Keltner Channel Breakout

How the rule works

An ATR-based channel adapts its width to recent ranges. A first completed close outside the channel is a volatility-expansion hypothesis in the EMA100 direction. Repeated outside closes are deliberately not counted as new breakouts. Reversions through EMA20 can be frequent when prices oscillate around a flat trend filter.

Setup

• Use EMA20(Close) ± 2 × MT5 ATR20 for channel bands; EMA100(Close) is the trend filter. Long signal requires Close[1] > upper[1] and EMA100[1], with Close[2] <= upper[2]. Short requires Close[1] < lower[1] and EMA100[1], with Close[2] >= lower[2].
• This is the modern EMA/ATR channel definition. It differs from Chester Keltner’s original typical-price/average-range construction. One position per symbol.

Indicators

• EMA(20)
• ATR(20)
• Keltner multiplier 2
• EMA(100)

Long entry

• Buy at next-bar ask after the confirmed upper-band close. Do not take an intrabar band touch because it may disappear by close.

Short entry

• Sell at next-bar bid after a confirmed lower-band close while close is below EMA(100).

Stop loss

• Initial stop is the EMA(20) at signal close or 2×ATR(20) from entry, whichever produces the smaller loss distance only if it lies on the correct side; otherwise use 2×ATR.

Exit

• Exit a long after a completed close below EMA20, or a short after a close above EMA20, at the next executable quote. Time exit: 18 completed H4 bars.
• Optionality is removed from this specification: also maintain a 3×ATR20 chandelier stop. Long uses highest completed high since entry minus 3×current ATR; short uses lowest completed low plus 3×ATR. Tighten only after a close.

Risk and position size

• Risk 0.50% equity. Skip signals where spread exceeds 10% of stop distance, because trading costs dominate a narrow channel trade.

Costs and execution

• Keltner implementations differ in ATR versus average-range choice. State this formula in code, use finished H4 data, log executable spread/slippage, and account for rollover swap.

When the method can fail

• Avoid low-volatility compressions immediately ahead of event risk and repeated outer-band closes without follow-through. Breakouts can fail back through the average.

Worked example

• Hypothetical channel: EMA20 is 1.1000 and ATR20 0.0020, making the upper band 1.1040. A close of 1.1045 qualifies only if the preceding close was not above its own upper band and price is above EMA100. With entry 1.1047, the nearer valid 2×ATR stop is 1.1007.

Historical test results

No tested pair meets the complete research-screen criterion. Positive cells alone do not qualify the strategy.

24 mechanical runs across 4 pairs on H4. Decisions use completed strategy bars; fills use native ICMarketsEU-Demo M15 Bid candles with explicitly modelled costs. Full window: 1 January 2025–18 September 2026. Shorter available histories: USDCAD: 2025-03-13. Each row starts with an independent USD10,000 account.

Interpretation: these are mechanical bar-model tests, not verified real-tick or live performance. Price rules were held fixed; subjective market-quality exclusions remain outside the mechanical model. Strategy adaptation requires a new test.

Download all test metrics and assumptions (JSON) · Download simulated trade logs and reproducible research code · Broker commission and swap documentation

Execution conventions

• Bar [0] is forming; [1] is the most recently completed bar. Unless explicitly stated otherwise, indicators use Close with zero plot shift and signals use completed bars. EMA is exponential, SMA is arithmetic and MT5 ATR is its native true-range average. Use at least 300 completed warm-up bars; a longer rule-specific requirement takes priority.
• A close-based signal is executed at the first available quote after the close, never retrospectively at the signal close. Buy at ask, sell at bid; close longs at bid and shorts at ask. A standing stop or target can fill within a bar. A gap through a stop fills at the first executable price, which can be worse than the stop.
• R means the initial entry-to-protective-stop distance, fixed at entry. Convert that distance into account-currency loss using the symbol tick size and loss-side tick value. Lots = cash risk / loss per lot, rounded down to the broker volume step. Round prices to the trade tick and recheck risk plus minimum stop/freeze distances before submitting. Budget commission and slippage separately; skip if minimum size exceeds the risk limit.
• Use one position per symbol and no averaging down. A trailing stop only tightens after a completed bar and becomes active on the next quote. The entry bar is holding bar one; a time exit executes on the quote after the specified holding bar closes. When a hard stop and a close-based exit compete, the stop already active takes precedence. With OHLC-only data, assume stop before target if both can occur in one bar; tick data is needed to resolve the order.
• The article defines the price rule. Any discretionary news exclusion changes the tested variant and must be recorded in advance. Map broker timestamps to the named London/New York time zone using historical daylight-saving offsets; do not assume a fixed UTC offset. Include spread, commission, slippage, overnight financing and triple-swap days in a profitability test.

