Most oscillators we use on Indian charts have soft, rounded tops. By the time the curve visibly rolls over, the stock has often already moved a fair distance. The Fisher Transform indicator, built by trader and engineer John Ehlers, attacks that problem in an unusual way. It doesn’t add more smoothing. It changes the shape of the data so that extreme readings stand out as rare, sharp spikes, and turning points show up as clean V-shapes rather than lazy hills.
That sounds great on paper. In practice it’s a twitchy tool that needs a firm hand. Below we go through where the idea comes from, Ehlers’ actual calculation, a bar-by-bar worked example, and the habits that keep it from flooding you with bad signals.
In this article
Where the Fisher Transform Indicator Comes From
The maths is borrowed from statistics. The Fisher transformation was originally applied to the Pearson correlation coefficient. When a sample correlation sits close to +1 or -1, its distribution is badly skewed, which makes confidence intervals awkward. Taking the inverse hyperbolic tangent (artanh) of it gives a variable that is approximately normally distributed.
Ehlers saw a parallel in price data. In his paper Using the Fisher Transform, he argues that the common assumption that prices follow a bell curve is “a really bad assumption” and one reason many indicators disappoint. He points out that a sine wave spends most of its time near its highs and lows, so its probability distribution looks nothing like a bell curve. Cycles behave the same way, which is why they are hard to trade.
His fix: squeeze price into a range between -1 and +1, then run it through the Fisher formula. The output has a nearly Gaussian distribution, which means big readings become genuinely rare events. And a rare event is easier to spot than a reading that looks like every other reading.
The Formula, Without the Fog
The core equation is short:
y = 0.5 × ln[(1 + x) / (1 – x)]
Here x is the input, which must lie strictly between -1 and +1, y is the output, and ln is the natural logarithm. That’s the same artanh function the statisticians use.
What matters to a trader is how this function bends. Ehlers notes that near the middle of the range the gain is roughly one, so small inputs come out almost unchanged. Near the edges the output gets amplified hard. We ran the numbers to show it:
| Input x | Fisher output y |
|---|---|
| 0.10 | 0.100 |
| 0.30 | 0.310 |
| 0.50 | 0.549 |
| 0.75 | 0.973 |
| 0.90 | 1.472 |
| 0.95 | 1.832 |
| 0.99 | 2.647 |
| 0.999 | 3.800 |
Going from 0.1 to 0.3 barely changes anything. Going from 0.9 to 0.99 nearly doubles the output. That stretching at the edges is the whole trick. When price is pinned near the top of its recent range, the indicator doesn’t drift sideways near a ceiling the way a bounded oscillator does. It shoots up.
The Fisher Transform doesn’t predict anything new. It just refuses to let an extreme reading hide among ordinary ones.
How Ehlers Actually Calculates It on a Chart
The formula alone isn’t the indicator. The EasyLanguage code in Ehlers’ paper adds a few steps, and these details explain most of how the line behaves:
- Price input. He uses the bar’s midpoint, (High + Low) / 2, rather than the close.
- Lookback channel. The default length is 10 bars. The code finds the highest and lowest midpoint over that window.
- Normalise. Price’s position in the channel is worked out as (Price – Low) / (High – Low), which runs from 0 to 1. Subtract 0.5 and double it, and you get a raw value between -1 and +1.
- Smooth. The normalised value is smoothed with an EMA-style step: 0.33 × today’s raw value + 0.67 × yesterday’s smoothed value.
- Clamp. Because smoothing can push the value past the channel limits, the code caps it: above 0.99 it is set to 0.999, below -0.99 it is set to -0.999. Without this, the logarithm would blow up at exactly 1.
- Transform and smooth again. Fish = 0.5 × ln[(1 + Value) / (1 – Value)] + 0.5 × yesterday’s Fish.
- Trigger. The second plotted line is simply Fish delayed by one bar. Crossovers between the two are the basic signal.
