Scalping

Mechanism

A scalping EA opens and closes within minutes, targeting a small move rather than a large one, and relies on a statistical edge repeated across many trades. Entry signals are fast-oscillator extremes, micro-structure patterns, or spread rebound. Because the target is small, a fixed round-trip cost is a large fraction of it: at the 2-8 pip target these EAs are usually sold with, one pip of spread and commission is 12.5% to 50% of the whole trade. That fraction, not the holding time, is what decides whether the record survives contact with a real account.

Suitability

Requires an account whose total round-trip cost is small relative to the target — raw-spread pricing, low commission, and execution that does not widen at the moment the EA fires. Dense-liquidity windows matter for the same reason: the strategy is a bet that the cost stays below the edge. Unsuitable wherever spread widens on news or where fills are slow, because the loss is not a bad trade but a permanent tax on every trade. Before buying one, measure the actual holding-time distribution rather than reading it off the chart period: on the records published here the fastest charts produced the longest holds.

Notes

Scalping is the only strategy on this list named after a holding time. Every other label describes what the EA thinks — a trend, a reversion, a range. Scalping describes only how long it stays. That makes it the easiest label to apply to something that is not one, and the hardest to verify from a product page.

It also makes it checkable. Holding time is recorded on every closed trade, so the question “is this actually a scalper” has an arithmetic answer. This page uses our own data — the closed-trade lists published on each EA page, 9,339 trades with exits dated 2 January 2019 to 31 March 2026 — to show what the label does and does not predict, and what a short-hold EA has to clear to survive. None of the EAs published here is a scalper, and saying so is the point: the records are useful precisely as the control group. The runs and testing conditions behind them are published under the site’s methodology.

How it works: the trade the label describes

A scalping EA takes a small move and repeats it. The entry can be almost anything — a fast oscillator at an extreme, a micro-structure pattern, a rebound off the spread — because the entry is not the distinguishing feature. What distinguishes it is the exit distance: a target close enough that the trade is over in minutes.

Two consequences follow, and only the first is usually stated.

The first is high trade frequency. Short holds mean more completed trades in the same period, which is the “volume game” framing every scalping listing uses.

The second is the one that decides the outcome. A fixed cost is a variable fraction of a variable target. Spread, commission and slippage do not shrink because your target did. At the 2-8 pip target these EAs are typically sold with, a one-pip round-trip cost is 12.5% to 50% of the entire trade. The same one pip against a 100-pip swing target is 1%. Nothing about the market changed between those two cases; only the denominator did.

That is the whole mechanism, and it is why execution quality is not a footnote for this strategy. It is also why the strategy is so often profitable in a backtest and not in an account: the backtest is where the denominator is honest and the numerator is optional.

What the published records say about “fast”

Every figure below comes from our own data — the closed-trade lists on each EA page — with holding time measured from entry to exit on each closed row. Trades are sorted by exit. The cost column is explained in the next section.

EA	Symbol / TF	Trades	Per month	Median hold	Cost that zeroes the record
Peregrine	BTC/USD H4	163	2.90	9.12 h	14.78%
Zerqon	US30 H4	232	4.17	10.51 h	27.04%
Cairn	EUR/USD H1	4,583	80.02	19.06 h	20.75%
Iridescence	EUR/USD H1	872	10.03	23.82 h	14.81%
Tessera	USD/JPY H1	624	7.19	24.36 h	11.14%
Orrery	JP225 H1	195	3.45	29.66 h	12.26%
Lattice Weave	GBP/AUD H1	492	5.73	41.47 h	15.48%
Lanternfish	USD/JPY M5	407	4.70	50.91 h	14.12%
Beacon	AUD/CAD M15	384	4.69	58.52 h	15.78%
Tidewell Slack	GBP/JPY M15	303	3.49	65.77 h	10.43%
Kestrel Hover	USD/JPY M15	284	3.34	67.05 h	7.63%
Ballast	EUR/GBP H1	307	3.71	79.16 h	22.38%
Windrose	AUD/CAD H1	101	1.47	212.89 h	14.12%
Thunderhead	EUR/GBP H1	392	7.46	255.38 h	21.27%

None of them is a scalper. Not one has a median hold under four hours; the shortest is Peregrine at 9.12 hours. Only one EA takes more than 50 trades a month. The catalogue’s median holding time is 46.19 hours. That is the honest starting point, and it is why the table is useful here: it is a control group against which the label can be tested.

The chart period points the wrong way. Read the table top to bottom, sorted by holding time, and the two H4 records sit at the top while the M5 and M15 records sit near the bottom — the H1 group straddles both ends. Grouped:

• H4 — median hold 9.82 h (two EAs)
• H1 — median hold 35.57 h (eight EAs)
• M5 — median hold 50.91 h (one EA)
• M15 — median hold 65.77 h (three EAs)

The rank correlation between chart period and holding time is negative (−0.31 on the published rows, −0.21 if sub-minute rows are excluded). Both treatments give the same four group medians. A faster chart did not produce a faster trade; it produced a slower one. The mechanism is not mysterious: the chart period sets how often the rule is evaluated, while the stop and target set how long the position lasts. An M15 EA with a wide invalidation level and a distant target will sit for days.

One timeframe does not mean one behaviour either. Inside the H1 group alone the median holds run from 19.06 hours to 255.38 — a factor of more than thirteen, from EAs sharing a chart period.