Adapt and test your version

• Treat these inputs as a reproducible starting specification. Choose a currency pair, session and holding period that match the time you can monitor the market, then record any changes to indicator periods, thresholds, stop distance or exits as a new version.
• Change one hypothesis at a time. Reserve a later period and additional pairs for evaluation before looking at their results. Include your account’s actual trading and financing costs, compare small parameter changes, and forward-test on demo. A selected profitable window does not validate the strategy or its modified version.

Reproduce the chart in MT5

• Open EURUSD on H4. Add the indicators listed above through Insert → Indicators; use Close and zero shift unless a different setting is specified. The Data Window displays exact values for the selected completed candle.
• Compare the signal candle with the required earlier bars before moving forward. Use the crosshair to read prices and times. For custom channels, session levels or Supertrend, reproduce the formula in this guide and check the calculation before relying on an external indicator.
• Draw entry, initial SL and target only after a candle has satisfied every condition. Label planned levels as plans; they do not demonstrate a historical fill. The featured chart is genuine historical MT5 context and is not a trade-performance report.

Sources

• MT5 Help: Average True Range — Indicator/platform definition; not strategy performance.
• StockCharts: Keltner Channel Parameters — EMA/ATR channel construction. This guide uses ATR20 rather than the documented ATR10 default.

Multi-year EURUSD benchmark — EA 10 v1.01

Keltner Channel Breakout · H4. All 20 planned base/stress cases completed and reconciled. The research binary uses the version 1.01 strategy with the documented tester cost overlay.

EURUSD only, using genuine historical bid/ask ticks in MT5’s real-tick mode. Ten independently started segments span selected dates from 2 July 2018 to 18 September 2026 (end exclusive). Excluded broker-calendar dates and short residual windows are listed in the downloadable benchmark data. Broker time is UTC+2 in US winter and UTC+3 in US summer; source ticks are UTC. Earlier history supplies warmup only, with at least 400 observed completed bars required at the first decision; warmup gaps remain a limitation. The full ten-year data archive is not a ten-year continuous performance test.

Each segment starts flat at USD 10,000. Base uses USD 3.50 per lot per side and no added execution delay. Stress uses USD 7.00 per lot per side plus a 200 ms delay. Observed spread is included in price P/L. Stress changes both commission and delay, so its difference from base does not isolate either effect. Remaining positions are closed and charged at each segment end.

Results below sum independent segment cash outcomes; they are not a compounded account return. Drawdown is the worst observed individual segment, not a continuous full-period or intratick worst-case drawdown. It uses observed native equity after modeled fees and native booked swap. Profit factor groups complete positions, including partial exits; n/a means the loss denominator is unavailable. Native booked swap is shown separately and has not been checked against historical financing rates. Native equity can affect subsequent sizing, so subtracting swap does not recreate a separate no-financing simulation.

The rules were fixed before this campaign, but parts of the historical period had already been inspected during development. This is not an untouched out-of-sample test, and no automatic profitability badge is awarded. Ledger reconciliation is not a complete independent replay of every indicator and decision. These figures describe this exact implementation and sample; they do not establish why an apparent edge would persist. Account-level and portfolio controls described elsewhere on this page still apply.

Base: Summed price P/L was non-positive; modeled commission reduced the result further. The sample lost money before financing.

Stress: Summed price P/L was non-positive; modeled commission reduced the result further. The sample lost money before financing.

Download the benchmark figures, exact segment dates and evidence hashes.