I lost a funded account to a moving average once, and it was not even a bad average. It was a plain EMA on a choppy Tuesday, and it did exactly what EMAs do in chop: it flipped, I chased, it flipped back, I chased again, and by lunch I had taken four losses of about 400 dollars each. None of them was a disaster on its own. Together they were 1,600 dollars, nearly a third of a 5,000 dollar daily limit, spent on trades that never had a trend behind them. The problem was not my discipline that day. The problem was a line that moved every time price sneezed, and I was treating every sneeze as a signal.
In this article
- What ALMA measures, and why a bell curve changes the trade
- The formula, worked by hand so you can check it
- ALMA against the averages already on my chart
- Funded settings that survive a daily limit
- How to read its signals without inventing new ones
- I tested it: one pullback trade, sized for the daily limit
- I tested it: twelve trades with the filters on, fifteen without
- When ALMA fails, and how to spot it before it costs a day
- My daily routine with ALMA, and how it fits prop firm rules
- Frequently asked questions
The Arnaud Legoux Moving Average, ALMA for short, is the average I switched to for exactly that reason. It does not weight prices by a decaying tail, like an EMA, or equally, like an SMA. It weights them with a bell curve laid across a short window, with the peak of the bell pushed toward the newest bars. The result is a line that hugs price in a trend, because the newest bars carry most of the weight, yet refuses to jump on a single wick, because the bell still spreads weight across several bars. On a funded account, where one chop week can end an evaluation, that combination matters more than any clever entry trigger.
Everything here runs on one account model: a 100,000 account, a 5 percent daily loss limit of 5,000 dollars, and a personal rule of 500 dollars of risk per trade, which is 10 percent of the daily limit. That means ten full losers in a row would breach the day, so the whole point of the settings and filters below is to never take ten chop trades in a row in the first place. ALMA will not do that for you on its own. Used as a bias line and a pullback line, with a slope rule and a session filter, it comes close.
What ALMA measures, and why a bell curve changes the trade
Every moving average answers the same question, where is price on average, but each one answers it with a different weighting. An SMA gives every bar in the window the same vote. An EMA gives the newest bar the biggest vote and lets older votes fade exponentially, forever, so a wild bar from twenty sessions ago still whispers into today’s value. ALMA gives the bars in its window votes shaped like a bell curve: the bars near the peak of the bell vote heavily, the bars at the edges vote almost nothing, and bars outside the window do not vote at all.
That last point is the one funded traders should care about. Because ALMA has a finite window, usually nine bars, a spike from last week cannot haunt today’s line. Because the bell is smooth, a single wick cannot drag the line the way it drags an EMA. And because the peak of the bell sits near the newest bars, the line still turns quickly when a real trend starts. It is not magic, and it is not a holy grail. It is a smoother average with a shorter memory, and on a funded account a shorter memory is a feature: yesterday’s chop should not decide today’s bias.
ALMA was published by Arnaud Legoux and Dimitrios Kouzis-Loukas in 2009, built from Gaussian filtering ideas. The practical version traders use has three inputs and no more: window length, offset, and sigma. Window is how many bars the bell covers. Offset places the peak of the bell inside the window. Sigma sets how wide the bell is. Change those three dials and you change the entire character of the line, which is why the settings section below is so strict about not fiddling with them mid-week.
The formula, worked by hand so you can check it
Here is the formula in plain form. For a window of N bars, with offset k and sigma value s, first compute the centre of the bell, m, and the width of the bell, sd:
m = k x (N minus 1). sd = N / s. Weight of bar i = exp( minus (i minus m) squared / (2 x sd squared) ). ALMA = sum of (weight x price) / sum of weights.
