Look at any indicator that redraws its trend line and you will notice it treats two very different markets the same way. A clean rally that walks almost straight up the chart and a nervous range that jitters sideways both get smoothed by the same fixed setting. So the line lags badly on the move you wanted to catch and flips constantly on the noise you wanted to ignore. What if the average could tell those two shapes apart on its own?
That is the idea behind the fractal adaptive moving average. By the end of this guide you will know how FRAMA reads the shape of price, what its settings do, how it differs from the Kaufman average, and whether it repaints.
Key Findings
- FRAMA adapts to the shape of price: it speeds up when the price path is smooth and slows down when the path is rough, so a single line handles trend and chop.
- Its dial is the fractal dimension: a near-straight move reads as low roughness and turns the smoothing up; a jagged, doubling-back move reads as high roughness and turns it down.
- It differs from the Kaufman average by what it measures: FRAMA judges the geometry of the curve, KAMA judges how efficiently price travelled.
- It does not repaint history: closed-candle values are locked; only the point on the live bar keeps moving until that candle closes.
What is the fractal adaptive moving average?
The fractal adaptive moving average is a moving average that sets its own smoothing speed by measuring how rough recent price action looks. The engineer John Ehlers described it in a 2005 write-up on his MESA Software site , and his aim was blunt: build one average that reacts fast when price is genuinely trending and stays calm when it is only shuffling around.
Every moving average has a smoothing setting that decides how tightly it follows price. On an ordinary average that setting is fixed. FRAMA keeps changing it, bar by bar, based on one reading of the current market. When that reading says the market is moving in a clean line, FRAMA tightens up against price. When it says the market is a mess of reversals, FRAMA loosens off and flattens out.
The reading it uses is what makes FRAMA distinctive. Instead of watching price levels, it watches the shape of the price path and asks a simple question: is this closer to a straight line or to a scribble?
What does “fractal dimension” actually mean for a price chart?
Fractal dimension is a way to score how much space a wiggly line fills. The concept comes from the mathematician Benoît Mandelbrot, whose 1982 book The Fractal Geometry of Nature popularised the idea that a rough coastline sits somewhere between a one-dimensional line and a two-dimensional plane. The rougher the coast, the higher its dimension.
Price does the same thing. A market that runs almost straight for a stretch barely fills any vertical space, so its dimension sits close to 1, the value of a plain line. A market that chops violently up and down covers a lot of vertical ground while going nowhere sideways, so its dimension climbs toward 2, the value of a filled plane. Everything real lands between those two.
FRAMA turns that single score into a speed. Here is the mechanical part, in plain terms. It splits its lookback window into two halves and compares how much price travelled in each half against how much it travelled across the whole window. A smooth move gets from one end to the other without wasted motion, which produces a low dimension. A choppy move backtracks constantly, filling space, which produces a high dimension. Low dimension turns the smoothing up toward fast; high dimension turns it down toward slow.
You never have to calculate any of this by hand. But knowing that the dial is the geometry of the curve, not the raw price, tells you exactly when FRAMA will engage and when it will step back.
FRAMA vs KAMA: two ways to build an adaptive average
Both averages solve the same complaint about a fixed-speed line, and on many charts they look almost interchangeable. The interesting part is where they part ways, because they are reading different signals to make the same decision.
The Kaufman adaptive moving average measures directional efficiency: how much net ground price covered against how far it actually travelled to get there. FRAMA measures geometric roughness: how much space the price curve fills. Those usually agree, but not always. A market can trend directionally while still whipping around a lot internally, and the two will grade that situation differently.
| Adaptive average | What it measures | Reads market as fast when | Tends to suit |
|---|---|---|---|
| Kaufman (KAMA) | Directional efficiency of travel | Price covers net ground with little backtracking | Markets that alternate clean trends and flat ranges |
| Fractal (FRAMA) | Roughness of the price curve | The price path is close to a straight line | Markets where noise texture, not just direction, matters |
| Fixed EMA | Nothing, one speed always | Never adapts | Traders who want a predictable, unchanging line |
Neither adaptive average is the “right” one. If you already trust the efficiency-ratio idea, KAMA is the natural pick. If you think in terms of how jagged or clean the tape looks, FRAMA maps onto that instinct more directly. For a broader view of how much weight any single line deserves, our note on leading versus lagging indicators is worth reading before you build around one, and if your real goal is simply the earliest low-lag turn, the zero lag EMA chases that from a different angle entirely.
What settings does FRAMA use, and what do they change?
The main input is the lookback window, the number of candles FRAMA scores for roughness. Ehlers used 16 in his original description, and it has to be an even number because the method splits the window in half. That is the one setting that changes the character of the line.
A shorter window makes FRAMA jumpy: it declares a trend sooner and gets fooled more often by brief bursts of noise. A longer window makes it patient: steadier and harder to fake out, but slower to wake up when a real move begins. Some implementations add fast and slow bounds that cap how quick or how sluggish the smoothing is allowed to become, which is worth using if you find the line either too twitchy or too dead at the extremes.
