Is the Keltner Channel built into MetaTrader 5?

No. There is no iKeltner function. It is built from two that do exist — iMA for the centre line and iATR for the width — which is what the Builder's node does. That is also why the node has two period settings instead of one.

What is the difference between Keltner Channels and Bollinger Bands?

The input. Bollinger's width is the standard deviation of the closing prices; Keltner's is the average true range, which is computed from each bar's high, low and previous close. Measured, taking a series and tripling the wick length without moving any close leaves Bollinger's half-width unchanged to the last bit and exactly doubles Keltner's. The two are usually within about 11% of each other in size, which is why the difference is easy to miss.

Usually Bollinger, but not by much and not always. Measured at the same period with the conventional constants, Keltner ran 0.89 to 0.96 times as wide across three synthetic series. The gap was largest on the trending one, because a sustained move inflates the standard deviation of the closes while leaving true range roughly where it was.

Does MetaTrader's ATR match Wilder's?

No, and this is the caveat that matters most here. The bundled ATR.mq5 computes a rolling simple average of true range; Wilder's ATR is a 1/N smoothing that never discards anything. Their long-run averages agree — 9.57 against 9.57 pips measured — but after volatility tripled, the platform's version reached 90% of the new level in 19 bars against Wilder's 59.

What ATR period matches Wilder's at 20?

32 or 33, measured. The analytic answer is 39, from matching the centre of mass, and the standard deviations do agree there — but the correlation between the two series is already falling by 39, and the closest overall match lands between the two effects. Expect a close match, not the same line.

Should the centre be an EMA or an SMA?

EMA is the convention and the node's default. It is not a small choice: measured, the two centres differ by 0.89 to 2.58 pips at period 20, and moving to a simple average raised the count of bars closing above the upper band from 955 to 1439 on the same series.

What is the Keltner squeeze?

The state where Bollinger's bands sit entirely inside Keltner's — the closes have stopped scattering relative to how far the bars are still travelling. Measured, it holds on 12.7% to 16.6% of bars, in runs averaging about six. It needs both indicators on the chart, because it is a comparison between two different measurements rather than a reading of one.

How much history does it need?

About 100 bars, and the centre line is what decides that. Measured at period 20 against a full-history baseline: the platform's ATR was exact at 40 bars, Wilder's was off by 0.038 there and reached 0.0009 at 100, and the EMA centre was off by 0.25 at 40 and 0.028 at 60.

Is this the same as the original Keltner Channel?

No. The 1960 version averaged high minus low around a simple average of the typical price; the modern one uses true range around an EMA. With no gaps the two widths are identical, measured to within 0.00%. With gaps on 15% of bars the original reads 23.87% narrower, because plain range does not count the distance covered while the market was closed.

Can I build this without coding?

Yes. The template is three Keltner nodes at period 20 with a 2.0 multiplier — one on the upper line, one on the lower, and one on the centre read from H4 — a price node, two Cross nodes for the band breaks, two Compare nodes against the H4 centre, two And gates, two Close nodes wired ahead of the entries, and Open Buy / Open Sell on a fixed lot. The periods and the multiplier are set in the Builder; the compiled EA exposes them along with the stop distances, lot size and maximum spread as inputs.

Personalizados e da comunidade VolatilityChannelNot built in Disponível no Builder

Keltner Channel

A moving average with bands set at a multiple of average true range. It looks like Bollinger Bands and is built from a different measurement: Bollinger's width is the standard deviation of the closes, Keltner's is the size of the bars. Widen the bars without touching a single close and Bollinger does not move at all while Keltner doubles.

Default: 20 ± 2 × ATR(20)
Raschke variant: 20 ± 2 × ATR(10)
vs Wilder's ATR: iATR(32–33) ≈ Wilder(20)

Criar no Builder Receita de estratégia

Ilustrativo — dados sintéticos, não uma cotação em tempo real.