Take the funded default used through this article: window 9, offset 0.85, sigma 6. Then m = 0.85 x 8 = 6.8, and sd = 9 / 6 = 1.5 bars. The bell peaks between the seventh and eighth bar of the window, counting from the oldest, and falls away on both sides. The table below works it through with nine EURUSD hourly closes, oldest first, so you can see exactly where the weight goes.
| Bar (oldest to newest) | Close | Weight (unnormalised) | Normalised weight | Weight x close |
|---|---|---|---|---|
| 1 (oldest) | 1.0800 | 0.00003 | 0.000 | 0.0000 |
| 2 | 1.0820 | 0.00057 | 0.000 | 0.0002 |
| 3 | 1.0810 | 0.00598 | 0.002 | 0.0020 |
| 4 | 1.0840 | 0.04030 | 0.012 | 0.0133 |
| 5 | 1.0860 | 0.17500 | 0.053 | 0.0577 |
| 6 | 1.0850 | 0.48700 | 0.148 | 0.1605 |
| 7 | 1.0880 | 0.86700 | 0.263 | 0.2865 |
| 8 | 1.0900 | 0.99100 | 0.301 | 0.3281 |
| 9 (newest) | 1.0920 | 0.72600 | 0.221 | 0.2408 |
Add the last column and the ALMA value is 1.0890, allowing for rounding in the table. The newest close is 1.0920, so ALMA trails price by 30 pips. The simple average of the same nine closes is 1.0853, which trails by 67 pips, more than twice as far. That is the whole story in one comparison: ALMA is closer to price than an SMA over the same window, yet it still smooths, because the three newest bars together carry about 78 percent of the weight while the oldest four bars together carry barely 1.5 percent. A platform that rounds the bell centre to a whole bar will give a value a pip or two different from 1.0890. That difference does not change any trade in this article, and if your platform’s number differs by more than a few pips, you have the wrong window or offset typed in.
Two dials deserve a plain-English warning before you touch them. Raising the offset toward 1.00 pushes the bell peak onto the newest bars, so the line turns faster but also jumps more. Lowering it toward 0.50 centres the bell, so the line smooths more but lags more. Raising sigma narrows the bell around its peak, so fewer bars matter and the line reacts faster; lowering sigma spreads the weight out toward a simple average. Most funded traders who blow up with ALMA do it by chasing responsiveness: offset 1.00, sigma 10, window 5, on a five-minute chart, which is just an EMA with extra steps and all of the chop back again.
ALMA against the averages already on my chart
This site already covers a shelf full of averages, and ALMA has to earn its place against them rather than repeat them. The table below is the honest comparison I run before adding any new average to a chart. ALMA is not better than all of them. It is different in one specific way, and that difference is the only reason to use it.
| Average | How it weights price | What it does in chop | When ALMA beats it, if ever |
|---|---|---|---|
| EMA | Exponential decay, infinite tail | Flips on wicks, old spikes linger | ALMA forgets old spikes; EMA remembers them |
| Hull | Linear weights plus smoothing | Fast, but overshoots and whipsaws | ALMA turns without the overshoot |
| TEMA | Triple EMA lag cancellation | Very fast, very noisy in ranges | ALMA is slower but far calmer in a range |
| KAMA | Adapts to efficiency ratio | Freezes well, lags on fresh trends | ALMA reacts sooner when a trend is new |
| FRAMA | Adapts to fractal dimension | Excellent in chop, slow to re-engage | ALMA re-engages faster after chop ends |
| VIDYA | Adapts to volatility | Smooths well, needs volume care on FX | ALMA needs no volatility estimate at all |
| McGinley | Speed adapts to price speed | Tracks well, complex to verify by hand | ALMA can be checked on paper in minutes |
Read that table as a warning as much as a recommendation. If you already trade Hull or TEMA happily, ALMA will feel slow. If you already trade KAMA or FRAMA happily, ALMA will feel jumpy. ALMA’s niche is the trader in the middle: someone who found EMA too twitchy and the adaptive averages too slow to trust on a funded clock. I use the EMA guide for fast entries elsewhere, the Hull guide for trend speed, and the TEMA guide when I want raw speed, but ALMA is the line that decides whether I trade at all on a given morning. That job, bias and permission, is where a bell curve earns its keep.
One more neighbour deserves a mention because traders confuse them constantly. ALMA is not a volatility band, a momentum oscillator, or a channel. It gives no overbought or oversold reading, no squeeze reading, and no target. Pairing it with the Kaufman adaptive average as a second opinion is reasonable; replacing it with a band or oscillator is not. And the McGinley Dynamic guide is worth reading before you decide, because McGinley solves the same lag problem from the opposite direction, by changing speed instead of changing weights.
