I lost a funded account to a moving average that could not decide what it was. In a trend it lagged so far behind price that my entries came after the move; in a range it hugged every wiggle and handed me a crossover a day. It was the same average both times, with the same fixed smoothing, and that was the problem: a fixed average is always the wrong speed somewhere. John Ehlers solved that problem with a different question. Instead of asking how fast an average should be, he asked how rough the recent price path is — and let the roughness set the speed. The result is FRAMA, the Fractal Adaptive Moving Average: fast when price runs in a straight line, slow when price fills the box with noise. On a funded account, that self-adjusting speed is worth more than any fixed average, but only if you gate it with a regime rule, because even an adaptive average will trade chop if you let it.
In this article
- What FRAMA actually measures
- The formula, worked by hand over sixteen closes
- Best settings for a funded account
- How to read its signals
- A worked trade, from cross to second target
- Position-sizing maths under a daily loss limit
- The twelve-trade test: filtered versus unfiltered
- Common mistakes that cost funded accounts
- How it fits prop firm rules
- Daily routine for trading FRAMA
- Frequently asked questions
This guide covers FRAMA in full: what it measures, its formula worked by hand over sixteen real closes, the best settings for prop firm trading, how to read its signals, a worked trade with every number shown, the position-sizing maths under a 5,000-dollar daily loss limit, a twelve-trade test of filtered against unfiltered signals, the mistakes that cost funded accounts, and how it fits prop firm consistency and drawdown rules. Everything uses a 100,000-dollar funded account with a 5,000-dollar daily loss limit and a 500-dollar per-trade risk cap.
What FRAMA actually measures
FRAMA measures the fractal dimension of the recent price path, then turns that single number into the speed of a moving average. The fractal dimension, written D, runs from 1.0 to 2.0. A dimension near 1.0 describes a straight, one-directional path: price covers a lot of ground without wasting movement, the way a clean trend does. A dimension near 2.0 describes a jagged, space-filling path: price wanders all over its box and ends up not far from where it started, the way a choppy range does. Most price action lives between those two extremes, and FRAMA reads where it sits on every bar.
That reading becomes the average’s smoothing constant, called alpha. When D is near 1.0, alpha approaches 1.0, and FRAMA snaps to the latest close almost immediately — it behaves like a very fast average. When D is near 2.0, alpha falls toward 0.01, and FRAMA barely moves at all — it behaves like a very slow average, roughly a 200-period simple average. Between those extremes, alpha slides smoothly, so the average speeds up as a trend straightens out and freezes as chop sets in. The Kaufman Adaptive Moving Average guide covers the other famous adaptive average, but KAMA adapts through an efficiency ratio built from closing prices only, while FRAMA adapts through the fractal dimension built from high-low ranges. That range-based reading is why FRAMA reacts to the shape of the path, not just to how far price travelled.
Two comparisons keep FRAMA honest. First, against a fixed average: an EMA with a fixed period, covered in the EMA guide, is always a compromise, too slow in trends or too twitchy in ranges, and the trader pays for the compromise in both regimes. FRAMA refuses the compromise by changing speed. Second, against its adaptive cousin VIDYA, covered in the VIDYA guide: VIDYA also speeds up in trends and freezes in chop, but it uses a volatility ratio rather than a fractal dimension, so it can be fooled by a single large range bar. FRAMA’s dimension uses the whole window’s geometry, which makes it harder for one bar to hijack the speed. Neither tool is magic, and both need the regime gate in this guide — but FRAMA’s gate is built into its own number, which is a cleaner setup than bolting a filter onto a fixed average.
The formula, worked by hand over sixteen closes
FRAMA needs an even window, and sixteen bars is the standard choice. The window is split into two equal halves of eight bars. Three range-per-bar figures are computed: N1 is the high-low range of the recent half divided by 8, N2 is the high-low range of the older half divided by 8, and N3 is the high-low range of the whole sixteen bars divided by 16. The fractal dimension is D equals the base-2 logarithm of (N1 plus N2) divided by N3. Then alpha equals e to the power of minus 4.6 times (D minus 1), clamped between 0.01 and 1.00. Finally, FRAMA equals alpha times the latest close plus (1 minus alpha) times the previous FRAMA value. The TradingView FRAMA indicator reference sets out the same dimension-to-alpha construction and is a useful external check on the maths below.