Two things jump out. First, there are two layers of smoothing, one on the normalised price and one on the Fish value itself. Second, the trigger is not a separate moving average in Ehlers’ original; it’s yesterday’s value. Some charting platforms describe the trigger as a moving average of the indicator, as Investopedia and TradingView’s help pages do, so check what your software plots before you compare signals with someone else’s screen.
A Worked Example, Bar by Bar
Let’s run four bars of a hypothetical NSE stock. These are made-up numbers, picked so the arithmetic is easy to follow. We’ll use Ehlers’ settings and assume the previous bar left us with a smoothed value of 0.40 and a Fish of 0.60.
Bar 1. The 10-bar channel runs from ₹480 to ₹520 and today’s midpoint is ₹515. Position in the channel is (515 – 480) / 40 = 0.875. Centre and double it: 2 × (0.875 – 0.5) = 0.75. Smooth it: 0.33 × 0.75 + 0.67 × 0.40 = 0.5155. The Fisher of 0.5155 is 0.570. Add half of yesterday’s Fish (0.30) and Fish = 0.870. The trigger is yesterday’s 0.60, so Fish is above it.
Bar 2. Price pushes to a new channel high: midpoint ₹521, so the channel is now ₹480 to ₹521. Raw value = 2 × ((521 – 480) / 41 – 0.5) = 1.000. Smoothed = 0.33 × 1.000 + 0.67 × 0.5155 = 0.675. Fisher of that is 0.821, plus half of 0.870, gives Fish = 1.256. Trigger is 0.870. Still rising.
Bar 3. Now the stock slips. Midpoint ₹509, same channel. Raw value falls to 0.415, but the smoothed value only eases to 0.589 because 67% of yesterday is still in it. Fisher of 0.589 is 0.677, plus half of 1.256, gives Fish = 1.305. Here’s the catch: price fell, yet Fish ticked up, and it is still above its trigger of 1.256. The double smoothing is doing its job, and costing us a bar.
Bar 4. Midpoint drops to ₹495. Raw value turns negative at -0.268. Smoothed = 0.306, Fisher of that is 0.317, plus half of 1.305, gives Fish = 0.969. The trigger is 1.305. Fish has crossed below it. That’s the sell-side signal in Ehlers’ setup.
| Bar | Midpoint (₹) | Raw value | Smoothed | Fish | Trigger |
|---|---|---|---|---|---|
| 1 | 515 | 0.750 | 0.516 | 0.870 | 0.600 |
| 2 | 521 | 1.000 | 0.675 | 1.256 | 0.870 |
| 3 | 509 | 0.415 | 0.589 | 1.305 | 1.256 |
| 4 | 495 | -0.268 | 0.306 | 0.969 | 1.305 |
Notice what the cross told us and when. The stock peaked on bar 2. The signal came on bar 4, by which point the midpoint had already given up ₹26 from that peak. Faster than many oscillators? Often, yes, which is Ehlers’ claim. Instant? No. Anyone who sells the Fisher Transform as a zero-lag reversal detector hasn’t run the numbers.
Reading the Signals
Traders tend to use the Fisher Transform indicator in three ways.
1. Trigger-line crossovers
The simplest approach, and the one Ehlers plots. Fish crossing above its trigger after a low reading suggests an upturn; crossing below after a high reading suggests a downturn. Investopedia gives the example of a cross below the signal line after an extreme high being used to exit a long position.
2. Extremes followed by a turn
Because the indicator is unbounded, there’s no fixed overbought line like 70 on RSI. Investopedia notes that what counts as extreme depends on the asset’s own history: for one asset a high reading might be seven or eight while a low might be -4, and other assets will look different. An extreme reading on its own is only a warning. The confirmation is the line changing direction.