The cost a strategy can absorb

Cost cannot be compared across these records in pips. A pip on BTC/USD, on JP225 and on EUR/USD are different quantities, and ranking them together would produce a number that means nothing. So the cost column above is expressed as a share of that EA’s own average winning trade: the round-trip cost, worth that fraction of the average win, at which the whole record reaches zero.

The result is unexpectedly tight. Across all fourteen the figure runs from 7.63% to 27.04%, with a median of 14.80%. Roughly one-seventh of the average win, and every strategy type on the site sits inside that band — trend, mean reversion and composite alike.

Two things follow, and both contradict the usual account.

Cost sensitivity is not a scalping property. If short holds were what made a strategy fragile to cost, the cushion would widen as holding time grows. It does not: the rank correlation between median hold and cushion is −0.10, which is close to nothing, and the sign is the opposite of the story. Kestrel Hover has the thinnest cushion in the catalogue at 7.63% while holding its median trade for 67.05 hours.

Nor is trade frequency. Cairn takes 80.02 trades a month, nearly eight times the next busiest EA, and has one of the larger cushions at 20.75%. The rank correlation between frequency and cushion is +0.38 — weak, and pointing the wrong way for the folklore.

What sets the cushion is the size of the target relative to the cost, and nothing else. That is why scalping is genuinely more fragile than the strategies here, and the reason is arithmetic rather than temperament: at a 2-8 pip target, a one-pip round-trip cost is 12.5% to 50% of the trade. The upper half of that range is outside the band that ends every record in this catalogue. A scalping EA has to clear, on every trade, a cost fraction that would have finished any strategy on this site.

Market conditions: when the edge survives the cost

The session argument that fills scalping listings — trade the London open, trade the overlap — is really the cost argument in disguise. Those windows are recommended because quotes are tight and stable there, not because price behaves differently. When the spread doubles, a strategy targeting eight pips has lost a quarter of its trade before it starts; a strategy targeting eighty has lost 2.5%.

This is also the honest reason to be sceptical of a scalping backtest before looking at anything else. A test configured with a fixed average spread and no commission has removed the variable that decides the strategy, and it will do so most flatteringly on exactly the EAs whose targets are smallest.

Parameters and settings in MT5

Input	Where the risk sits	What it controls
TakeProfitPips	The denominator of everything	The target; halving it doubles cost as a share of the trade
MaxSpreadPoints	The one input that must not be optional	Blocks entries when the quote is too wide to leave an edge
SlippagePoints	Silent when set too generously	How much worse a fill may be before the order is refused
CommissionPerLot	Usually absent from the tester	Must be modelled, or the whole record is measured without its main cost
SessionFilter	The cost argument in disguise	Restricts trading to hours where quotes are tight
MinTimeBetweenTrades	Cheap protection	Stops one signal firing repeatedly through a single wide-spread moment

Three MT5 realities decide whether a short-hold record survives:

• The tester’s spread model is the experiment. SYMBOL_SPREAD_FLOAT tells you whether the symbol floats; a run configured with the current fixed spread measures a market that does not exist. Use real tick data and read the modelling quality before reading the profit.
• The stop level bounds how tight a target can be. SymbolInfoInteger(symbol, SYMBOL_TRADE_STOPS_LEVEL) is the minimum distance the broker accepts for stops and targets. A strategy whose target is inside that distance cannot be placed as designed, and what runs instead is a different strategy.
• Commission is not in the profit column by default. Read HistoryDealGetDouble(ticket, DEAL_COMMISSION) and include it, along with DEAL_SWAP, when computing per-trade results. On a small target the commission is often larger than the edge.

Failure modes: how a scalping EA loses money

• The cost is a larger share of the target than the edge is. The dominant path, and the one that does not appear in an unmodelled test. It is not a losing streak; it is a constant subtraction.
• The spread widens where the EA trades. Signals that fire on volatility arrive with the quote at its worst. Testing on average spread hides precisely the fills that matter.
• Slippage measured in pips looks small. Half a pip is unremarkable against a 100-pip target and catastrophic against a four-pip one. The absolute number is the wrong unit.
• The target is tightened to raise the win rate. It works, and it moves the strategy toward the fragile end of the cost ratio. A rising win rate and a falling edge often arrive together.
• The chart period is treated as the holding time. On these records that assumption is wrong in direction as well as size, and it leads to buying a multi-day strategy as a fast one.
• Frequency is mistaken for the risk. It is not: across this catalogue frequency correlates positively with the cost cushion. The risk lives in the target size.

How to test a scalping EA in mt5depot before it costs anything

The useful test is not whether a scalping EA is profitable in a clean backtest. Almost all of them are, because the clean backtest is where the cost was removed. The useful test is where the profit reaches zero as cost rises, and how far that point sits from your broker’s real number.

Scalping versus the slower strategies

	Scalping	The strategies published here
What the label names	A holding time	An entry idea
Median hold	Minutes, by definition	9.12 to 255.38 hours
Cost as a share of the target	12.5%–50% at a 2-8 pip target	Ends the record at 7.63%–27.04% of the average win
What a backtest omits	The variable that decides it	A second-order correction
What to read first	Breakeven cost	Worst losing run

The comparison is not an argument that scalping cannot work. It is an argument about what has to be true for it to work, and about where to look. For a trend or mean-reversion EA the number that decides whether you can hold the position is the worst losing run. For a scalper it is the breakeven cost, because the losing run that ends it is not a run of bad trades — it is the spread, charged on every good one.