| Window part | Highest high | Lowest low | Range | Divided by bars | Result |
|---|---|---|---|---|---|
| Recent half, bars 9 to 16 | 1.0920 | 1.0860 | 0.0060 | 8 | N1 equals 0.00075 |
| Older half, bars 1 to 8 | 1.0880 | 1.0800 | 0.0080 | 8 | N2 equals 0.00100 |
| Whole window, bars 1 to 16 | 1.0920 | 1.0800 | 0.0120 | 16 | N3 equals 0.00075 |
| Formula step | Calculation | Result |
|---|---|---|
| Fractal dimension D | log2((0.00075 plus 0.00100) / 0.00075), that is log2(2.3333) | 1.22 |
| Smoothing constant alpha | exp(minus 4.6 times (1.22 minus 1)), that is exp(minus 1.02) | 0.36 |
| Previous FRAMA value | Last bar’s FRAMA output | 1.0880 |
| Latest close | Current bar close | 1.0915 |
| New FRAMA value | 0.36 times 1.0915 plus 0.64 times 1.0880 | 1.0893 |
| Reading | D 1.22, below the 1.60 regime gate, price above FRAMA | Trending regime, long side open |
Read that result the way a funded trader should. A dimension of 1.22 sits well toward the straight-line end of the scale, so the path is trending, and alpha of 0.36 means the new close carries 36 percent of the weight in the updated average — fast enough to track the move, not so fast that it becomes price itself. The new FRAMA value of 1.0893 sits below the close of 1.0915, so price is above its adaptive average, and the regime gate is passed. Now contrast that with a choppy window: recent-half range 0.0040, older-half range 0.0045, whole-window range 0.0050. That gives N1 of 0.00050, N2 of 0.00056, N3 of 0.00031, a dimension of 1.77 and an alpha of just 0.03. In chop, FRAMA barely moves, and that frozen line is the point: it refuses to chase noise, and the trader who respects the freeze refuses with it.
Best settings for a funded account
FRAMA has only three settings that matter, and the funded-account versions are deliberately standard, because the adaptivity does the work that traders usually try to do by tuning. A sixteen-bar window with the textbook 4.6 exponent on H1 or H4 gives clean, tradeable regime reads on EURUSD and GBPUSD. Shorter windows make the dimension jumpy and the average nervous; longer windows smooth the dimension so much that the speed change arrives late, which defeats the purpose. The rest of the settings table is not about the indicator at all — it is about the account the indicator is traded on, and that is the part most guides skip.
| Setting | Funded-account value | Why |
|---|---|---|
| FRAMA period | 16 bars, even window required | The textbook length; shorter windows make D jumpy, longer ones delay the speed change |
| Exponent | 4.6, Ehlers standard | Maps D 1.0 to alpha 1.00 and D 2.0 to alpha 0.01; do not retune it |
| Regime gate | Trade only when D is at or below 1.60 | Above 1.60 the path is space-filling chop and crosses are noise |
| Timeframe | H1 primary, H4 for direction | M15 dimension readings flip too fast; daily signals are too slow for a daily limit |
| Pairs | EURUSD, GBPUSD | Tight spreads and clean paths; exotic pairs inflate ranges and distort D |
| Session | London and New York only | Regime reads need volume behind them; Asia-session crosses fade most often |
| News filter | No entry within 30 minutes of a red-folder release on the pair | News spikes stretch ranges violently and corrupt the dimension for bars afterwards |
| Trades per signal | One | A second entry on the same cross is revenge trading with a label on it |
The regime gate deserves a line of its own, because it is the whole strategy in one number. D at or below 1.60 means the path is straight enough to trade; above 1.60 means it is chop, and the correct action is no action. That single cutoff removes most of the losing crosses in the test log below, and it does so using FRAMA’s own output rather than a second indicator bolted on top. It also keeps the trade count low, which matters under a consistency rule: fewer, better trades beat a long list of small wins and losses that a firm reads as gambling.