What a Keltner Channel tells you

A Keltner Channel is a moving average with a band drawn at a fixed multiple of average true range above and below it. That is one word different from the description of Bollinger Bands — multiple of average true range instead of multiple of standard deviation — and the difference is not a matter of degree. Standard deviation is computed from closing prices; true range is computed from each bar's high, low and previous close. They answer different questions: how far the closes scatter, against how much ground the market covers within a bar. The cleanest way to see it is to take a series, leave every close exactly where it is, and make the bars taller. Measured that way, Bollinger's half-width does not change by a single unit and Keltner's doubles. The two lines are usually within about a tenth of each other in size, which is precisely why the distinction gets lost: they look interchangeable and are not.

The width comes from the bars, not from the closes
MetaTrader has no Keltner — it is an iMA and an iATR combined
And MetaTrader's iATR is not Wilder's ATR, which is what the usual definition assumes

1 Centre line: a moving average of the applied price, EMA by convention
2 Width: average true range over its own period, times a multiplier
3 Bands: the centre plus and minus that width, recomputed every bar

Mostrar a fórmula e detalhes de cálculo

MetaTrader has no Keltner Channel, so the calculation below is the one the Builder’s node performs with two handles. The centre line and the width are separate indicators, and each brings its own definition.

mid   = MA(applied_price, period)          <- iMA, EMA by convention
width = ATR(atr_period) * multiplier       <- iATR
upper = mid + width
lower = mid - width

The second line is where the platform detail lives. ATR.mq5, bundled under Indicators/Examples, computes:

TR[i]  = max(high[i], close[i-1]) - min(low[i], close[i-1])
ATR[i] = ATR[i-1] + (TR[i] - TR[i-N]) / N        <- a rolling window, i.e. an SMA

Wilder’s ATR — the one nearly every description of this indicator assumes — is:

ATR[i] = (ATR[i-1] * (N-1) + TR[i]) / N          <- nothing leaves, it only decays

Both are linear filters whose weights sum to one, so their long-run averages must agree, and measured over 2,700 bars at period 20 they do: 9.571 against 9.573 pips. What separates them is timing. On a series whose volatility tripled at a fixed bar:

                    before    after    bars to reach 90% of the new level
iATR(20)             3.31      9.63                 19
Wilder ATR(20)       3.30      9.64                 59

Matching the period does not close the gap. Wilder’s 1/N smoothing has the centre of mass of a simple average over 2N-1 = 39 bars, so iATR(39) is the analytic answer — but sweeping the period, the closest match was 32–33. Decomposing shows why:

iATR(n) vs Wilder ATR(20), walk-high series
  n     mean          sd            corr
 20   9.571/9.573   1.100/0.778   0.8897
 25   9.572/9.573   0.977/0.778   0.8947   <- correlation peaks
 32   9.573/9.573   0.855/0.778   0.8782   <- |diff| minimised
 39   9.574/9.573   0.767/0.778   0.8482   <- sd matches, as predicted

The means agree at every period. The standard deviations line up at 39 exactly as the theory says, and the correlation is already falling by then; the smallest average difference sits between the two.

The other half of the page is what the width is measured from. Take a series, leave every open and close where it is, and triple the wick length:

                 closes        Bollinger half-width      Keltner half-width
walk-low       max|d| = 0.0        max|d| = 0.0          3.18 -> 6.38  (x2.00)
walk-high      max|d| = 0.0        max|d| = 0.0          9.57 -> 19.18 (x2.00)
trending       max|d| = 0.0        max|d| = 0.0          4.81 -> 9.62  (x2.00)

Bollinger cannot see the change, because it never looks at a high or a low. That is the whole distinction, and it survives into the trade count — at widths that differ by only 11%, the trending series produced 1358 closes above Bollinger’s upper band and 2393 above Keltner’s.

Finally, the 1960 original used plain high-minus-low rather than true range. With no gaps the two are the same width to within 0.00%; with gaps on 15% of bars the original reads 23.87% narrower, which is exactly the distance true range was invented to count.

Sinais

Os sinais distintos e testáveis que Keltner oferece — e o regime adequado para cada um.

