When a strategy stops working

Every strategy that makes money has some edge – a reason why it should make more than it loses. But edges are not forever. The market changes, other traders learn, and what worked three years ago may today only pay fees.

Why an edge fades#

• The market has changed. You know from the first lesson that Bitcoin's median daily ATR fell from 6.1% in 2021 to 2.9% in 2025. A strategy built on wild moves has fewer opportunities in a calmer market, and fees take a bigger bite out of them.
• Others have found the edge. When many people trade the same signal, they get in earlier and the price moves before you get to it.
• The edge was never real. The most common case. The strategy was overfitted to the past – like our 40-day average, which had a profit factor of 7.6 during tuning and 2.4 out of sample.

Bad luck or a broken strategy?#

This is where the number of trades helps. Say the backtest promises a 40% win rate and you have only 30%. How likely is it that this is just bad luck and the strategy is fine?

After 20 trades, a win rate of 30% or less happens one time in four – a normal swing. After 100 trades only in 2.5% of cases. There it is much more likely that something has changed.

The same goes for the other numbers from your review: the average win, the drawdown, the length of losing streaks. One number out of range is a warning. Several numbers out of range after dozens of trades are a signal.

What to do when a strategy is failing#

1. Are you trading it by the rules? Look at your journal. If not, the problem is not the strategy.
2. Is the drawdown within the backtest range? If so, it is most likely a normal period – stick to your drawdown rules.
3. Are the numbers out of range after enough trades? If so, switch the strategy off by the rule set in advance.
4. Has the market changed? Look for an explanation – a different regime, lower volatility, higher costs. If you find none, take it as a warning that the edge may never have been real.
5. Back to research. You test a modified strategy all the way again: backtest, out of sample, forward test. A new rule "fixed" on the latest losses is just more overfitting.

Common myths#

• "The strategy worked for three years, it will keep working." Edges fade. Watch the numbers, not the past.
• "Five losses in a row, the strategy is dead." After a few trades it is almost always noise.
• "I will add a filter and it will work again." On past data, it always does.