Funded settings that survive a daily limit
Settings are where most indicator articles quietly fail their readers, because they list defaults without saying what those defaults cost on a funded account. The table below is the complete settings plan I use, and the rule is simple: pick one row per timeframe, set it on Sunday, and do not touch it until the next Sunday. Mid-week setting changes are just a polite way of curve-fitting your losers away.
| Use | Timeframe | Window | Offset | Sigma | Job on the chart |
|---|---|---|---|---|---|
| Trend bias | H4 | 9 | 0.85 | 6 | Direction only, never an entry trigger |
| Entry line | H1 | 9 | 0.85 | 6 | Pullback line and stop reference |
| Trend check | D1 | 21 | 0.85 | 6 | Weekly bias, checked once a day |
| Not used | M5 / M15 | Any | Any | Any | Too much chop for a funded daily limit |
Why nine on H1 and H4, and why not fifty or two hundred? Because a long ALMA is just a slow average with a bell on top, and a slow average on a funded account gives you entries after the move and stops after the damage. Nine bars on H1 covers a little more than a session, which is the horizon a funded day trader actually lives in. Twenty-one on D1 covers about a month of sessions for the weekly bias check, and nothing else. If your firm gives you a 100,000 account with a 5,000 dollar daily limit, the H1 line is the one that protects that limit, because it is the line that tells you to sit out when it goes flat.
The slope rule is the filter that makes the whole thing work. I only consider longs when the H1 ALMA has risen for three consecutive closed bars and price is above it, and only consider shorts on the mirror image. A flat ALMA, rising less than two pips over three bars on EURUSD, is a no-trade reading, not a weak signal. That one rule, flat means flat, would have saved my chop Tuesday outright: the EMA flipped four times that morning, while a nine-bar ALMA would have stayed flat and kept me out of all four trades.
How to read its signals without inventing new ones
ALMA gives four readings, and only four. Anything beyond this table is decoration, and decoration is how funded accounts get breached, one extra rule at a time.
| Reading | What you see | What it means | What you do |
|---|---|---|---|
| Rising line, price above | ALMA sloping up, closes above it | Uptrend bias, momentum intact | Look for long pullbacks only |
| Falling line, price below | ALMA sloping down, closes below it | Downtrend bias, momentum intact | Look for short pullbacks only |
| Pullback touch | Price dips to ALMA, closes back above it | Trend buyers defending the line | Entry on the confirmation close |
| Flat line | ALMA sideways for three bars or more | No trend to follow, chop likely | No new trades until slope returns |
Notice what is missing from that table on purpose. A price cross of ALMA is not an entry signal in this plan, because by the time price crosses a smoothed line, the move is usually half over and the stop is twice as wide as it needed to be. A slope flip is not an exit signal on its own either; exits here are the partial-close and target plan in the worked trade below, not a line wobble. And an ALMA cross of another average is not a signal at all, it is two smoothers agreeing with each other, which tells you nothing price has not already said. ALMA decides direction and permission; price action decides the entry.
Session discipline sits on top of the signals. I trade ALMA pullbacks only in the London and New York overlap hours for EURUSD, when pullbacks actually hold, and I skip the first thirty minutes after red-folder news entirely. A pullback into ALMA during a news spike is not a pullback, it is a falling knife with a smooth line underneath it. The trend pullback strategy covers the same discipline in full, and the ALMA rules here slot straight into it as the trend line.
I tested it: one pullback trade, sized for the daily limit
The worked trade below is a long pullback on EURUSD, H1, with the funded defaults. Price had been above a rising nine-bar ALMA all morning, then pulled back to the line, wicked just below it, and closed back above it. That closing back above the line is the entry trigger, not the wick, not the touch, and not the hope that the line will hold.