Band Break

Pronto no Builder

Condição: The close moves outside the upper band (mirrored for shorts)
Melhor regime: Trending
Uso típico: The template's entry, gated by a higher timeframe so the break has a direction

Centre-line Bias

Pronto no Builder

Condição: Price above or below the centre line on a higher timeframe
Melhor regime: Trending
Uso típico: One node and one Compare — the gate the template uses, and useful on its own

Return to the Centre

Pronto no Builder

Condição: Price crosses back through the centre line after a band break
Melhor regime: Either regime
Uso típico: An exit that scales with volatility instead of a fixed distance

The Squeeze

Interpretação manual

Condição: Bollinger's bands sit entirely inside Keltner's
Melhor regime: Ranging
Uso típico: A compression reading that needs both indicators, since it compares the two measurements against each other

Implementação no MT5

O que o MetaTrader 5 realmente calcula e desenha — a referência para cada regra desta página.

Notas da plataforma

Função MQL5

iMA / iATR

Preço aplicado

Close Close · Open · High · Low · Median · Typical · Weighted

No Builder

The applied price goes to the centre line only. The width cannot take one: true range is defined from the high, the low and the previous close, so there is nothing to choose. The node exposes the centre period, the ATR period, the multiplier, which of the three lines to publish, the averaging method, shift, timeframe and a variable name.

Notas de implementação

★★★ MetaTrader's iATR is not Wilder's ATR. The bundled source (ATR.mq5, under Indicators/Examples) computes `ExtATRBuffer[i] = ExtATRBuffer[i-1] + (ExtTRBuffer[i] - ExtTRBuffer[i-N]) / N`. The recurrence makes it look like a smoothing, but subtracting the true range from N bars ago is what a rolling window does: this is a simple moving average of true range, seeded with the mean of the first N values. Wilder's ATR, which is what almost every description of the Keltner Channel assumes, is `(prev*(N-1) + TR)/N` — nothing leaves the average, it only decays. Both are legitimate; they are not the same function, and the channel's width is whichever one your platform calls.
★★★ The two ATRs have the same average and different timing. Measured over 2,700 bars on each of three synthetic series at period 20, the mean absolute difference was 0.15 to 0.43 pips on average levels of 3.18, 9.57 and 4.81 — a relative gap of 4.5% to 6.4% — and the means themselves agreed to three significant figures (9.571 against 9.573). That agreement is not a coincidence: both are linear filters whose weights sum to one, so in a steady state they must land on the same number. What differs is what happens when the state is not steady. On a series whose volatility tripled at a fixed bar, the simple-average version reached 90% of its new level in 19 bars and Wilder's took 59.
★★★ Matching the period does not fix it, and the analytic prediction is wrong. Wilder's 1/N smoothing has the same centre of mass as a simple average over 2N-1 bars, so ATR(20) Wilder should be matched by iATR(39). Measured, the closest iATR period was 32 or 33 depending on the series, not 39. Decomposing the difference explains why: the means agree at every period, the standard deviations agree at 39 (0.767 against 0.778) and the correlation is highest around 25 (0.8947, falling to 0.8482 by 39). Mean absolute difference is minimised where those two pull against each other, which is between them. The lesson is the same one the ADX Wilder page reaches from the other direction — a smoothing constant can be matched, a filter cannot.
★★★ Bollinger and Keltner are not the same measurement, and a synthetic series proves it rather than arguing it. Take a series, keep every open and close exactly where it is, and multiply the wick length by three. The closes are then identical bar for bar — measured, maximum difference 0.0 — and so is Bollinger's half-width, to the last bit. Keltner's half-width went from 3.18 to 6.38, 9.57 to 19.18, and 4.81 to 9.62 pips: a factor of exactly 2.00 in all three. Any statement of the form 'Keltner is Bollinger with ATR instead of standard deviation' is true and hides that the two are reading different columns of the same data.
★★ The widths are close and the break counts are not. Measured at period 20 with a multiplier of 2 against Bollinger at 20 and two deviations, the average full widths were 12.74 against 13.39, 38.28 against 39.69, and 19.24 against 21.64 pips — Keltner between 0.89 and 0.96 times as wide. But the number of bars closing above the upper band was 955 against 865, 749 against 713, and on the trending series 2393 against 1358. That last pair is a factor of 1.76 from a width difference of 11%, because on a trending series the standard deviation of the closes grows with the trend while true range does not.
★★ The squeeze is a real, countable state. Counting bars where Bollinger's two-deviation bands sat entirely inside Keltner's 1.5-ATR bands: 14.93%, 16.60% and 12.67% of bars across the three series, arriving in 328, 359 and 281 separate runs averaging just over six bars each. That is frequent enough to be a filter rather than an event, which is the argument for treating it as a regime reading and not as a signal.
★★ Warm-up is decided by the centre line, not by the width. Recomputing from a truncated history and comparing the last value against a full-history baseline at period 20: the simple-average ATR was exact at 40 bars — a rolling window cannot remember further back than its own length — while Wilder's ATR was still off by 0.038 pips at 40 and reached 0.0009 at 100. The EMA centre was the slowest of the three, off by 0.25 pips at 40 bars and 0.028 at 60. If a flow reads this node soon after a symbol's history has been trimmed, it is the centre that is not ready yet.
★ The 1960 original is a third indicator, and the difference is gaps. Chester Keltner's version averaged the plain high-minus-low range, not true range, around a simple average of the typical price. On a series with no gaps the two widths are identical — measured, they agreed to within 0.00%. Introduce gaps on 5% of bars and the high-minus-low version reads 9.13% narrower; at 15% of bars, 23.87% narrower. True range exists precisely to count the distance a market travelled while it was closed, and the original formula predates the convention.
★ The MQL5 source writes true range in an unfamiliar but equivalent form. ATR.mq5 computes `MathMax(high, close_prev) - MathMin(low, close_prev)` rather than the usual maximum of three distances. Measured across 44,985 bars on gapped series, the two forms differ by exactly zero. It is the same quantity, written to avoid three comparisons.

