Regime Trading: Why One Strategy Can't Cover All Markets

The one-strategy trap

Most bot platforms give you a strategy and tell you to pick a pair. SMA Cross on BTC. Bollinger Bands on ETH. Set it up, let it run, hope for the best.

The problem: markets change. BTC can trend hard for three weeks, chop sideways for two, then break out again. A trend-following strategy prints money during the trend and bleeds during the chop. A mean-reversion strategy does the opposite.

Running one strategy through all conditions is like wearing a winter coat in July. It's not that the coat is bad — it's that conditions changed.

ADX: the regime classifier

The Average Directional Index (ADX) doesn't tell you which way the market is moving. It tells you how strongly it's moving. That's a crucial distinction:

• ADX > 25 — strong trend. Doesn't matter if it's up or down. The market has directional momentum.
• ADX 20–25 — transition zone. The old regime is fading or a new one is forming. Volatility is compressing.
• ADX < 20 — ranging. Price is oscillating without clear direction. Mean reversion territory.

We compute ADX on the 4-hour timeframe using a 14-period lookback. This gives us a smooth, reliable read on the current market regime — not too noisy, not too lagged.

One strategy per regime

Instead of forcing one strategy to work in all conditions, we built three:

Notice the overlap: SMA Cross fires at ADX > 15, Trailing at ADX > 20. This isn't a bug. Both strategies have additional confluence gates — volume, momentum, trend alignment — that prevent them from firing on the same signal. The ADX gate is a minimum, not the only filter.

How the dispatcher works

Every scan cycle, the bot evaluates all active strategies in order. Each strategy checks its own preconditions — ADX gate, confluence score, specific signal requirements. The first strategy to hit its minimum score threshold fires.

This means in a strong trend (ADX 30+), both SMA Cross and Trailing could qualify. But only one position per pair is allowed. The first to reach confluence wins. In practice, SMA Cross fires first during sustained trends (it needs fewer confirming signals), while Trailing catches the initial breakout from compression (it needs the Keltner squeeze).

There's no manual switching, no parameter adjustment, no "I think the market is about to change." The regime detection is built into each strategy's entry conditions. The market tells the bot which strategy to run.

What the numbers say

We backtested each strategy in isolation across BTC, ETH, SOL, and BNB over a full year:

SMA Cross is the workhorse: positive on all four pairs, consistent 1.58–1.95 profit factor. Trailing Breakout is the specialist: massive when it hits (ETH +49%), but lower win rate means some pairs underperform.

Together, they cover the trend and breakout regimes. In strong trends, SMA Cross carries. During breakout transitions, Trailing catches the move that SMA Cross hasn't confirmed yet.

Why not just run everything?

We have a third strategy — Ranging — that covers the low-ADX regime. We built it, backtested it, and it was negative on every pair. Crypto trends too hard for mean reversion to work reliably. We shelved it.

The honest answer: regime trading only works if each strategy actually has an edge in its regime. Adding a strategy that doesn't have an edge just adds losses. Two profitable strategies covering two regimes beats three strategies where one bleeds.

The goal isn't to trade every market condition. It's to trade only the conditions where you have an edge — and sit out the rest.