Parabolic SAR: Settings, Strategy, and What Tests Reveal

Parabolic SAR places a trailing level below price during an upward leg and above price during a downward leg. SAR stands for Stop and Reverse. Its familiar dots give traders a simple way to discuss trend direction, exits, and potential reversals.

The settings deserve a closer look. Across 175,000 completed historical configuration evaluations, we varied starting acceleration, acceleration increment, maximum acceleration, and entry confirmation. One useful finding was easy to miss in a results table: raising the cap changed the timing of a few trades while leaving the execution count unchanged. Raising it further then produced exactly the same recorded trades.

How Parabolic SAR works

J. Welles Wilder introduced the Parabolic System in New Concepts in Technical Trading Systems, alongside other familiar technical tools. The publisher’s description lists the system among the book’s original concepts.

During an upward leg, SAR moves toward the highest high retained from that leg, called the extreme point, or EP. The core update is:

Next SAR = current SAR + AF × (EP − current SAR)

AF is the acceleration factor. If SAR is 90, EP is 100, and AF is 0.02, the next unbounded value is 90.20. If the extreme and AF remain unchanged, the following value is 90.396.

The complete calculation also bounds the level using recent candle ranges and resets its state on a reversal. For a downward leg, the retained extreme is the lowest low. TradingView’s SAR documentation explains these mechanics and its implementation choices.

What the .02 / .02 / .20 settings mean

The customary defaults contain three separate controls:

The factor does not increase automatically on every candle. Fresh extremes cause the increments. Even with no new extreme, an existing positive factor can keep moving SAR toward the retained extreme, subject to the range bounds.

The cap limits AF. It does not set a maximum price level for the SAR dots.

SAR can keep tightening without a fresh high

In this selected historical episode, the upward leg went 44 updates without a fresh extreme or an increase in AF. SAR nevertheless advanced toward the retained high. Inspect the same candle in the price and acceleration panels.

This is an indicator calculation, with no simulated trades. We selected the longest qualifying stretch by duration to make the mechanism visible. It does not predict the length of the next stretch.

The distinction helps explain the cap. Giving AF permission to rise above 0.20 has no effect while the calculation continues using a smaller factor. Even when the factor changes, the recent-range bound can restrict how far the final level moves.

How traders use the buy and sell signals

A simple SAR strategy treats a change from above price to below price as a bullish reversal, and the opposite change as bearish. TradingView’s SAR strategy describes a system that alternates between long and short positions.

Our executable comparison uses a long-or-cash adaptation. It buys after the required number of consecutive bullish SAR candles, then sells when the bullish state ends. Signals use completed candles, with market execution on the following eligible open. This differs from an intrabar order placed at a SAR stop and from an always-reversing long/short system.

The usual weakness remains relevant: repeated reversals can be costly in a choppy range. A directional level alone does not establish trend strength. A separate trend-strength filter is a familiar extension; the experiments here isolate SAR’s own controls and entry confirmation.

A higher cap can change timing without adding trades

We held starting acceleration and increment at 0.02, with immediate bullish-state entry. In one matched historical comparison, raising the cap from 0.20 to 0.25 left the total at 462 executions. One execution means one buy or one sell.

The returns still changed: −31.79% versus −27.33%. Both results were negative.

We inspected the complete recorded trade lists. Only five execution locations differed when matched by their position in the list. The first changed purchase arrived two candles earlier with the higher cap. A strategy can keep the same number of buys and sells while buying or selling at different moments.

In the same sample, caps 0.25, 0.30, 0.40, and 0.50 shared the displayed curve values. A focused repeat at 0.25 and 0.50 went further: all 462 recorded execution rows matched in every displayed cell.

The broader grid also contained many outcome plateaus. Raising only the cap from 0.40 to 0.50 kept all seven exported summary fields identical in 80% to 100% of matched settings within each historical sample. The exact field list and comparison counts are in the methods below.

These are conditional plateaus in recorded outcomes. They do not establish that 0.25 is a universally sufficient cap, or that indicator values were identical on every candle. Matching summaries alone also do not prove matching trade lists; the complete trace check above applies to that one fixed configuration. The full parameter comparisons are available in the evidence download.

Confirmation reduces opportunities and changes the price paid

Waiting for an extra bullish candle rejects short bullish episodes and delays surviving entries. That can reduce activity while worsening the outcome.

For example, at the familiar defaults in one historical sample, increasing confirmation from one to two bullish candles reduced executions from 963 to 921. Recorded return changed from −80.03% to −81.47%. Fewer transactions did not rescue that example.

Across the tested parameter comparisons, confirmation helped some settings and hurt others. Its benefit must cover the cost of entering later. Counting the signals it removes is only part of that assessment.

Is there a best Parabolic SAR setting?

The defaults are a useful reference. Our tests do not justify a single setting for every trading environment.

Read the controls together. A low increment may leave a large cap unused; a faster increment can reach it sooner. Once an earlier reversal resets the calculation, the future extreme and acceleration history can also differ. Increasing a number therefore does not provide a simple ranking of complete trading outcomes.

A useful comparison holds the other controls fixed, checks nearby settings, and inspects recorded trades when summary metrics conceal a change. A plateau can save needless fine tuning; a sharp isolated winner calls for more scrutiny.

For related entry and exit questions, compare the Alligator strategy and Turtle channel strategy. To repeat these SAR comparisons, open TrainBard with the downloadable script and hold the other controls fixed while changing one setting.

Copy the strategy script

Copy this tested long-or-cash adaptation into your editor. It starts with the familiar 0.02 / 0.02 / 0.20 settings and one bullish candle for entry. Set the study’s combined trading cost to 0.1% per execution in your backtest settings. The tested SAR calculation convention is documented below.

//@version=6
strategy("Parabolic SAR Lab: acceleration and confirmation", overlay=true, initial_capital=10000, default_qty_type=strategy.percent_of_equity, default_qty_value=99, process_orders_on_close=false)

sar_start = input.float(0.02, "Starting acceleration", minval=0.002, maxval=0.05, step=0.002)
sar_step = input.float(0.02, "Acceleration increment", minval=0.002, maxval=0.05, step=0.002)
sar_max = input.float(0.2, "Acceleration cap", minval=0.1, maxval=0.5, step=0.05)
confirm_bars = input.int(1, "Consecutive bullish bars", minval=1, maxval=5)

sar_value = ta.sar(sar_start, sar_step, sar_max)
bullish = close > sar_value
bearish = close <= sar_value
bullish_age = ta.barssince(bearish)
valid = sar_start < sar_max and sar_step <= sar_max and not na(sar_value) and bar_index >= 500
enter_long = valid and bullish and bullish_age >= confirm_bars
exit_long = valid and not bullish

plot(sar_value, "Parabolic SAR", color=color.blue)

if exit_long and strategy.position_size > 0
    strategy.close("Long")
if enter_long and not exit_long and strategy.position_size == 0
    strategy.entry("Long", strategy.long)