Fisher Transform Extremes: Why Price Reversals Can Wait

Published November 9, 2026 · View Markdown

A Fisher Transform reading can look dramatic while price carries on in the same direction. A high value describes the indicator's present calculation. It does not put an appointment for a reversal on the market's calendar.

That distinction matters in TradingView automation. “Sell when extreme” and “sell when the indicator turns down after an extreme” produce different events. Before using the AlgoWay TradingView alert workflow, decide which event your alert represents. A connector cannot resolve a trading rule that has two meanings.

What the transformation actually magnifies

The mathematical transformation is F(x) = 0.5 × ln((1 + x) / (1 - x)), with x strictly between -1 and 1. TradingView's Fisher Transform explanation describes normalizing price over a lookback window and using the result to highlight potential turning points. Extreme readings depend on the instrument and its history; they are not universal reversal thresholds.

Here is an illustrative calculation of the transformation alone, not a reproduction of a platform's complete indicator:

• At x = 0, F(x) = 0.
• At x = 0.5, F(x) is approximately 0.5493.
• At x = 0.9, F(x) is approximately 1.4722.
• At x = 0.99, F(x) is approximately 2.6467.

The last input increases by only 0.09, yet its transformed value rises by about 1.17. The logarithm stretches values near the boundary. Negative inputs produce matching negative outputs. Neither the formula nor those calculations contain a future price.

Check the implementation before comparing readings

Normalization, smoothing and boundary handling belong to the calculation too. For example, the thinkorswim FisherTransform documentation describes range normalization, a five-period EMA and a restriction inside the open interval from -1 to 1. It also documents a trigger plot that is the Fisher line delayed by one bar. Do not assume a similarly named TradingView script uses precisely those choices.

A plotted value of 2 is not, by itself, a two-standard-deviation observation or a reversal probability. The formula changes the scale of an input. To make a statistical probability claim, you would need evidence about the resulting distribution and future outcomes, not merely the indicator's name.

Three rules that disagree on the same sequence

Consider hypothetical completed-bar Fisher values of 1.8, 2.2, 2.4 and 2.1. Use 2.0 only as an example threshold, not a recommended setting.

• Above 2.0: true on the second, third and fourth bars.
• Crosses above 2.0: true only on the second bar, because the previous value was at or below the threshold.
• Turns down while still above 2.0: true on the fourth bar, because 2.1 is below 2.4 but remains above 2.0.

With a one-bar-delayed trigger, the fourth bar also crosses below that trigger: the current line is 2.1 against 2.4, whereas on the preceding bar it was 2.4 against 2.2. This equivalence depends on the delayed-line definition. A moving-average trigger can cross on a different bar.

None of these invented indicator values tells us whether the next trade wins. They demonstrate event timing. The distinction resembles the separation between a Bollinger squeeze setup and its breakout trigger, although the underlying calculations are different.

Turn the observation into one auditable alert

Write down the script, input price, lookback, smoothing, trigger definition and threshold. Specify whether a confirmed close is required. TradingView's Pine alert documentation explains bar-close alert frequency and that running alerts retain a snapshot of the script and inputs. Recreate an alert when the settings it should use change.

Then state the action: open a short, close a long, or merely record a candidate. Those are separate instructions. Define whether another signal is allowed while a position remains open, and specify size and exits independently of the oscillator reading. Compare the recorded alert event with the intended completed bar before relying on the rule for execution.

The useful question is not “How high is too high?” It is “What exact change should cause this strategy to act, and what evidence supports that choice?”