Kestrel EA Risk Management and Capital for MT4 Portfolios

Recommended capital per portfolio and level

Kestrel EA is a grid expert advisor for MetaTrader 4: it opens from its own signal, adds to the basket in small, capped steps when price moves against it, and closes the whole basket together at its target. It never closes a single trade at a loss; until the basket closes, the loss is floating. You choose your pairs — 1 to 4 of EURUSD, GBPUSD, EURCAD and AUDCAD — then Patient, Active or both for each, one set file per chart: Patient adds to a basket less often and holds a smaller worst case; Active adds more often, returned more in the backtest and reached a larger worst case.

The worst case is the larger of the worst floating loss and the deepest fall from an equity peak, in euros at fixed rates. From it come the four capital figures below, in ascending order: the two stop-out floors, the joint worst moment reading, and the one we recommend. They are calculation labels on a historical simulation, never requirements: the account is yours and the published figure is always a recommendation.

Record	Period	Worst case	Max open lots	No stop-out · Standard STP (20%)	No stop-out · Raw and ECN (50%)	If you size on the joint worst moment	What we recommend
EURUSD Patient	2010 – Sep 2026	€2,788	5.46	€3,000 21 Aug 2024	€3,400 21 Aug 2024	—	€8,500
EURCAD Patient	2014–2025	€3,408	3.98	€3,600 13 Mar 2017	€3,800 13 Mar 2017	—	€10,300
GBPUSD Patient	2010–2025	€3,885	6.64	€4,200 14 Nov 2024	€4,700 14 Nov 2024	—	€11,700
AUDCAD Patient	2010 – Sep 2026	€3,703	13.28	€4,100 22 Oct 2012	€4,600 22 Oct 2012	—	€14,000
EURUSD Active	2010 – Sep 2026	€4,946	14.95	€5,400 18 May 2017	€6,100 18 May 2017	—	€14,900
EURCAD Active	2014–2025	€5,887	8.17	€6,200 13 Mar 2017	€6,700 13 Mar 2017	—	€17,700
AUDCAD Active	2010 – Sep 2026	€4,699	21.86	€5,200 13 Feb 2011	€6,000 13 Feb 2011	—	€20,300
GBPUSD Active	2010–2025	€7,122	10.68	€7,600 31 Jul 2020	€8,400 31 Jul 2020	—	€21,400
K3 (AUDCAD Patient + EURUSD Patient)	2010 – Sep 2026	€3,704	13.30	€4,100 22 Oct 2012	€4,600 22 Oct 2012	€14,000	€22,500
K1 (AUDCAD Active + EURUSD Patient)	2010 – Sep 2026	€4,701	21.95	€5,200 13 Feb 2011	€6,000 13 Feb 2011	€19,000	€28,800
Full portfolio (AUDCAD Patient + AUDCAD Active + EURUSD Patient)	2010 – Sep 2026	€6,165	23.70	€6,700 4 Mar 2016	€7,300 4 Mar 2016	€25,200	€42,800

Simulated on Dukascopy data, Horizon level, euros at fixed rates; windows differ by pair and each row states its own. In every published record the worst case is the floating reading, not the fall from a peak. For a single record the joint-worst-moment column is the same figure as the recommendation, so it is left blank.

Summit multiplies every euro figure by five and Zenith by ten. The percentages do not change. The first two capital figures are limits, the third is one way of reading the record, and the fourth is our recommendation: between two and a half and nine times the balance that would have been closed out. Every combination of the eight records, with the same ladder: Kestrel Pair Combinations: What Fits Your Capital, or build your own set in the Kestrel Portfolio Calculator: Build Your Set and See the Capital.

What a stop-out is

Your broker lets you hold open positions against the money in your account. If the account stops covering those positions by enough, the broker closes them for you, at whatever price is on the screen. That is a stop-out, and the level at which it happens is set by your broker and by the type of account you opened.

It matters more here than in most strategies. Kestrel holds a basket open until it closes at its own target, so the loss is floating until then. A stop-out is the one event that turns that floating loss into a real one, and by definition it does it at the worst moment. So we publish, for every pair, variant, level and combination, the balance below which it would have happened in the record.

The two levels are the ones published by VT Markets, the broker where our demo accounts have been running since September 2026: a stop-out at 20% on Standard STP accounts, the usual one, and at 50% on Raw and Pro ECN accounts (VT Markets Help Centre, checked 17 September 2026). A higher level closes you sooner, so the Raw and ECN figure is the larger of the two, and it is the one to plan with.

