Live trade sequence replayed across six capital levels.
Three cost models applied. Slippage drawn from 4,957 actual broker fills.

We did not model blind.
We pulled 4,957 actual fills.

Before scaling a single trade, we went into the IC Markets EU execution log and extracted every take-profit fill from the live record. Exit fill price against requested take-profit price, divided by mid, multiplied by ten thousand. That is the slippage in basis points for every exit. These figures — not industry assumptions — became the calibration anchor for both cost scenarios.

How is slippage modelled — and how conservative is it?

The strategy exits exclusively on limit orders — take-profit targets placed at the time of entry. Limit order exits do not push price. They do not manufacture their own slippage. This property holds at $30M and remains true at $100M in trillion-dollar daily turnover markets. The execution log is the proof. Limit orders are also unaffected by spread widening — they wait for their price.

Slippage distribution across 4,957 take-profit exits · IC Markets EU MT5 · Developer-Attested API Record

What this shows

Each bar is a bucket of execution outcomes from real live trades. 84.1% of all exits landed within 0.5 basis points of the requested price. This matters because the capacity model is only as credible as its slippage assumptions — and these assumptions are grounded in 4,957 actual fills, not estimated. The near-zero median (0.15 bps) confirms X45 exits passively, not aggressively. Outliers above 5 bps (70 total) are clustered at news events in a single pair — not structural.

Three scenarios. Built to break
The thesis if it can be broken.

Every trade in the live record was scaled proportionally across six capital levels. Real commissions and swap costs from the broker log were applied at scale. Slippage was modelled three ways — one conservative, one deliberately punishing, one optimistic ceiling. CAGR is annualised across 19 active trading months. The honest calculation.

Net annualised return across six capital levels under each cost model. Denominator: 19 active trading months. All figures derived from the same 4,999-trade live record scaled proportionally. Linear CAGR is flat across all levels — proportional drag means proportional cost, so the annualised rate is invariant to capital. Nonlinear crosses linear at ~$24M where power-law compounds faster.

The linear floor holds flat at 29.3% across all capital levels — proportional scaling means proportional cost, so the annualised rate is capital-invariant. The nonlinear model starts above the linear floor at small scale but crosses it at approximately $24M AUM, where the power-law exponent causes drag to compound faster than capital growth. At $30M: nonlinear generates $3.7M of additional drag versus $3.5M linear. The stress floor is the harder test above $24M. The thesis holds under both. At $30M, X45 tolerates approximately 4 basis points of all-in drag before the economics break.

The nonlinear effective rate grows with capital multiplier M = C / $7,700. Formula: eff_bps = 0.20 × M^0.2. Below ~$24M the nonlinear model is cheaper than linear. Above it the power-law bites harder — making it the genuine stress floor at institutional scale.

All drag figures from live engine. Base NAV $7,700. Exponent 1.2 recalibrated Sep 2026 (Jawad review). $30M row = proposed capital level only — not a confirmed safe figure.

Net CAGR across six capital levels · Zero (35.5% flat) / Linear floor (29.3% flat) / Nonlinear stress: 32.2% at $1M → 31.0% at $5M → 30.3% at $10M → 29.5% at $20M → 29.0% at $30M → 28.3% at $50M · Exponent 1.2 · Modelled from 4,999 live trades

How to read this

Three lines, six points each. The gold flat line is the theoretical ceiling — zero additional friction. The lower flat line is the linear floor — 29.3% at every capital level because proportional drag is invariant to size. The descending line is the stress test: nonlinear cost compounding with capital. It starts above the floor at small scale (M: 32.2%) and crosses below it at ~4M — the only region where institutional size creates a genuine cost penalty. At 0M, the stress floor is 29.0%: 0.3pp below linear. The thesis holds.

Where does the nonlinear model cross the linear floor — and why does the crossover point matter?

The point at which increasing capital begins to degrade the CAGR curve — plotted across all three cost scenarios from $1M to $50M. The frontier shows where, and by how much, the edge compresses. Zero friction is flat. Linear is flat. Nonlinear curves downward — the power-law compounds. The crossover at ~$24M is where nonlinear becomes the binding constraint.

Capacity frontier · CAGR degradation $1M–$50M · Crossover derivation: eff_bps = 0.20 × M^0.2 = 1.00 → M = 5⁵ = 3,125 → C* = 3,125 × $7,700 = $24.06M · At $30M: Zero 35.5% / Linear 29.3% / Nonlinear 29.0% · 6.5pp cost-sensitivity envelope

What the frontier reveals

The frontier shows where each additional dollar of capital begins to cost more than the last. For zero and linear models, there is no frontier — drag is flat or proportional, so CAGR never bends. For the nonlinear model, the curve bends downward at scale. The crossover at 4M is mathematically derived: when the effective nonlinear rate (0.20 × M^0.2) equals the fixed linear rate (1.00 bps), the models cross. Above that, nonlinear is the harder constraint. Below it, the stress test is actually cheaper than the floor case.

Execution drag ($M) by capital level · Linear (1.00 bps fixed): $0.12M at $1M · $0.58M at $5M · $1.17M at $10M · $2.34M at $20M · $3.50M at $30M · $5.84M at $50M · Nonlinear (0.20 × M^0.2): $0.06M · $0.43M · $0.98M · $2.25M · $3.66M · $6.76M · NL cheaper below $24M, NL = stress floor above

Why absolute drag matters alongside CAGR

CAGR percentages alone can obscure the dollar cost of friction. At 0M, .66M of nonlinear drag is 60k more than the .50M linear floor — that marginal difference is what exponent 1.2 adds versus a proportional model. Below 4M, the nonlinear model is cheaper in absolute terms than linear. The lines cross because the power-law effective rate (0.20 × M^0.2) rises slowly — it only exceeds 1.00 bps at M > 3,125, i.e. capital above 4M. This is not an assumption; it is a derived threshold.

Observed equity curve · 4,999 live trades · IC Markets EU · 21-month window Feb 2024 – Nov 2025 · TWR 51.34% · CAGR 31.28% (historical basis)

This is the foundation

Every calculation in this analysis descends from this curve. 4,999 trades. No backtest. No simulation. IC Markets EU live account, Feb 2024 to Nov 2025. The three-month pause (Nov 2024–Jan 2025) is the governed seasonal blackout — not a drawdown event. It is annotated, not hidden. TWR of 51.34% is the return a passive investor would have seen. CAGR of 31.28% is annualised on an MWR basis across the 19 active trading months. These are the inputs to every scenario in this report.

The capacity test confirms what
The architecture already implied.

These are not incidental properties. The algorithm was built on structural characteristics that are scale-invariant by design. Each is verifiable from the live execution record.

Account structure analysis · Single vs multi-account deployment · Simultaneous and staggered execution modelled · $30M total capital

Does splitting across accounts reduce cost?

Under a nonlinear impact model, smaller per-account positions carry a lower effective rate than a single large account. The multiplier M = C/NAV is smaller per account, so eff_bps = 0.20 × M^0.2 is lower. This means the same total capital, split across multiple accounts and staggered, produces less aggregate drag than a single consolidated account. The graph shows how that benefit scales from 1 to 30 accounts at 0M total. Single account at 0M: worst case. Multi-account staggered: measurably better.