edge that can disappear once you subtract fees. On top of that, if the sample count is small, that 2-point gap could itself be luck. Never look at win rate alone — check sample size, the size of wins and losses, and fees together.