Both are limits, not recommendations. At either balance the account spends its worst day one tick from liquidation, with no room for a swap charge, a wider spread, a gap, or an episode slightly worse than the worst one in this record — and a worse one remains possible. What we recommend is well above both: for a single pair, about two and a half to three times the Raw and ECN floor; for a combination, up to nine times.

One more thing the numbers say. In most combinations, the moment that would have closed the account is not the moment of the worst floating loss: it is a different day, when a smaller loss met a larger margin requirement. That is why we measure it hour by hour instead of estimating it from the worst day.

We are naming a broker here because these are its published numbers, not because we are recommending one. Stop-out levels differ by broker and by account type: check yours. You keep your account, your broker and your money.

The floor, the recommendation and the distance between them

The distance between the first step of the ladder and the last is the point of the table: the floor spends its worst day one tick from a forced close, and we recommend being several times away from it. A lower stop-out is not a better account, by the way — it only means the broker waits longer before closing the basket for you, and by then the loss it closes is larger.

And one limit of the figure itself: it comes from a simulation with a fixed spread and no commission, so it measures the stop-out rule, not the cost difference between a Raw or ECN account, which charges a commission on a tighter spread, and a Standard STP one, which charges a wider spread instead.

How we calculate it

Every figure starts from the simulation of each record over its window: the floating loss of the open baskets and the fall from the equity peak, each with its own date. The larger of the two is the worst case, and for a single record the capital we recommend is three times that — which is why the deepest fall reads at about a third of the recommendation in every single-record row: that third is the construction, not a property of the market.

For one record, the capital we recommend is three times the worst it ever went. For several, we add those recommendations up instead of sizing on the single worst moment they happened to share, because that moment is history and the next one has not happened yet.

Simulated monthly averages over the backtest, not an income expectation: €131 (EURUSD Patient) to €735 (Full portfolio) at Horizon, and five or ten times that at Summit and Zenith.

What happens below the recommendation

We checked every hour of the sixteen-year GBPUSD record as a possible opening day. With the capital we recommend, an account opened on any of them survived to the end of the record in both variants; even with €10,000 on Active — below what we recommend — it survives every start date, and below €8,000 it stops doing so. That is what the recommendation is for.

Margin at the worst moment

Like the kestrel that hangs still in the wind, the strategy’s cost is paid while holding — the recommendation exists so the hold is never forced to end early. On the capital we recommend at 1:500 leverage, the margin level at the worst moment reads between 333% (EURUSD Active) and 866% (EURCAD Patient) across the eight single records, simulated; on the €3,400 Raw and ECN floor, the EURUSD Patient episode of 21 August 2024 would have ended in a forced close instead.

Combinations sized on the added-up recommendations read higher still, because the capital grows faster than the shared worst moment: across the combinations we recommend by capital band, the margin level at the worst moment runs from 763% (GBPUSD Patient + AUDCAD Active) to 1,989% (all eight records), simulated at 1:500 on the capital we recommend.

Choose your level by capital

Summit scales the starting lot by five and Zenith by ten; the worst case, the floors, the joint reading and the recommendation scale with them. Do not move up a level without moving the capital up by the same factor: the same lot on five times less capital carries five times the drawdown.

From here it is five steps: choose pairs and variant, choose the level whose recommended capital fits your account, buy the annual licence (€999, €4,999 or €9,999, VAT included), install it on MT4 with its set file, and follow the record.

What these tables do not cover

• Swap and commission: the estimate is 900 to 1,400 dollars a year per Horizon portfolio, and it is an estimate, not a measurement.
• Slippage, and the broker stop-out actually firing — the floors above measure where it would have fired in the record, not what your broker will do next year.
• A future episode worse than every episode in the sample: it remains possible.

Beyond the backtest, the software has run on two demo accounts since September 2026 — a forward test on demo, not a real account. The two AUDCAD variants are described on Kestrel AUDCAD Expert Advisor for MT4: Patient and Active, which level fits a given account is the whole argument of Kestrel Forex Robot Price and Licence Levels, and the separation of the floating loss from the fall from the equity peak is on Is Kestrel a Safe Expert Advisor? Drawdown and Worst Float.