Volatility Regimes: Why Your Bot Wins in Chop but Loses in Trends
Most strategies do not fail because they are bad. They fail because the market changed character. A grid or DCA bot that prints steady profits in a calm, ranging market can get run over the moment volatility spikes and price trends hard in one direction. The strategy did not break. The regime did.
A single blended backtest result buries this. It averages your calm-market wins and your volatile-market losses into one number that describes neither. Freya's backtest instead sorts your trades by the market's volatility at the time and shows how you performed in each regime, so you can see the exact conditions your strategy needs to survive.
Key Takeaways
- Freya classifies every trade by market volatility at the time, using ATR, into low, medium, and high regimes.
- For each regime it reports trade count, win rate, average PnL, total PnL, and average trade duration.
- It names your best-performing regime, which is the market character your strategy is actually built for.
- A bot that is strongly positive in one regime and negative in another is not broken; it is specialized, and it needs the right market.
- The ATR percentiles show how volatile the backtest period was, which tells you whether your result generalizes or was regime-specific.
- The goal is not a bot that wins everywhere. It is knowing which regime yours needs and not running it in the one that kills it.
This pairs with walk-forward validation: regime analysis shows where a strategy works, walk-forward shows whether that edge holds up out-of-sample.
Market regimes shift without warning, and past volatility does not predict future volatility. The examples here are illustrative. Crypto trading can lose money. Read the Risk Disclosure before going live.
What a Volatility Regime Is
Volatility is how much price moves around, and Freya measures it with ATR, the Average True Range, a standard gauge of a market's typical move size. Using ATR, the engine classifies the market at the time of each trade into three regimes:
- Low volatility: calm, quiet, often ranging or slowly drifting. Small candles, tight moves.
- Medium volatility: normal, active trading conditions.
- High volatility: violent, fast-moving markets, often strong trends, breakouts, or crashes. Large candles, big swings.
The same strategy behaves completely differently across these. A mean-reversion or grid bot loves low volatility, where price oscillates in a range it can harvest, and hates high volatility, where a trend blows through its levels. A breakout or trend bot is the opposite: it needs the big moves of high volatility and bleeds in the chop of a quiet market.
Reading the Regime Breakdown
For each of the three regimes, Freya reports the same set of numbers, so you can compare them directly:
- Trade count: how many of your trades happened in that regime. If almost all of them were in one regime, your backtest mostly tested that condition and says little about the others.
- Win rate and average PnL: how well you did in that regime. This is where specialization shows up.
- Total PnL: how much of your overall profit came from that regime. A strategy whose entire profit came from one regime is fragile if that regime is rare.
- Average duration: how long trades lasted in that regime, which often reveals whether the strategy was riding moves or getting chopped in and out.
And one summary number ties it together: your best-performing regime, the market character in which the strategy did best.
| Regime pattern | What it means | How to act |
|---|---|---|
| Strong in low, weak in high | A range or mean-reversion edge that trends break | Pause or reduce size when volatility spikes |
| Strong in high, weak in low | A trend or breakout edge that chop erodes | Avoid running it in quiet, ranging markets |
| Positive across all three | A genuinely robust, adaptable strategy | Rare and valuable, if it holds up out-of-sample |
| All profit from one regime, few trades elsewhere | The backtest only really tested one condition | Test across a period with more of the other regimes |
The Sample Trap
A backtest run over a single quiet stretch will show almost all its trades in the low-volatility regime and tell you nothing about how the strategy handles a spike. This is why the ATR percentiles, the 25th, 50th, and 75th percentile of volatility over the period, matter. They tell you how volatile the backtest window actually was.
If the whole period was calm, a great result is only evidence that the strategy works in calm markets. To trust it more broadly, you need a backtest that spans different conditions: a quiet range, a strong trend, and a violent selloff. A strategy that stays positive, or at least protects capital, across all three is one you can lean on. A strategy that only ever traded one regime has simply not been tested against the others yet.
Putting It to Work
- Find your best regime and respect it. If your strategy makes its money in low volatility, running it into a high-volatility breakout is choosing to trade in the conditions it loses in.
- Do not demand a bot that wins everywhere. Specialization is fine and often stronger. Knowing the specialization is what protects you.
- Check where the profit came from. If total PnL is concentrated in one regime that rarely occurs, the strategy is fragile even if the headline looks good.
- Match the market, or step aside. The most reliable use of regime analysis is defensive: recognize when the market has moved into your strategy's weak regime and reduce size or pause, rather than feeding it into conditions the backtest already showed it cannot handle.
- Validate across regimes. A single-regime backtest is a partial test. Confirm the edge on a period rich in the other conditions, and with walk-forward analysis, before trusting it.
Frequently Asked Questions
How does Freya classify market volatility?
It uses ATR, the Average True Range, a standard measure of how large the market's typical price moves are. Based on ATR, the engine sorts the market at the time of each trade into low, medium, or high volatility, then reports your win rate, average PnL, total PnL, trade count, and duration in each, plus your best-performing regime.
Why does my bot win in some market conditions and lose in others?
Because most strategies have an edge that depends on market character. Range and mean-reversion strategies profit from the oscillation of low-volatility markets and get run over by trends. Breakout and trend strategies need the big moves of high volatility and get chopped up in quiet markets. This is normal specialization, not a flaw, as long as you know which regime yours needs.
Should I only trust a bot that is profitable in every regime?
Not necessarily. A strategy that is strongly profitable in one regime and simply avoids or protects capital in others can be excellent, if you run it in the right conditions. What you should not trust is a backtest that only ever traded one regime, because it has not been tested against the others. Aim for a backtest that spans calm, trending, and volatile periods.
What are the ATR percentiles for?
They describe how volatile the backtest period itself was, at the 25th, 50th, and 75th percentiles. If those values are all low, your backtest was run in a calm market and your result mainly proves the strategy works in calm markets. High percentiles mean the period included violent conditions. They help you judge whether a good result generalizes or was specific to one kind of market.