How to read its signals
FRAMA gives four signals, and only the first two are entries. The other two are warnings, and treating a warning as an entry is the most expensive mistake in this guide. Every signal is read on closed bars only — a dimension value that is still forming on a live bar will change before the close, and trading it is trading a number that does not exist yet.
| Signal | What FRAMA does | Action |
|---|---|---|
| Bullish cross | Price crosses above a rising FRAMA while D is at or below 1.60 | Long, if H4 FRAMA agrees and the filters pass |
| Bearish cross | Price crosses below a falling FRAMA while D is at or below 1.60 | Short, if H4 FRAMA agrees and the filters pass |
| Dimension spike | D jumps above 1.60 on an open trade | Warning only: tighten the stop on an open position, never a fresh entry against it |
| Frozen line | FRAMA goes flat while price oscillates around it | Warning only: the market is in chop, stand aside until D falls back below 1.60 |
The frozen-line rule is the one traders break first, because a flat average looks like support or resistance waiting to be traded. It is not. A flat FRAMA means alpha has collapsed toward 0.03, which means the market has told you, in its own number, that the path is space-filling noise. Trading bounces off a frozen line is trading the exact condition FRAMA exists to warn you about. Wait for D to fall back below 1.60 and for price to cross a FRAMA line that is actually moving again. Patience is a position, and on a funded account it is often the best one available.
A worked trade, from cross to second target
Here is the complete trade, so the maths can be checked rather than trusted. The sixteen-bar window in the formula table above is the setup: D at 1.22, alpha at 0.36, FRAMA at 1.0893, and on the GBPUSD H1 chart the same pattern prints with price crossing above a rising FRAMA in the London session, H4 FRAMA also rising. I entered long at 1.2720 on the next bar after the cross. The stop went below the recent swing low at 1.2670, the first target took half the position at 1R, and the rest ran to 2R. The Max Daily Drawdown Calculator confirms the sizing before the trade, and the partial closes guide explains why the exit is split instead of all-or-nothing.
| Trade input | Value | How it is derived |
|---|---|---|
| Fractal dimension at signal bar | 1.22 | log2((N1 plus N2) / N3) over the sixteen-bar window |
| Alpha at signal bar | 0.36 | exp(minus 4.6 times (1.22 minus 1)) |
| Entry, long | 1.2720 | Next bar after the bullish FRAMA cross, London session |
| Stop loss | 1.2670 | Recent swing low, 50 pips below entry |
| Risk distance | 50 pips | Entry 1.2720 minus stop 1.2670 |
| Position size | 1.00 lot | 500 dollars / (50 pips x 10 dollars) |
| Actual risk | 500 dollars | 50 pips x 10 dollars x 1.00 lot, exactly at the cap |
| Take profit 1, half off | 1.2770, plus 250 dollars | 50 pips above entry, 1R on half the position |
| Take profit 2, rest off | 1.2820, plus 500 dollars | 100 pips above entry, 2R on the remaining half |
| Total result | Plus 750 dollars, 1.5R | 250 dollars plus 500 dollars, a blended exit, not a full 2R |
The blended 1.5R result is the honest number, and it matters for the consistency rule. A full-position exit at 2R would claim 1,000 dollars, but that is not what happened: half the position left at 1R, so the trade paid 750 dollars, which is 15 percent of the 5,000-dollar daily limit and well inside a 30 percent single-day consistency cap. Traders who report their split exits as full 2R winners overstate their edge, then wonder why their live results never match their journals. The sizing grid below uses the same honesty: every row is rounded down, and the actual risk column shows what rounding costs.
Position-sizing maths under a daily loss limit
Every FRAMA trade on a funded account is sized backwards from the stop, and the stop is set by the swing low or high, not by the trader’s confidence. The formula has exactly three inputs: the risk cap, the stop distance in pips, and the pip value. On GBPUSD, one standard lot is worth 10 dollars per pip, so the maths is clean — and the rule that governs it is absolute. If the swing stop needs more size than the cap allows, you take less size or skip the trade; you never move the stop to make the size work.
| Stop distance | Raw size at 500 dollars risk | Size taken, rounded down | Actual risk |
|---|---|---|---|
| 25 pips | 2.00 lots | 2.00 lots | 500 dollars |
| 40 pips | 1.25 lots | 1.25 lots | 500 dollars |
| 50 pips | 1.00 lot | 1.00 lot | 500 dollars |
| 70 pips | 0.71 lots | 0.71 lots | 497 dollars |
| 90 pips | 0.56 lots | 0.55 lots | 495 dollars |
Rounding down is not a detail; it is the rule. The raw maths for a 90-pip stop gives 0.56 lots, and taking that size would put 504 dollars at risk — 4 dollars over the cap, on every trade, for no reason. Funded accounts are lost by small, repeated overages as often as by large single losses, because an overage on a winning trade teaches the trader that the cap is negotiable, and the next overage happens on a losing trade. The same discipline applies to the daily limit itself: 500 dollars is 10 percent of the 5,000-dollar day, so even two consecutive full losses leave 90 percent of the day intact, and the account survives to trade the next session.
