Parabolic SAR: Read the Flip, Acceleration Factor, and Stop

Parabolic SAR means Stop And Reverse. In a rising state its dots normally sit under price; in a falling state they sit above. That side shows which state the chosen calculation currently assigns. TradingView describes the same display convention and treats a price cross as a potential reversal event {source:tradingViewParabolicSar}. It does not measure trend strength, predict how far price will travel, or prove that a new trend has begun. A high, a close, and a venue-triggered stop are different observations, so decide which one your rule uses.

The recurrence is SAR[next] = SAR[current] + AF[current] × (EP[current] − SAR[current]). EP is the highest high recorded in the active rising state or the lowest low in the falling state. AF is the acceleration factor. The projected value moves partway from the previous SAR toward that state’s extreme point; it is a price level generated from past state, not a moving average or a price target. TA-Lib documents this recurrence and identifies SAR as path-dependent {source:taLibParabolicSar}. The selected lookback and initial state matter because every later projection depends on earlier values.

AF starts at a configured value. In an up-state, each new high that extends EP adds the configured increment; in a down-state, a new low does the same. AF stops increasing at its maximum, and a bar that does not extend EP leaves AF unchanged. Common examples use a 0.02 start and increment with a 0.20 maximum, but those are settings, not universal constants. MQL5 exposes the step and maximum as inputs and documents those usual values {source:mql5ParabolicSar}. A larger step or cap draws the projected level closer to price sooner. That can create earlier state changes as well as tighter trailing levels; it does not mean the signal is more accurate.

The projected curve must respect nearby bars. In the rising state, the SAR is kept at or below the prior bars’ lows; in the falling state, it is kept at or above the prior bars’ highs. If price penetrates the active SAR, the state switches. The reversal SAR is reset to the previous state’s EP, AF returns to its starting value, and the new state begins tracking its opposite extreme. TradingView describes the two-bar range constraint and reset, while TA-Lib’s implementation specifies its own penetration inequalities and update order {source:tradingViewParabolicSar}. Those edge details matter when a low or high exactly touches the line, so use one documented convention throughout a calculation.

Suppose an up-state has SAR 96.00, EP 108.00, and AF 0.08. Its next raw projection is 96 + 0.08 × (108 − 96) = 96.96. If the move later makes a new high at 110 and the configured increment is 0.02, AF becomes 0.10 for a subsequent projection: 96.96 + 0.10 × (110 − 96.96) = 98.264, or about 98.26 before range constraints. This arithmetic only illustrates the recurrence. The bar’s earlier lows may cap that candidate, and a later penetration may cause a reset. The calculation says nothing about a fill or future return.

A SAR touch can be a backtest signal, an exit condition, a stop trigger, or a stop-and-reverse rule. Those are different strategies. State whether a high or low can trigger intrabar, whether a close must confirm the event, and whether a reverse position is opened after an exit. OHLC bars do not show the order of every trade inside the candle. A gap can open beyond the plotted level, and a stop order’s execution depends on venue rules and available liquidity. Do not record the SAR value as if it were a guaranteed execution price; model the next executable price and costs.

Wilder’s original description leaves room for different initialization choices. Implementations can infer the first direction differently, seed the first SAR with different bars, or begin output after different warm-up histories. TA-Lib explicitly notes that the algorithm is path-dependent and that bootstrap choices can vary across software {source:taLibParabolicSar}. SAR normally uses highs and lows, so session boundaries, missing bars, adjusted prices, and the selected timeframe change its path. An unfinished candle can also change its high and low. Record the symbol, venue, timeframe, session, start/increment/maximum AF, initialization rule, history length, and whether signals use intrabar values or completed bars.

Write down the entry, active SAR state, price field that triggers a flip, order timing, position reversal policy, invalidation, and size rule before reviewing results. Compare the SAR rule with a baseline using the same data, dates, execution assumptions, and spread, commission, slippage, and gap treatment. Keep a chronological holdout or walk-forward validation; do not choose AF values after seeing which curve best fits the full chart. For related mechanics, see ADX strength and direction, Donchian breakout timing, and ATR stop distance. Historical behavior depends on the chosen instrument, settings, sample, and costs. It does not establish future performance.

• Walk-forward validation
• ADX strength and direction
• Donchian breakout timing
• ATR stop distance