From two validated strategies to a portfolio you can live with

The certification walkthrough ends with one strategy and a prop-firm certificate. This one ends with a book: two real NinjaTrader 8 strategies on two instruments (WebinarTisHmaSlope on MNQ, goAlgoBosChoChStrategy on MGC), tested as a combination, held to rules you write. Every screenshot is from those runs.

The route

The question a portfolio answers
Two members, validated separately
Portfolio walk-forward — is the combination worth building?
Correlation measured out-of-sample
Earning membership
The fixed portfolio
Living with it
Certify against your own standard
The report

Before you read the numbers: everything below is hypothetical, model-based simulation — not a prediction or guarantee. Backtested results have inherent limitations, past performance does not indicate future results, and trading futures carries a substantial risk of loss. This walkthrough shows how the tool interrogates a combination of strategies; it is not a claim that any strategy or portfolio makes money. Full guide · see the disclaimer in every report.

1/The question a portfolio answers

Six stages, then a fork. The prop-firm route ends in a certificate; this route ends in a portfolio you intend to sit through drawdowns with. The validation underneath is the same — Optimize, Walk-forward, Audit, Perturbation — the difference is who sets the rules at the end. A prop firm's evaluation has a target, a daily loss limit and a trailing drawdown. Your book has the drawdown you would actually tolerate, the months underwater you would actually wait out, and the diversification you actually need.

AlgoCrucible home page showing the six validation stages and the two destinations: prop-firm certification and a portfolio — **Two destinations.** Same funnel, different ending. The certificate says "this strategy's history would have passed that firm's rules N% of the time." The portfolio says "this combination, held to my rules, would have stayed inside them N% of the time."

2/Two members, validated separately

Two strategies on two instruments. Each must earn its place on its own before they are allowed to share a book: an optimization, a walk-forward, an audit on two held-out periods and a perturbation of its settings. That per-strategy work is the whole of the certification walkthrough; here we pick up where it ends.

Runs page listing each member strategy's chain of runs: optimize, walk-forward, audit, perturbation, robustness, certify — **Two chains.** `WebinarTisHmaSlope` on MNQ, 15-minute bars, and `goAlgoBosChoChStrategy` on MGC, 5-minute bars. Both reached a ROBUST perturbation verdict — the one thing the portfolio page will insist on later. Validated on a laptop? Runs move between computers from this page (*Export selected* / *Import runs*), and a paid plan covers more than one machine.

3/Portfolio walk-forward — is the combination worth building?

Before freezing anything, ask whether these two have a joint edge on data nobody tuned on. A portfolio walk-forward is each member's own walk-forward plus a merge: every member enters as a search space, not a setting; every window re-optimizes it and trades the result on the next window; then the out-of-sample trades of all members are laid on one calendar. Members that already have a finished walk-forward are reused as-is — the merge takes seconds and needs no engine.

Portfolio walk-forward form with both members ticked, each carrying its search space, and the option to reuse a finished walk-forward run — **Each member enters as a search space.** Learn 120 days, grade 30, step 30. The *Walk-forward* column either reuses a finished run (#182: 20 windows, 549 out-of-sample trades; #247: 27 windows, 334) or re-runs the member on the engine. Every trade you are about to see came from data the optimizer hadn't seen.

Combined out-of-sample result of the portfolio walk-forward: 883 trades, net profit, profit factor 1.08, max drawdown, and the joint out-of-sample period — **The combined out-of-sample curve.** Joint period 2024-10-29 to 2026-06-17 — the span *both* members were tested on, stated on the screen. 883 trades, net +$2,772, profit factor 1.08, worst drawdown $2,703. This is the honest answer to "do these two have a joint edge?" — and it is *not* a setting you can trade. What it validated is the procedure "re-optimize every 30 days and trade whatever comes out."

4/Correlation measured out-of-sample

Correlation on one hindsight-fitted history flatters everything. Here it is measured on the out-of-sample trades only — the days each member actually earned while trading settings it had not been tuned on.

Correlation matrix of daily out-of-sample P&L between the two members, and the count of days both were in the market — **−0.01.** Low is the whole reason to hold both: one's drawdown is the other's ordinary month. The overlap count underneath says how often they were in the market at the same time — and the conflict scan says how many hours they were long and short the same contract family at once (here: none).

Next step panel on the portfolio walk-forward result explaining that a walk-forward validates a procedure, not a deployable configuration, and routing to the members — **The app says it on the screen.** A walk-forward validates a procedure. A portfolio you trade needs each member frozen at one setting that survived Audit and Perturbation — so the next step routes you to the members, not to a Deploy button. The tail check — the last 35% of the period graded on its own — came back *marginal*, which is worth knowing before Part 6.

5/Earning membership

The fixed portfolio has one entry rule: a strategy joins at a specific setting, and that setting must have passed Perturbation with ROBUST. Not "optimized well". Not "audited marginal". The neighbours of the setting held the edge out-of-sample, or the strategy stays out.

Portfolio page 'Add a strategy' dropdown open, listing eligible members and greying out one with the reason 'cliff edge, not eligible' — **Only ROBUST gets in.** The greyed entry had a profitable optimization and still isn't allowed in — its settings sat on a cliff edge, and a portfolio built on a mirage certifies a mirage. A member that stored its trades joins without a re-backtest; an imported trade record (a locked vendor algo, a Range-bar system, an Analyzer export from another machine) can join *by record* — and the portfolio then carries a banner saying exactly that.