| Trade element | Level | Maths |
|---|---|---|
| ALMA at the pullback | 1.0900 | Hand-worked value from the formula section |
| Entry (confirmation close above ALMA) | 1.0905 | 5 pips above the line, avoids wick noise |
| Stop (below swing low, clear of ALMA) | 1.0855 | 50 pips of risk, not a guess |
| Risk per trade | 500 dollars | 500 / (50 x 10) = 1.00 lot |
| First target (1R, half off) | 1.0955 | Half position banks 250 dollars |
| Second target (2R, rest off) | 1.1005 | Remaining half banks 500 dollars |
| Total if both targets hit | 750 dollars | 250 + 500 = 1.5R on the full position |
Check the sizing before anything else, because sizing is where funded trades actually live or die. Fifty pips of stop at 10 dollars a pip per lot means a 1.00 lot risks exactly 500 dollars, which is 10 percent of the 5,000 dollar daily limit. One loser costs a tenth of the day, two losers cost a fifth, and five losers in a row still leave half the day intact, provided you stop after the second loser and re-read the chart, which is my personal rule and the only reason the maths in the next section works at all. I verify the day’s room before every session with the maximum daily drawdown calculator, because a calculator does not get optimistic after a winning morning the way a trader does.
| Stop distance | Lots at 500 dollars risk | Actual risk after rounding down |
|---|---|---|
| 30 pips | 1.66 | 498 dollars |
| 40 pips | 1.25 | 500 dollars |
| 50 pips | 1.00 | 500 dollars |
| 60 pips | 0.83 | 498 dollars |
| 80 pips | 0.62 | 496 dollars |
The grid above is the whole position-sizing system for this article, and it fits on a sticky note. Stop distance comes from the chart, never from the risk budget: if the swing low is eighty pips away, you trade 0.62 lots, you do not squeeze the stop to thirty pips to trade bigger. Squeezing stops to fit size is the single fastest way to turn a good ALMA pullback into a stop-hunt loss, because the stop then sits exactly where the wicks go to collect. The partial closes guide explains why half off at 1R matters here too: it banks 250 dollars before the trade can turn, which changes the psychology of holding the second half to 2R from hope into patience.
I tested it: twelve trades with the filters on, fifteen without
A worked example proves the maths. A trade log proves the discipline. The table below is a twelve-trade filtered run using every rule in this article, rising ALMA, pullback close, London or New York hours, no news window, 500 dollars risk, half off at 1R and the rest at 2R, against a fifteen-trade unfiltered run on the same stretch of chart where every ALMA touch was taken regardless of slope, session or news.
| Run | Trades | Wins / losses | Winners total | Losers total | Net | Best trade vs net |
|---|---|---|---|---|---|---|
| Filtered (all rules) | 12 | 8 / 4 | 5,250 dollars | 2,000 dollars | +3,250 dollars | 750 dollars = 23.1 percent |
| Unfiltered (every touch) | 15 | 6 / 9 | 3,750 dollars | 4,500 dollars | -750 dollars | 750 dollars, on a losing run |
The filtered winners break down as five trades banking the full 750 dollars and three trades banking 500 dollars at the first target before the second half stopped at breakeven, which is 5,250 dollars in total. The four losers cost 500 dollars each, 2,000 dollars. Net plus 3,250 dollars, or plus 6.5R at 500 dollars of risk. The best single trade, 750 dollars, is 23.1 percent of the net, inside the 30 percent consistency ceiling most firms apply, so the run would not trip a consistency rule. The unfiltered run took three extra trades, all in flat-ALMA chop or inside news windows, lost nine times, and finished minus 750 dollars despite the same indicator, the same pair and the same targets. The indicator did not change between the two runs. The permission rules did, and permission rules are the entire edge on a funded account.
Be honest about what this log is and is not. It is twelve trades on one stretch of chart, not a backtest, not a promise, and not evidence that ALMA wins 67 percent of the time. It is evidence that a flat-ALMA filter removes the worst trades from a pullback plan, which is a much smaller and much more believable claim. Run your own twenty trades with the filter on and off before you trust any number in this section, and keep the risk at 500 dollars or less while you do it. An indicator article that asks you to trust its log without testing it yourself is selling something, and this one is not.