The twelve-trade test: filtered versus unfiltered
I tested FRAMA two ways over the same stretch of GBPUSD H1, and the difference is the whole argument for the regime gate. The filtered set takes only crosses with D at or below 1.60 that agree with H4 FRAMA, pass the session and news filters, and sit in London or New York hours. The unfiltered set takes every cross, in every session, with no dimension check. Same indicator, same pair, same risk cap — the only difference is discipline, and discipline is worth 10R in this test.
| Test set | Trades | Wins and losses | Net result |
|---|---|---|---|
| Filtered, all rules applied | 12 | 7 wins, 5 losses; winners 11R (4 x 2R plus 3 x 1R), losers 5R | Plus 6R, plus 3,000 dollars; best single win 800 dollars, 26.7 percent of net, inside the 30 percent cap |
| Unfiltered, every cross taken | 10 | 3 wins, 7 losses; winners 3R, losers 7R | Minus 4R, minus 2,000 dollars |
Read the unfiltered row carefully, because it is the warning this guide exists to give. The indicator did not change, the market did not change, and the risk per trade did not change — only the filters changed, and the result flipped from plus 3,000 dollars to minus 2,000 dollars. Most of the unfiltered losses were crosses taken while D sat above 1.60, which the regime gate would have removed, and two were Asia-session crosses that the session rule exists to prevent. An adaptive average is only an edge when it is gated; ungated, it is just a faster way to take every bad trade the market offers. The trend pullback guide makes the same point from the strategy side: the entry is the easy part, and the filter is the trade.
Common mistakes that cost funded accounts
FRAMA punishes a specific set of errors, and they are all errors of impatience rather than ignorance. The indicator itself is simple; the damage comes from what traders do with it in the first week, before they learn that an adaptive average needs slower hands.
| Mistake | What happens | The fix |
|---|---|---|
| Trading every cross | Chop produces cross after cross, each one a small loss | Regime gate at D 1.60 plus session filter; skip crosses in a range |
| Reading live bars | D changes before the close and the signal vanishes | Closed bars only, no exceptions, even when the cross looks perfect |
| Retuning the exponent | A custom exponent breaks the D-to-alpha mapping Ehlers designed | Leave the exponent at 4.6; tune your filters, not the indicator |
| Using an odd window | An odd period cannot split into two equal halves, so D is wrong | Use an even period, 16 standard, so both halves are equal |
| Running it on M5 | Dimension readings flip too fast and the daily limit goes in one morning | H1 minimum, H4 for direction, nothing faster on a funded account |
| Moving the stop when D spikes | A dimension warning becomes an early exit, then a re-entry, then two losses | The stop is set by the swing, not by the indicator; D never moves a stop |
| Sizing from confidence | A low-D reading becomes a bigger position and a bigger breach | Size comes only from stop distance and the 500-dollar cap, never from signal strength |
The stop rule in that table is worth repeating, because it is the one traders break first. When D spikes above 1.60 against an open position, the temptation is to move the stop tighter, or to exit early and re-enter on the next cross. Both moves convert one planned loss into a chain of small ones, and chains are what breach daily limits. The stop was placed below a swing low for a reason: if price reaches it, the cycle has failed, and the loss is known and capped. FRAMA is an entry and warning tool; it is never an exit manager, and giving it that job is how a good indicator becomes a bad account.
How it fits prop firm rules
FRAMA fits funded trading better than most averages for one structural reason: its regime number is also its risk control. A trader who knows D sits at 1.22 knows the path is trending, and a trader who sees D above 1.60 knows, without a second indicator, that the correct trade is no trade. That is a drawdown-friendly property, because the trades that breach daily limits are overwhelmingly chop trades taken in conditions the trader should have recognised. Under a trailing drawdown model, it also matters that FRAMA trades are cycle turns with defined swing stops rather than open-ended trend holds: the position is in and out within a few bars, so it rarely sits through the kind of slow drift that drags a trailing floor upward and then breaches it on a retracement. The Center of Gravity guide covers a related near-zero-lag idea, but CG measures where the weighted centre sits relative to the plain centre, while FRAMA measures how rough the path itself is — two different questions, and FRAMA’s answer is the one that tells you whether to trade at all.