Funciona melhor / use com cuidado

Nenhum indicador tem vantagem universal. Veja onde Keltner ajuda — e onde ele engana.

Funciona melhor

Breakout rules, where the threshold should scale with how far the market has been travelling
Instruments that gap, where true range counts the gap and a close-based width does not
Stops and targets expressed as a multiple of the same width the entry used
Any rule whose multiplier was measured on this platform's ATR rather than borrowed

Use com cuidado

Assuming it is Bollinger with different arithmetic — the input series is not the same
Porting a multiplier from a chart drawn with Wilder's ATR; the width turns three times slower there
Trading band breaks without a direction filter — measured, they are 1.76× as frequent as Bollinger's
Reading the original 1960 formula and the modern one as the same indicator

Criar uma estratégia Keltner

Conecte o sinal às regras de entrada e saída, depois exporte um EA MT5 compilável — sem código.

Three Keltner Channel nodes at 20 with a 2.0 multiplier — the upper edge, the lower edge, and a centre on H4
The Bid crossed against the upper edge, once upward and once downward
The Bid compared against the H4 centre, feeding both directions
Cross up while above the H4 centre → Close short, then Open Buy · SL 60 / TP 180
Cross down while below it → Close long, then Open Sell, mirrored

Combinar Keltner com outros indicadores

Um indicador raramente funciona sozinho. Essas combinações cobrem os pontos cegos do Keltner.

Keltner + Bollinger Bands

Keltner Channel
Bollinger Bands
And gate

Por quê: The pairing that explains both. Put them on the same chart at the same period and they will usually be within about a tenth of each other in width — measured, Keltner ran 0.89 to 0.96 times as wide across three synthetic series. That closeness is what makes the difference easy to miss, and the difference is structural: hold every close fixed and widen the bars, and Bollinger's half-width does not move by a measurable amount while Keltner's doubles. The classic use of the pair is the squeeze — Bollinger's bands falling entirely inside Keltner's, which happened on 12.7% to 16.6% of bars in the same measurement, in runs averaging about six bars
Melhor regime: Either regime

No template holds both. Add a Bollinger node beside the Keltner nodes and compare the two upper values to detect the squeeze.

