PSAR Tight
Parabolic Stop and Reverse (Welles Wilder, 1978). A trend-following indicator that places trailing dots that accelerate toward price, eventually triggering a reversal.
How it works
PSAR places a series of dots that track price from behind:
- In an uptrend, dots sit below the candles and rise with each bar.
- In a downtrend, dots sit above the candles and fall with each bar.
- When price touches the dot, the trend reverses — the dot jumps to the other side.
The speed at which dots approach price is controlled by an Acceleration Factor (AF):
SAR(next) = SAR(current) + AF * (Extreme Point - SAR(current))The AF starts at its initial value and increases each time a new extreme point (highest high in uptrend, lowest low in downtrend) is reached, up to a maximum. As AF grows, the dot accelerates toward price — creating a natural tightening stop.
Tight variant parameters
| Parameter | Standard PSAR | PSAR Tight |
|---|---|---|
| AF start | 0.02 | 0.02–0.03 |
| AF step | 0.02 | 0.02–0.03 |
| AF max | 0.20 | 0.20–0.30 |
Higher acceleration produces earlier exits from losing trends, more frequent reversals, and tighter stops once in a trend. These characteristics suit crypto's volatile, momentum-driven price action.
Performance
| Metric | Value |
|---|---|
| Validated timeframes | W1, D1 |
| Sharpe (SOL) | 1.43 |
| Sharpe (BTC) | 1.08 |
| Sharpe (ETH) | 0.95 |
| Mean Sharpe range | 0.69–1.43 |
Performance varies significantly across assets. SOL's strong directional moves are a natural fit. Assets with more choppy price action see lower Sharpe ratios.
Natural stop loss
Like Supertrend, PSAR has a built-in stop level: the current SAR value. In a long position, the SAR dot below price is the invalidation point.
Quantity = (Account * Risk%) / |Entry - SAR Value|The stop tightens automatically as AF accelerates. No manual trailing stop logic is needed.
When it works well
- Strong, sustained trends: the accelerating dot rides the trend and only exits when momentum fades.
- Weekly and daily timeframes: W1 and D1 filter noise effectively.
- Volatile assets with directional bias: SOL and BTC during trending phases are ideal candidates.
When it struggles
- Sideways markets: PSAR reverses constantly, generating many small losses. Higher acceleration makes this worse.
- Lower timeframes: below D1, frequency of false reversals increases. Not recommended below 8h.
- Gradual trend changes: PSAR is binary (long or short). Slow rotations produce whipsaw exits.
Implementation
Source: gordon-strategy/src/engines/psar/mod.rs
The algorithm is byte-equivalent to the archived monolith (gordon-trading/crates/gordon-indicators/src/psar.rs). Parity fixtures live in src/engines/psar/parity_fixture.rs.
Related
- Supertrend — similar timeframe range (D1–8h), different mechanism (ATR bands vs accelerating dots)
- Adding a Strategy — promotion flow from implementation to production