Portfolio Studio

Portfolio Studio is the workspace for a single saved portfolio. Where the Portfolios Hub is about building portfolios, the Studio is about understanding and preparing one: what it does, how its capital should be split, and whether it can survive the real world.

Where to find it: In the app sidebar, open Portfolio Studio — it opens your most recent portfolio (or pick one from the Library). The Studio has three main tabs:

Every portfolio also earns a portfolio-level CSI — the same 0–100 scale as individual strategies, so you can directly compare a portfolio to the strategies inside it.

The example portfolio

Your Library includes AlgoChef Sample: Five Markets, an example portfolio built from five of your sample strategies (S&P 500, Nasdaq, gold, natural gas and the yen, one contract each, $250,000 starting capital). Open it to see every Studio tab filled in before you build your own.

It is marked Example · read-only: like the sample strategies, it can't be renamed, deleted, resized or rebalanced, and it doesn't count toward your plan's portfolio limit. To experiment with it, choose Duplicate from its row menu in the Library; the copy is an ordinary portfolio you can change freely.

Analyze Tab

The default landing tab. It reads top-to-bottom like a briefing:

Portfolio Summary

A row of chips at the top gives the headline results of the historical backtest: net profit, return, win rate, CSI, Sharpe, max drawdown, diversification and the number of active strategies. The portfolio's starting capital sits in the card's top corner, because every dollar and percent figure is measured against it. An Edit Portfolio button lets you adjust the composition without leaving the page.

At the bottom of the card, Robustness check shows the verdict of the latest robustness run you saved (Robust, Caution, Needs more capital or Not Robust) and when it was saved, with a link to the Robustness tab for the details. A backtest can look excellent and still be fragile, so read the two together. If no run is saved yet it says Not yet stress-tested. It warns when the saved run may be out of date: made with an older version of the check, or at a different starting capital than the portfolio has now. Robustness runs are kept only when you save them, so save the run you want shown here.

Combined Equity Curve

The portfolio's merged equity curve with a $ / % toggle and brush zoom for inspecting specific periods. This is the curve that matters — individual strategy curves can look ragged while the combination is smooth (that's diversification working).

Contracts Open Over Time

A stacked bar chart of how many contracts each strategy holds through time. Reveals concentration moments — periods when many strategies are positioned at once and your margin and risk usage spikes.

Drawdown Analysis

Portfolio-level drawdown depth and duration, following the same $/% mode as the equity chart. Compare it against the worst single strategy's drawdown — measured on your whole account, at the same level of day-to-day risk as the portfolio — to see how much combining strategies has helped historically.

Strategy Allocation

How capital and risk are distributed across members. A sortable allocation table is always shown, with a Bars / Treemap toggle beneath it and a $ / % switch for the figures.

Period Performance & Calendar Returns

Performance sliced by period, plus a month-by-year heatmap — the portfolio version of the same views in the Quality Report.

Strategy Breakdown

Detailed per-strategy metrics inside the portfolio context: who's contributing profit, who's contributing risk, and who's dead weight. This section also holds the Correlation Matrix, a heatmap of how each pair of strategies moves together (see How portfolio numbers are measured).

Risk Decomposition

Splits portfolio risk by source. A portfolio where one strategy contributes 70% of the risk isn't really diversified, no matter how many names are in it.

An insights footer links you to the logical next steps in Allocate and Robustness.

How portfolio numbers are measured

• Sharpe ratio is the portfolio's average daily result divided by how much it swings day to day, scaled to a year. Nothing is subtracted for interest rates: futures profits already come on top of what your cash earns. So the Sharpe ratio is the same whatever starting capital you enter.
• Correlation between two strategies is measured only over the days both were trading. If two strategies share less than about six months of history, or one of them did not trade at all during the shared period, the cell shows "—" (not enough shared history) instead of a number, and those pairs are left out of the average. If no pair can be measured, the diversification score shows "—" too.
• Risk Concentration (an optional column in the Portfolios Library) is how much of the portfolio's day-to-day swing comes from a few strategies: near 1 means one strategy drives almost all of it, lower means the risk is spread out. It is measured on each strategy's actual dollar results and uses the same correlation as the heatmap, so a strategy that started trading later counts its full swing rather than a diluted one. It has been measured this way since October 2026.
• "vs best single strategy" and "vs worst single strategy" use the same formula and the same starting capital as the portfolio itself. The drawdown comparison sizes the single strategy to swing as much day to day as the portfolio, so the two falls are compared at the same level of risk.