RelicusRoad Pro
Have you been trading for a while but have never made consistent profits or are you new to FOREX trading and want to get a head start? Try RelicusRoad and you'll never look back.
Get RelicusRoad ProDoes the fractal adaptive moving average repaint?
Built on closed candles, FRAMA does not repaint. When a bar closes, the value it printed is fixed, and pulling up the chart a week later shows the same line you watched live. On that count it is as honest as any properly coded average.
What moves is the live bar. While the current candle is still open, its price keeps shifting, so FRAMA redraws its newest point on every tick. Because FRAMA can be running at its fast setting during a strong push, that last point sometimes travels a fair distance before the candle finally closes. None of that is repainting. The signal can keep sliding while the bar is unfinished; the version that counts is the one printed once the candle closes, because that one does not move again. If you want a repeatable way to prove a tool actually holds its reading after the close, the non-repaint screenshot test walks through it.
How do you actually trade with FRAMA?
The cleanest job for FRAMA is a trend filter, and it plays straight to its strength. When the line is sloping and price sits on one side of it, you have a trend worth taking signals in. When the line goes flat and price saws across it, FRAMA is telling you the market has turned rough, and the honest move is usually to stand aside rather than trade every cross.
That second half is the part most traders skip. A flat FRAMA is not a failed signal. It is a signal that says the tape is too noisy to trust a direction right now. Reading the flatness as useful information, instead of forcing entries anyway, is where the tool actually protects your account.
Stay honest about its limit, though. An adaptive average sharpens when you engage; it does nothing for the trade once you are in it. Position size, a stop set before you click, and the discipline to pass on a marginal setup are still what keep you solvent. FRAMA points at the right conditions. Your risk rules decide whether being right pays.
Where RelicusRoad Pro fits
The frustration underneath all of this is trusting the read. A line that swings hard mid-candle, a turn that looks decisive and then quietly unwinds by the close. RelicusRoad Pro is built to fix its signals at the candle close on MT4, MT5, and TradingView, so the read you act on is the one that stays on the chart tomorrow. It does not replace an adaptive filter like FRAMA; it removes the doubt about whether the signal in front of you is final. To see how a filtered average slots into a full entry system, our moving average crossover strategy guide covers the confirmation rules that stop a smooth line from trading you into every range.
Frequently asked questions
What is the fractal adaptive moving average?
The fractal adaptive moving average, or FRAMA, is a moving average that adjusts its own smoothing speed by measuring how rough or smooth recent price action is. John Ehlers published it in 2005. When price traces a fairly straight path, FRAMA reads that as a real trend and speeds up to track it closely. When price is jagged and full of reversals, FRAMA reads that as noise and slows down so it stops chasing every wobble. The measurement it uses for rough-versus-smooth is the fractal dimension of the price curve, which is where the name comes from.
How does FRAMA measure the fractal dimension of price?
It splits its lookback window into two halves and looks at how much ground price covered in each half compared with the whole window. A smooth, near-straight move covers its distance efficiently, which reads as a low fractal dimension, close to that of a simple line. A choppy stretch doubles back on itself constantly and fills far more vertical space for the same horizontal distance, which reads as a higher dimension, closer to that of a rough plane. FRAMA turns that single number into a smoothing setting: low dimension means fast, high dimension means slow.
What is the difference between FRAMA and KAMA?
Both are adaptive averages that speed up in trends and slow down in chop, but they decide when to do so from different signals. The Kaufman adaptive moving average measures directional efficiency, the ratio of net travel to total path length. FRAMA measures the geometric roughness of the price curve itself through its fractal dimension. In many conditions they behave alike, but they can disagree: a market can grind directionally with a lot of internal noise, and the two averages will read that situation differently. Neither reading is universally better; they are two answers to the same question.
What are good FRAMA settings?
The main input is the lookback window, which Ehlers set to 16 in his original description, and it must be an even number because the calculation splits it in half. A shorter window makes FRAMA quicker to declare a trend and quicker to be fooled by short bursts of noise; a longer one makes it steadier but slower to engage a fresh move. Some versions add fast and slow bounds that cap how responsive or how sluggish the line is allowed to get. The sensible approach is to run the defaults on your own pair and timeframe, watch the line through both a trend and a range, and change one input at a time only if it is misreading your market.
Does the fractal adaptive moving average repaint?
A FRAMA built on closed candles does not repaint its history. Once a bar closes, the value it printed is fixed and stays there when you scroll back later. The only part that moves is the newest point, on the live candle, because that bar’s price is still changing and the average recalculates its latest value on every tick. That movement is expected behaviour for any moving average, not repainting. The mistake is treating a sharp FRAMA turn mid-candle as settled when the bar can still close in a different place.
Want the turn to stop moving after you commit to it? RelicusRoad Pro locks its signal at the candle close on MT4, MT5, and TradingView, so you act on a read that will still be there tomorrow.