3. Rate-of-change crossings
In the same paper Ehlers takes the sharp turns further. Since the steepest moves happen at turning points, he multiplied the Fisher Transform’s rate of change by 10 (matching his 10-bar channel) and plotted it over the Fisher line. He reports that the crossings of the two identified the major turning points on his example chart. It’s an interesting idea, but it’s one chart in one paper, not a tested trading system, so treat it as something to research rather than a rule.
An unbounded indicator has no fixed overbought line. Your only reference for “extreme” is the same stock’s own history.
Where the Fisher Transform Goes Wrong
We’d be doing you a disservice if we stopped at the good parts. The weaknesses are real and they show up fast on Indian mid-caps and intraday charts.
- It’s noisy. Investopedia flags that the indicator can be noisy despite its aim of making turns clearer, and that it throws off many signals, plenty of which aren’t profitable. With a 10-bar channel on a 5-minute chart, you’ll see crossovers all session long.
- Extremes don’t guarantee reversals. Price can go sideways or reverse only a little after a peak reading. A stock in a strong trend can keep printing new channel highs, keeping the input near +1 for days.
- “Extreme” keeps moving. A level that was a reliable ceiling for years can start showing up regularly. Levels you picked from last year’s chart may not hold this year.
- Prices still aren’t normal. The transform makes the indicator’s output closer to a bell curve. It doesn’t make the market behave. Investopedia makes the same point: asset prices aren’t normally distributed, so trying to normalise them won’t always give reliable signals.
- The channel resets. When a big bar drops out of the 10-bar window, the high or low of the channel jumps, and the normalised value can jump with it even if price barely moved. That can produce a crossover that has nothing to do with today’s trading.
Making It Usable on NSE Charts
A few habits help a lot.
Trade with the trend, not against it. Investopedia suggests a common filter: in an uptrend, take only buy and sell (exit) signals, not short signals; in a downtrend, use it for shorts and for timing when to cover. A 50-day or 200-day moving average is an easy way to define that trend. Our guide to technical analysis for Indian stocks covers how to set that context first.
Pair it with a different kind of tool. The Fisher Transform measures where price sits in its recent range. Volatility tools like Bollinger Bands look at spread around an average instead. Both lean on distribution ideas, but they answer different questions, and agreement between them is more useful than either alone.
Test the length on your timeframe. Ehlers used 10 bars. Investopedia and TradingView use nine periods as an example. Neither is magic. A longer window cuts noise and adds lag; a shorter one does the reverse. Backtest on the stocks you actually trade.
Use it on other indicators carefully. Ehlers’ conclusion suggests applying the transform to normalised indicators like RSI or the Stochastic, and Investopedia mentions RSI and MACD as candidates. If you go that route, remember you’re stacking one set of assumptions on another. Compare the result with the plain MACD before trusting it, and see whether the sharper turns actually arrive earlier or just look more dramatic.
Size small until you know its rhythm. Because signals arrive often, a fixed stop and modest position size matter more here than with slower tools. One bad week of whipsaws can wipe out a month of good crossovers.
Quick Checklist
- Confirm the trend with something slower before acting on a crossover.
- Know whether your platform’s trigger is the previous bar’s Fish or a moving average.
- Judge “extreme” against the same stock’s past readings, not a fixed number.
- Wait for the direction change; an extreme reading alone is not a signal.
- Expect one or two bars of delay from the built-in smoothing, as in our example.
- Backtest your chosen length before trading it with real money.
The Fisher Transform is a clever piece of engineering. It turns blurry tops into sharp ones and makes extremes easy to see. What it can’t do is decide for you whether a sharp top is the real top. That part is still on us.
Disclaimer: This article is for educational purposes only and is not investment advice. The worked example uses hypothetical prices to explain the calculation and does not refer to any real stock. Indicator settings and signals described here are not recommendations to buy or sell any security. Please do your own research or consult a SEBI-registered investment adviser before making investment decisions.





Leave a Reply
Join the conversation
Commenting is open to registered readers. Sign up with just your email — no password needed, and it takes under a minute.
Sign up to comment Log in