6/The fixed portfolio

Now each member at exactly the setting you would put on a chart, with the contracts you would trade. This describes something real — which is why it needed the per-member gates first.

Portfolio page with both members added at their audited settings, one contract each, and the rules panel: no hedging, maximum correlation, maximum family exposure — **The book, described.** One contract each; rules: no opposing positions, pair correlation under 0.6. Both members came in on their stored trades — the build took one second.

Combined portfolio statistics: 597 trades, net profit, profit factor 1.35, max drawdown $1,650, with the out-of-sample robustness verdict and the Monte Carlo retest — **One equity curve.** 597 trades, net +$6,454 after $1/round-trip commission and 2 ticks a side of slippage, profit factor 1.35, worst drawdown $1,650 — with the worst stretch underwater drawn, not just totalled. Correlation of daily P&L −0.19. The held-out tail (last 35%) graded *robust* on its own (PF 1.23).

7/Living with it

Totals don't tell you whether you would have kept trading it. These do.

'Living with it' panel: profit per year, return divided by worst drawdown, longest time underwater, share of the period underwater, daily Sharpe and Sortino — **What holding it would have felt like.** 119 trading days was the longest stretch below a prior high; 68% of the period was spent below one; net profit ran to 3.9× the worst drawdown; daily Sharpe 1.4, Sortino 2.0. How long underwater, how much return per unit of worst drawdown, how the daily swings compared with the drift. These are the numbers that decide whether you would actually sit through a bad quarter — and they are the numbers your own standard, next, is written in.

8/Certify against your own standard

You set the rules. Not a prop firm's — yours: the account it trades in, the return that justifies the work, the drawdown you would actually tolerate, the months underwater you would wait out, how much of the profit may come from one member. The standard is saved and judged against the portfolio walk-forward — the out-of-sample procedure — because that is the evidence that wasn't fitted.

Portfolio standard editor with the user's own limits filled in: $25,000 account, 6% a year, recovery factor 1.0, drawdown 12%, 200 days underwater, correlation 0.4, two members, all members Perturbation-passed — **"My 25K micro book".** A $25,000 account; at least 6% a year; net profit at least equal to the worst drawdown; drawdown under 12%; no stretch underwater longer than 200 days and no more than 85% of the period below a prior high; pair correlation under 0.4; at least two members, none earning more than 70% of the profit, every one profitable and Perturbation-passed; no hedging. Each rule on the form explains what it measures and what tightening it costs.

Verdict of the portfolio walk-forward against the user's standard: FAIL, 2 of 15 rules — return per year 5.0% against 6%, 86.6% of the period underwater against 85% — with three warnings and ten passes listed — **FAIL — 2 of 15.** Return per year 5.0% against the 6% bar; 86.6% of the period spent below a prior high against 85%. Three more rules passed by a hair (recovery factor 1.03, drawdown 10.8%, 187 days underwater). Diversification and member quality all clear. This is the honest reading: a combination with a real joint edge that doesn't yet earn its place on *this* account at *these* sizes — and the screen says exactly which bar it missed.

And against a prop firm's rules, for contrast. The same fixed book through a 50K one-step evaluation model: pass odds 89%, bust risk 11% (all of it trailing drawdown), a median of 5½ months to target. Below it, the cost-sensitivity slider: the book breaks even at about 8½ ticks a side — the result survives four times the slippage it was priced at. Both are estimates from history, not predictions — and the fact that a book can clear a prop preset and miss your own bar is the point of writing one.

9/The report

Everything on one document — members, their settings, the verdict each one earned, the correlation, the rules tested and the odds — with the methodology and its limits stated on the page.

First page of the printable portfolio report with members, settings, verdicts, correlation, certification results and the full-parameter settings appendix offered — **Save as PDF.** The full-parameter settings appendix records every input of every member as captured when the runs started — so the document describes what was tested, not what you remember testing.

∴Which route is yours?

If you'll re-optimize on a schedule

The portfolio walk-forward is your evidence. You are not trading a setting, you are trading a procedure — re-fit every N days, trade what comes out — and the out-of-sample curve, correlation and your standard's verdict describe that procedure. Run it again when you change the schedule or the members.

If you'll set parameters once

Then the fixed portfolio is the thing you trade, and every member must earn its seat with a ROBUST perturbation first. Certify the book against your own standard, and against any evaluation you intend to sit — and re-certify when a member's neighbourhood stops holding.

The reasoning behind the two routes, and why a walk-forward result is never a deployable configuration, is in the Help Center: Two routes to a portfolio.

Build your own book

The full pipeline — Optimize, Walk-forward and Audit — is free for 30 days on up to 3 strategies. Perturbation, portfolios and certification are in the Trader plan, which covers two computers — validate on the laptop, assemble on the desktop.

Download AlgoCrucible The single-strategy walkthrough →

Results shown are hypothetical, model-based estimates computed from limited historical data under simplifying assumptions — not a prediction or guarantee that any strategy or portfolio will be profitable or that any evaluation will be passed. Evaluation profiles are generic models of common industry structures, not the official rules of any firm; AlgoCrucible is independent and not affiliated with, endorsed by, or sponsored by any proprietary trading firm or by NinjaTrader. Trading futures carries a substantial risk of loss and is not suitable for everyone. Not financial advice.