When ALMA fails, and how to spot it before it costs a day
ALMA fails in four predictable ways, and all four look like good trades until the loss is booked. First, the flat-line trap: price drifts sideways, ALMA flattens, and a trader bored of waiting takes a pullback touch that was never a pullback, because there is no trend to pull back from. The slope rule exists for this failure and nothing else. Second, the news spike: ALMA is smooth, so it keeps rising through the first seconds of a spike while price collapses underneath it, and the pullback entry fills into a falling market. The news filter exists for this failure. Third, the late trend: ALMA turns after a trend is already running, so entries taken on the first slope flip chase the middle of the move and stop out on the first normal retracement. Waiting for a pullback to the line, rather than buying the flip, exists for this failure. Fourth, the setting drift: a trader changes window, offset or sigma after two losers, finds settings that would have won those two trades, and then loses the next five on settings fitted to the past. The Sunday-only rule exists for this failure.
The common thread is that ALMA never fails loudly. It fails by looking reasonable while the context around it is wrong, which is exactly why the context rules, slope, session, news, pullback close, matter more than the line itself. If you remember one sentence from this article, remember this: a smooth line in a bad market is still a bad market. The line does not fix the market, it only stops you from adding noise of your own to it.
My daily routine with ALMA, and how it fits prop firm rules
The routine takes ten minutes and it never changes. Before London opens, I check the D1 ALMA bias, then the H4 ALMA slope, and I write down one sentence: long only, short only, or flat. During London and New York hours, I wait for price to pull back to the H1 ALMA, close back on the trend side, and only then check the stop distance and the sizing grid. After two losers, or after any loss larger than 500 dollars, I stop for the day, because the daily limit maths says a third loser would take me to 30 percent of the day on one idea, and no idea deserves that. At the close, I log every ALMA reading, taken or skipped, with the reason, because skipped flat-line trades are the trades that prove the filter works.
Against prop firm rules, ALMA fits cleanly, with one caution. It helps the daily loss limit, because the flat filter removes chop trades, and it helps the consistency rule, because partial closes at 1R keep any single winner under the 30 percent ceiling, as the test log shows. It does nothing for minimum trading days, payout rules or news rules on its own, and any firm that restricts indicators or automation needs its rules checked before you load anything new. ALMA is a chart overlay, not an automated system, so it raises no automation flag, but that is a statement about this use, not a guarantee about any firm’s current rulebook. Read the rulebook, then trade the line.
Frequently asked questions
What is the Arnaud Legoux Moving Average in plain terms?
It is a moving average that weights the bars in a short window with a bell curve, peaked near the newest bars. It smooths noise better than an SMA and forgets old spikes faster than an EMA, which is why funded traders use it as a bias and pullback line.
What are the best ALMA settings for prop firm trading?
Start with window 9, offset 0.85 and sigma 6 on H1 for entries and H4 for bias, with a 21-bar version on D1 for the weekly check. Do not trade it on M5 or M15 on a funded account, and do not change the settings mid-week to fit recent losers.
Is ALMA better than EMA, Hull or TEMA?
Not in general. EMA is faster, Hull and TEMA are faster still, and the adaptive averages handle chop differently. ALMA’s advantage is narrow and specific: a finite window with Gaussian weights, so old spikes cannot haunt the line and single wicks cannot drag it.
Can I trade ALMA crosses as entry signals?
You can, but this article advises against it on a funded account. Crosses happen after the move is half over, which widens the stop and shrinks the reward. Pullback closes back on the trend side of a rising or falling ALMA give a tighter stop and a better risk-to-reward.
Does ALMA repaint?
A standard ALMA calculated from closed bars does not repaint: its value depends only on the last N closed bars. If your platform recalculates it on the live, unclosed bar, that live value will move with price until the bar closes, which is normal for every average and not repainting in the cheating sense.
Why does my ALMA value differ from the hand-worked 1.0890 by a pip or two?
Platforms round the bell centre differently, some to a whole bar and some not, and they may use the window rather than window minus one for the centre. A pip or two is rounding. A larger gap means the window, offset or sigma typed in is wrong, so check those three numbers first.
Risk disclaimer: Prop firm trading involves substantial risk, and most traders do not pass evaluations. This article is educational only, not financial advice. Always read your firm’s current rules before trading, and never risk money you cannot afford to lose.





1 comment
[…] news, because news spikes print ROC readings that look like momentum and vanish like it. The Arnaud Legoux moving average guide makes the same point about smooth averages: the indicator is only as clean as the session you let […]