Abrir o Builder →

Keltner + ATR

Keltner Channel
ATR
And gate

Por quê: ATR is the width, so putting it beside the channel is a way of reading the same number twice — once as a distance on the price scale and once as a level you can threshold. That is genuinely useful for stops: a stop at one ATR and a band at two are on the same scale by construction, which fixed-pip stops never are. The caution is the same one the rest of this page makes: MetaTrader's iATR is a simple moving average of true range, so both the band and the stop inherit its speed rather than Wilder's
Melhor regime: Trending

Add an ATR node with the same period as the channel's width and use it for the stop distance.

Abrir o Builder →

Keltner + Envelopes

Keltner Channel
Envelopes
Compare

Por quê: The third way of setting a width, and the one that makes the other two legible. Envelopes place the bands a fixed percentage away, so the width is a property of the price level and nothing else — it does not respond to the market at all. Keltner responds to the size of the bars, Bollinger to the scatter of the closes. Three channels, three different answers to the only question a channel asks, and the choice matters more than the period does
Melhor regime: Ranging

Add an Envelope node at the same period to see how differently a fixed percentage behaves when volatility changes.

Abrir o Builder →

Parâmetros

Valores iniciais para validar no seu próprio par e timeframe — não são configurações garantidas.

Parâmetro	Padrão	Faixa de teste sugerida	O que faz
Period	20	1–999	The centre line's length. Convention is 20 with an exponential average, which is what the template ships. Switching the centre to a simple average is not cosmetic: measured on the same series, the two centres differ by 0.89 to 2.58 pips on average and the number of bars closing outside the upper band rose from 955 to 1439 — about half again as many — because a slower centre lets price run further from it before the band catches up.
ATR period	20	1–999	The width's own length, kept separate from the centre so the common variant of a 20-bar centre with a 10-bar width can be expressed. Shorter widths react to a single violent bar; longer ones describe the session. Note that this period is fed to MetaTrader's iATR, which is a simple moving average rather than Wilder's smoothing — measured, the iATR period that comes closest to Wilder's ATR at 20 is 32 or 33.
Multiplier	2.0	0.1–100	How many average true ranges away the bands sit. Two is the usual figure and is what the measurements on this page use. It is the parameter most worth re-measuring on your own instrument, because it interacts with the ATR period: a shorter width with a larger multiplier and a longer width with a smaller one can produce bands of the same average size that behave completely differently when volatility turns.
Line	Upper	Upper / Main / Lower	Which of the three the node publishes. One node emits one line, so the template uses three; they are created with the same symbol, timeframe and periods, so the terminal serves them from the same pair of indicators rather than from three of each.
Method	EMA	SMA / EMA / SMMA / LWMA	How the centre line is averaged. EMA is the convention and the node's default. This one is worth setting explicitly rather than leaving to a default somewhere else in the chain: a centre that quietly falls back to a simple average is a different strategy, not a slightly different one.
Shift	1	0–100	Which bar the bands are read from. The template uses 1 so the channel is fixed for the duration of the bar being traded. With shift 0 the band moves as the bar forms, and a break can appear and disappear inside the same bar — the usual way a backtest and a live account stop agreeing.
Timeframe	Current	Current / M1–MN1	Which series the handles are created on. The template puts the centre-line gate on H4 while the bands stay on the trading timeframe. Warm-up is worth a thought here: measured, the centre line needs the most history of the three parts — about 100 bars against 40 for the ATR.

Predefinições iniciais

Template 20 ± 2 × ATR(20), EMA centre, H4 gate The conventional settings with a direction filter

Raschke variant 20 ± 2 × ATR(10) The version most modern descriptions use — a faster width on the same centre

Matched to Wilder ATR period 32–33 Measured as the closest match to Wilder's ATR at 20, if your multiplier came from a platform that uses it

Exemplos de mercado

Onde Keltner funciona, onde falha e como um filtro muda o resultado.