All of these are historical, simulated figures. They describe the past and do not predict future results.

Allocate Tab

Compares capital-split methods and sizes your portfolio in whole contracts. It's a big enough topic to have its own page — see Allocation Methods for every method and setting explained.

In short: you set your capital, pick methods from three tiers (Foundation, Professional, Institutional), click Run Comparison, and evaluate the results in Candidate Cards (one per method, each with its change vs your current sizes and a small equity sparkline), a Comparison Table, a Risk / Return Map, an Equity & Drawdown overlay, and a Capital Split chart. Save as new portfolio turns a result into a brand-new portfolio and leaves this one unchanged; to resize this portfolio itself, use Apply Rebalance on the drift strip.

Margin budget. Contract sizes are based on your capital and a margin budget: the share of your capital that may be used for exchange overnight margin. The default is 25%, and you can set anything from 5% to 100% under Advanced Settings. Each result shows how much margin it uses, how much buffer would be left after the worst drawdown in its own history, and a simulated stress check with margins 50% higher. Margin figures are estimates based on exchange overnight margin. Your broker may require more, and past drawdowns do not limit future ones. Portfolios saved before margin budgets existed keep their contracts exactly as they are.

Quick Build uses the same sizing and the same margin budget, so a new Quick Build can have different contract counts from one you built earlier.

How sizing uses correlation

Since 2026-10-07, sizing measures how strategies move together the same way the correlation heatmap on the Analyze tab does.

• Correlation is measured only over the period two strategies actually traded together. Before, days when only one of them was trading were counted too, which made strategies with different start dates look less related than they are. When two strategies share too little history to measure (a "—" on the heatmap), sizing uses the average of the pairs that could be measured.
• Each strategy's swings are compared per dollar of margin, when every strategy in the portfolio has an exchange margin figure. A strategy on a large contract and one on a micro contract are then put on the same footing before the sizes are worked out. If any margin is missing, swings are compared per contract instead.
• Methods affected in Allocate: HRP, Max Diversification, Markowitz and Black-Litterman. New runs can recommend different contract counts than runs made before this date.
• Quick Build "Diversified" now spreads risk across strategies that move differently, so strategies that move together share a slice. Before, it split capital evenly.
• Since 2026-10-08, HRP and Quick Build "Diversified" give each independent bet an equal share of the risk. Strategies that move independently each carry about the same risk, and strategies that move together count as one bet and share it. Before, these methods leaned toward the strategies that swing least for the margin they use, which could leave a few strategies carrying little of the portfolio's risk. Portfolios saved earlier keep the sizing they were built with and do not show a "Rebalance" because of this change.
• "Minimum size." When a strategy's share is too small for one contract, it is raised to the smallest tradeable size and its row in Quick Build says so (when every strategy has a margin figure; otherwise the build lists it in its warnings). That strategy then uses more margin than its share alone would give it.

The most any one strategy can take in a new Quick Build depends on how many you pick, for every Quick Build method:

So Equal Risk and Edge-Weighted builds with 2–7 strategies can size differently than before.

What does not change: portfolios you already saved keep their sizes. Their rebalance check keeps using the method they were sized with, so they won't show "Rebalance" because of this. Nothing changes unless you run again and save. Robustness runs saved before 2026-10-07 carry an Older method badge: run the check again to compare with today's results.

All sizes come from historical, simulated figures. They describe the past and do not predict future results.

Robustness Tab

One configuration panel, one Run Robustness Check button, one integrated answer. This replaced the older separate backtest tools (walk-forward, Monte Carlo, and stress testing now run together as a single check).

Configuration

The ruin rule for portfolios

Risk of ruin is the share of simulations that hit the rule, using the same rule as Monte Carlo Capital Sizing.

• Margin and Margin + Max drop from peak are offered only when every strategy in the portfolio is futures with a margin requirement set and recorded contract counts. The margin is the portfolio's peak margin: the most margin its positions need open at once, across contracts. The ruin line is that peak margin plus the buffer (default 10%).
• Otherwise the only option is Floor + Max drop from peak: ruined if the whole portfolio goes below a dollar floor (default $3,500; $0 means no floor) or falls the chosen % (default 50%) or more below its highest point so far, and the panel says why.
• Example (simulated): peak margin of $40,000 with a 25% buffer puts the line at $50,000.

