ATR Mean Reversion
Waits below the market for sell-offs that overshoot, buys the panic, and is usually out again by the close of the same day.
- Markets
- A broad set of liquid coins, futures
- Chart
- Daily candles
- Direction
- Long only
- Leverage
- 2x, isolated margin
- Positions
- Up to four, equal size
- Typical holding
- Less than a day
- Live since
- 24 February 2026
- Order not reached: after a weak day, a buy order waits below the low. Price turns before it gets there, so the order expires.
- Setup: another close below the previous day's low while the market is restless. A new order goes in, again below the low.
- Entry: the next day the sell-off overshoots and runs into the order.
- Exit: the day closes above the entry price, and the position is sold at the close.
- Daily ranges: gold while recent days swing more than usual. Only then are orders placed.
The idea
Crypto overreacts. On a day of heavy selling, liquidations and triggered stops push prices further than the news behind the move would justify, and part of that overshoot is often undone within hours. The strategy tries to be the buyer at the moment the sellers run out.
It does not chase. Instead of buying at the market, it places an order well below the day’s low, and that order only fills if the selling carries on into an extreme. The distance is measured in the coin’s own volatility, so a quiet coin and a wild one are treated the same way. Each trade is small and short; the edge comes from repeating the same pattern across many coins.
When it buys
An order is placed when two conditions meet at the daily close, and a trade only opens if the market then comes to that order:
A weak day
The coin closes at or below the previous day’s low. Sellers are in control.
A restless market
Recent daily swings are larger than the coin’s longer-term average. Calm drifts lower are ignored.
An order below the low
A limit order is placed below the day’s low, at a distance scaled to recent volatility. It is valid for one day; if price does not get there, it is cancelled.
Price reaches the order: the trade opens.
There is deliberately no trend filter. The strategy buys sharp dips in rising and falling markets alike; in testing, adding a trend filter made the results worse.
When it sells
First close in profit
If the day closes above the entry price, the position is sold at the close. Most trades end this way, on the same day they were bought.
No rebound
If the close stays weak compared with the coin’s short-term average, the position is sold as well. The bounce did not come, and the strategy does not wait for it.
Emergency stop
A wide stop is placed together with the order, so it is active from the moment of the fill. It is only meant for crashes that do not bounce at all.
A position is only held overnight when the close lands between the two exit rules, which is rare.
Risk
- Moderate leverage of 2x on futures. Every position uses isolated margin, so a single trade can never draw on the rest of the account.
- Capital is split into four equal slots, so a single coin is never more than a quarter of the strategy.
- Short holding times keep exposure low: most of the time the strategy holds nothing and waits.
- It buys into falling prices on purpose. When a sell-off turns into a crash, several positions can lose at once, and the emergency stop sits far below the entry.
Where it does well, and where it struggles
Does well in
- Volatile markets with sharp but short sell-offs, where panics are bought back quickly.
- Busy phases with many coins swinging at once, which produce more setups.
Struggles in
- Crashes that keep going for days, where there is no rebound to sell into.
- Calm markets: few setups, and orders below the low are rarely reached.
- Market-wide sell-offs, when all slots fill on the same day and the positions move together.
Backtest
How the strategy would have traded since 2018, with the same rules, limits and leverage as live.
Feb 2018 to Oct 2026. Coins: the 30 largest by market value at the start of each year. Up to four positions, 2x leverage. Starting balance $10,000, profits reinvested, fees included.
- Average return per year
- +47%
- Worst drawdown
- −70% −65% on closed trades
- Trades
- 600 About a day each, typically
- Winning trades
- 62% Profit factor 1.6
Including open trades at their lowest pointClosed trades only
| Year | Return | Worst drawdown |
|---|---|---|
| 2018 part | +52% | −18% |
| 2019 | +50% | −36% |
| 2020 | +25% | −70% |
| 2021 | +118% | −35% |
| 2022 | +1% | −39% |
| 2023 | +11% | −49% |
| 2024 | +100% | −30% |
| 2025 | +64% | −50% |
| 2026 to Oct | +24% | −35% |
In the backtest, 2 positions were liquidated, losing the margin set aside for them. Both curves include these losses.
Hypothetical results from a simulation on historical data, not live trading. Live results differ, and past performance is no guarantee of future results. How the backtests were made