Funciona

A break with the higher timeframe behind it

The close pushes outside the upper band while the H4 centre line is already below price. The gate is what makes this different from the break two bars earlier, which had the same shape and no direction.

Falha

The break inside a range

Price closes outside the band and returns within two bars. On a ranging series the bands are narrow, and a narrow band is broken often — this is the case the higher-timeframe gate exists to refuse.

Filtrado

The squeeze before it matters

Bollinger's bands have fallen entirely inside Keltner's — the compression state. Measured, this holds on 12.7% to 16.6% of bars in runs of about six, so it describes a regime rather than marking a moment.

Indicadores relacionados

Bollinger Bands Ver guia →

ATR Ver guia →

Envelopes Ver guia →

FAQ

Is the Keltner Channel built into MetaTrader 5?: No. There is no iKeltner function. It is built from two that do exist — iMA for the centre line and iATR for the width — which is what the Builder's node does. That is also why the node has two period settings instead of one.
What is the difference between Keltner Channels and Bollinger Bands?: The input. Bollinger's width is the standard deviation of the closing prices; Keltner's is the average true range, which is computed from each bar's high, low and previous close. Measured, taking a series and tripling the wick length without moving any close leaves Bollinger's half-width unchanged to the last bit and exactly doubles Keltner's. The two are usually within about 11% of each other in size, which is why the difference is easy to miss.
Which one is wider?: Usually Bollinger, but not by much and not always. Measured at the same period with the conventional constants, Keltner ran 0.89 to 0.96 times as wide across three synthetic series. The gap was largest on the trending one, because a sustained move inflates the standard deviation of the closes while leaving true range roughly where it was.
Does MetaTrader's ATR match Wilder's?: No, and this is the caveat that matters most here. The bundled ATR.mq5 computes a rolling simple average of true range; Wilder's ATR is a 1/N smoothing that never discards anything. Their long-run averages agree — 9.57 against 9.57 pips measured — but after volatility tripled, the platform's version reached 90% of the new level in 19 bars against Wilder's 59.
What ATR period matches Wilder's at 20?: 32 or 33, measured. The analytic answer is 39, from matching the centre of mass, and the standard deviations do agree there — but the correlation between the two series is already falling by 39, and the closest overall match lands between the two effects. Expect a close match, not the same line.
Should the centre be an EMA or an SMA?: EMA is the convention and the node's default. It is not a small choice: measured, the two centres differ by 0.89 to 2.58 pips at period 20, and moving to a simple average raised the count of bars closing above the upper band from 955 to 1439 on the same series.
What is the Keltner squeeze?: The state where Bollinger's bands sit entirely inside Keltner's — the closes have stopped scattering relative to how far the bars are still travelling. Measured, it holds on 12.7% to 16.6% of bars, in runs averaging about six. It needs both indicators on the chart, because it is a comparison between two different measurements rather than a reading of one.
How much history does it need?: About 100 bars, and the centre line is what decides that. Measured at period 20 against a full-history baseline: the platform's ATR was exact at 40 bars, Wilder's was off by 0.038 there and reached 0.0009 at 100, and the EMA centre was off by 0.25 at 40 and 0.028 at 60.
Is this the same as the original Keltner Channel?: No. The 1960 version averaged high minus low around a simple average of the typical price; the modern one uses true range around an EMA. With no gaps the two widths are identical, measured to within 0.00%. With gaps on 15% of bars the original reads 23.87% narrower, because plain range does not count the distance covered while the market was closed.
Can I build this without coding?: Yes. The template is three Keltner nodes at period 20 with a 2.0 multiplier — one on the upper line, one on the lower, and one on the centre read from H4 — a price node, two Cross nodes for the band breaks, two Compare nodes against the H4 centre, two And gates, two Close nodes wired ahead of the entries, and Open Buy / Open Sell on a fixed lot. The periods and the multiplier are set in the Builder; the compiled EA exposes them along with the stop distances, lot size and maximum spread as inputs.

Glossary

Termos-chave

Backtest Overfitting Spread