Runs saved before the current rule (including runs saved before the dollar floor was added) keep their old numbers and are labelled as calculated with the previous rule. Run the check again to update them.

Results — the five pillars

1. Robustness Verdict — the headline: Robust, Caution, Needs more capital (holds up in every stress test, but your capital is below what your risk targets need; the card shows both amounts) or Not Robust. Includes "How Much Money Does It Take?" at a glance.
2. Capital Sizing — required capital to meet your risk targets, the binding constraint (which target is the limiting factor), and the Out-of-Sample Haircut — how much more capital the conservative OOS view demands. A goal-seek panel lets you adjust targets and re-solve.
3. Outcome Distribution — the spread of simulated outcomes: percentile bands for profit and drawdown across all Monte Carlo runs.
4. Stress Scenarios — four made-up shocks applied to the portfolio's own trade history: correlations jump to 0.8, the most volatile quarter's swings double, the biggest earner stops working for the last quarter of the history, and margins rise 50%. These are not replays of real historical crises.
5. Out-of-Sample Validation — splits history into earlier and later periods and checks whether the same portfolio performs about as well later as it did earlier. Strategies with too few trades in an earlier period are left out of that window, and the section names them.

Saved Robustness Runs keeps prior runs so you can reload and compare configurations without re-simulating.

Rotation

Instead of trading every member strategy at once, Rotation keeps a bench of validated strategies and actively trades only the top slice of them, re-ranked on a schedule.

Rankings are point-in-time. At each re-pick, the ranking sees only trades that had already closed by that date. This is the heart of the rebuilt engine: an earlier version ranked on all-time scores, which included trades that had not happened yet at the simulated decision point and quietly flattered every backtest it produced. A rotation result is only worth anything if the simulated decision could actually have been made with the information available at the time.

Configuration. Six ranking methods: Quality, Quality (gated), Momentum, Risk-adjusted, Drawdown-conscious, and Custom blend. Health acts as an eligibility filter and as one input to the custom blend, rather than as a method of its own. You set the re-ranking cadence and the ranking look-back independently — they are two separate dials — and each rotation window is then sized by any method from the Allocate library.

Churn control. Four independent rules can block a swap that the raw ranking would otherwise make: a buffer on ranks, a minimum holding period, a cap on turnover, and a minimum score gap before a replacement is allowed. Each window records how many swaps each rule blocked, so you can see whether the brakes are doing useful work or just getting in the way.

Transaction costs are modelled per rebalance event, so a re-pick that changes nothing costs nothing. Costs are off by default; switch them on to see the fee drag alongside the headline result.

Results are always a three-way comparison, never a single number. Every run replays the same trades three ways through identical maths — rotating, holding everything with the same capital, and picking once and never rotating — and reports nine metrics across all three. That is deliberate: rotation is only worth doing if it beats simply holding the lot, and the comparison makes a negative answer as visible as a positive one.

Rotation ships with named presets so you don't have to tune every dial yourself. The recommended default is AlgoChef Standard — tier-gated quality selection, quarterly rotation, and a five-year ranking history — the configuration that held up best across the rotation lab's cadence tests. Pick a preset to populate the whole configuration, then adjust from there.

Rotation History shows the windows of your most recently saved rotation run — what was held when, and what changed at each re-pick. It is a record of a saved simulation, not a live log of trades you placed.

Common Questions

Analyze, Allocate, Robustness — in what order? That order. Understand what you own (Analyze), decide how to size it (Allocate), then verify it survives (Robustness).

Do the tabs affect my saved portfolio? Analyze is read-only. In Allocate, Save as new portfolio creates a separate portfolio and leaves this one as it is; only Apply Rebalance on the drift strip changes this portfolio, rewriting its contract counts in place after a confirmation. Robustness never modifies the portfolio — it saves separate run results.

Why does my portfolio's CSI differ from its strategies' scores? The portfolio is scored on its combined equity curve. Diversification can make that curve steadier than any member's, or correlation can make it rougher, so the portfolio's score can sit above or below its members'.