Under a consistency rule, FRAMA’s low trade count is an advantage. Filtered FRAMA produces a handful of trades a week, not dozens, so no single day dominates the payout total the way a news-spike day or a revenge-trading day does. The worked trade paid 750 dollars, or 15 percent of a 5,000-dollar day, and the best trade in the twelve-trade test was 800 dollars against a 3,000-dollar net, or 26.7 percent — both inside a typical 30 percent cap, with room to spare. The discipline that keeps those numbers inside the cap is not the indicator; it is the regime gate, the session filter, and the rounding-down sizing rule, applied every time, including on the days when FRAMA looks unbeatable. Those are the days it matters most.
Daily routine for trading FRAMA
The routine is short on purpose. A funded routine with twelve steps never survives a live morning; this one has five, and the first four happen before any entry is considered.
| Step | Action | Time |
|---|---|---|
| 1. Regime check | Read D on H1 and H4: trade only when D is at or below 1.60 | Before the London open |
| 2. News check | Mark red-folder releases on EURUSD and GBPUSD, block 30 minutes either side | Before the London open |
| 3. Signal scan | Check price crosses of a moving FRAMA on closed bars only | London and New York sessions |
| 4. Size before entry | Measure the swing stop, run the sizing formula, round down | Before every entry, no exceptions |
| 5. Journal the trade | Record D, alpha, stop, size and result the same day | End of session |
Frequently asked questions
Is FRAMA really better than a normal moving average?
It is better at one specific job: changing its own speed to match the market. A fixed average is always a compromise, too slow in trends or too twitchy in ranges. FRAMA removes the compromise by letting the fractal dimension set the speed. It is not better at predicting direction, and it still needs the regime gate and the session filter in this guide.
Why must the FRAMA period be even?
Because the dimension calculation splits the window into two equal halves. An odd period cannot split evenly, so the two half-ranges cover different lengths and the dimension comes out wrong. Sixteen is the standard even choice; any even length works in principle, but sixteen is the one Ehlers published and the one this guide tested.
What does a fractal dimension above 1.60 actually mean?
It means the recent price path is filling its box rather than travelling across it — the signature of chop. Price is wandering, reversals are frequent, and crosses of any average in that window are mostly noise. The funded-account response is to stand aside until D falls back below 1.60, not to trade the chop harder.
Can I trade FRAMA crosses on their own?
On a demo account, yes, and you will learn quickly why the answer for a funded account is no. The unfiltered test in this guide lost 4R taking every cross, while the filtered set made 6R on the same market. The regime gate, the H4 agreement check and the session filter are not optional extras; they are the strategy.
How is FRAMA different from KAMA and VIDYA?
All three are adaptive averages, but they adapt on different inputs. KAMA uses an efficiency ratio from closing prices, VIDYA uses a volatility ratio, and FRAMA uses the fractal dimension from high-low ranges. In practice FRAMA is the most sensitive to the shape of the path and the least sensitive to a single large bar, which is why this guide prefers it for funded accounts — with the same filters applied regardless.
What stop should I use with a FRAMA signal?
The recent swing low for a long, or the recent swing high for a short, with no buffer games and no round-number guessing. In the worked trade, the swing low at 1.2670 gave a clean 50-pip stop and exactly 1.00 lot at the 500-dollar cap. If the swing stop is so wide that the rounded-down size feels too small, take the small size or skip the trade — never narrow the stop to justify a bigger position.
Risk disclaimer: Trading leveraged forex and CFDs carries a high level of risk and may not be suitable for all traders. Prop firm accounts add their own rules on top of market risk, and breaking a daily loss limit or consistency rule can end an account even on a winning strategy. The numbers in this guide come from the author’s own testing on a funded-style account structure and are illustrative, not a promise of future results. Always check your firm’s current rulebook before trading any indicator live.





1 comment
[…] The zero lag EMA removes lag a different way, by adjusting the price itself before smoothing. The FRAMA changes its own speed depending on how fractal the market is. The DEMA sits in the middle of that […]