Getting Started

Portfolio Construction helps you compare, stress-test, optimize, hedge, and rebalance multi-asset portfolios. Most tabs use the same basic pattern: choose a date range, define one or more portfolio groups, click Run, then inspect charts, tables, and the Navigate to... dropdowns for follow-up analysis.

Where To Start

Intro

Intro explains the core investing ideas behind the tool: diversification, long-run compounding, risk, drawdowns, and why mixing different return drivers can matter more than simply adding more assets. Its examples use longer-history simulated proxies for global equities and compare both DBMF and KMLM trend-following proxies.

Backtest

Replay one or more portfolios over history. Use it after choosing an allocation to understand growth, drawdowns, rolling performance, correlation, and beta through different market regimes.

Output: growth, drawdown, rolling metrics, correlation or correlation matrix, rolling beta versus SPYSIM, and summary metrics.

Frontier

Search many allocation mixes and mark the portfolios with the best return/risk trade-offs. Use the scatter chart to see whether a high-return allocation is taking much more risk than alternatives.

Output: optimal portfolios for Sharpe, return, volatility, Sortino, drawdown, and worst drop. Click any scatter point to inspect weights or send it to another tab via Navigate to… — the per-asset maintenance, buffer, and leverage cap carry across so the Margin tab reproduces the same numbers.

Rolling optimal allocations. The rolling-window optimizer (best Sharpe / volatility / Sortino per window) scores allocations on a lag-adjusted covariance, the same Scholes-Williams correction used for beta and correlation. This matters for books that mix assets trading on different exchange schedules: without it, a fund closing hours before the others looks more diversifying than it really is and gets over-allocated. The correction is a no-op for assets that trade on the same schedule.

Funded constrained results: the dense cloud is a fast screening population for visual distribution; funding currencies are clipped to their common live calendar before any funded evaluation. The objective optima (stars) and any point you click are solved exactly on the full funding path, searching round currency blends within the Step % grid, so a screening estimate is never treated as a funded result on its own.

Margin proximity: funded optimum charts show the realized equity-to-assets ratio from the selected funding path, including FX-revalued debt and resets. Safe leverage is sized against a stricter intraday-stressed path, so the plotted minimum can remain above the sizing threshold.

Reading the scatter: every dot is a screening estimate, colored by its safe leverage, for visual distribution. Stars mark the best funded blend for each objective (the same portfolios the summary table lists), drawn at their position in the cloud. Clicking any dot exact-solves that portfolio's funded path before showing its details, so the numbers you read match the summary and what the Margin tab reproduces on hand-off.

Forecast

Estimate a portfolio's future value range by resampling contiguous historical daily return blocks. Asset returns and the USD risk-free-rate row are sampled together, so each synthetic path keeps the historical co-movement observed inside sampled blocks.

Output: aggregate metric percentiles (incl. Max SWR), a monthly portfolio-value range chart, and an Income card you can switch between income-vs-success %, income-by-confidence bars, and a depletion fan.

Margin

Estimate how much leverage a portfolio can carry before a historical loss would breach maintenance margin. Use this for leverage sizing rather than the simpler Leverage tab.

Output: safe leverage, margin-call proximity, leveraged growth, drawdowns, top episodes, weight drift, and hedge effect when enabled.

Funding and overlays: funding bounds must permit weights totaling 100%. Base and hedged tracks select blends independently; Match leverage applies the lower leverage afterward while retaining both blends. With TIMS enabled, scenario margin reduces the safety cushion only. Displayed hedged return and drawdown use pre-TIMS equity so margin demand is not counted twice. An infeasible result means no tested blend remained safe even at 1x on the selected history.

Margin-call days: proximity charts and counts use the realized funded equity-to-assets ratio. A day is counted only when that ratio is below the portfolio's raw maintenance requirement. MC Buffer changes the stricter threshold used to size leverage; it does not redefine an actual maintenance breach.

Hedge

Compare put-backspread overlay settings on top of a base portfolio. The goal is to reduce severe drawdowns without turning the overlay into a separate long-option sleeve.

Output: baseline versus hedged growth, drawdowns, optimal overlay settings, episode cards, and rolling views of which settings worked best over time.

Currency

Test how a USD-denominated portfolio would have behaved when funded through USD, EUR, GBP, JPY, CHF, or a mix of those currencies. Results combine portfolio returns, interest-rate carry, and FX moves.

Output: ranked funding mixes, optional scatter chart, policy-rate chart, pair FX chart, and funding-return correlation.

Interpretation: safe leverage is fitted to the same historical window shown in the result, including an intraday loss allowance. It is not a forecast or a guarantee against future margin calls. In sampled mode, the full cloud is first evaluated approximately at 2x leverage; only the strongest candidates and reference rows receive the exact path-based safe-leverage solve. Approximate or infeasible rows can remain visible in the scatter, but they are excluded from rankings and detail actions. Tight bounds can produce skipped candidate IDs when no valid mix is found within the fixed rejection budget.

CWARP

Candidate Weighting And Risk Parity ranks candidate assets by how much they improve a base portfolio. Use it before adding a new sleeve.

Output: CWARP score, ranked candidate table, and volatility-versus-return scatter.

Match

Put portfolios on equal footing by finding the leverage needed to match a target volatility, CAGR, or max drawdown.

Output: matched leverage, metrics before and after matching, growth chart, and drawdown chart.

Leverage

Sweep simple fixed leverage levels on a portfolio to see the broad risk/return trade-off. This tab is for quick exploration; use Margin when margin-call risk matters.

Output: table and chart of CAGR, volatility, Sharpe, Sortino, and max drawdown at each leverage level, using the portfolio's selected funding mix.

Correlation

Check whether assets or portfolios really diversify each other. Use two groups for rolling detail, or three or more groups for a matrix view.

Lag-adjusted beta and correlation. Both are computed with a lead/lag correction (Scholes-Williams) rather than a naive same-day calculation. The reason: a fund listed on a different exchange than the benchmark closes at a different local time (for example, a European-listed S&P 500 ETF on Xetra closes hours before the US close), so part of each day's market move lands in the fund's next-day return. A same-day beta would understate its real exposure (it can read near 0.5 for a fund that truly tracks the market one-for-one). The correction adds the neighboring-day benchmark relationship to recover the true economic beta and correlation, and normalizes so the benchmark scores exactly 1.0 against itself. For assets that trade on the same schedule as the benchmark the adjustment makes effectively no difference.

Rebalance

Turn current positions, cash, target weights, and margin debt into concrete buy, sell, borrow, repay, and hold actions.

Output: trade tables, before/after pies, financing bars, currency debt rebalance, priority ranking, and risk metrics comparing current and slider-adjusted portfolios. In the growth chart, color separates before/after and line style separates unlevered/leveraged. Exposure slider changes update the key metrics in the background, cash-only buys can be redirected among eligible rows, and the risk metrics table can send all four before/after unlevered/leveraged paths to Backtest or send pairs to Backtest, Match, or Correlation.

Trend

Inspect current positioning for KMLM, CTA, and DBMF. Use it to understand what managed-futures funds are currently long or short, how concentrated the exposures are, where funds overlap, and which normalized instruments have shifted across local snapshots.

BH Macro

BH Macro Limited is a closed-ended LSE fund feeding the Brevan Howard Master Fund, with separate GBP (BHMG) and USD (BHMU) share classes. This page compares the listed share price against the published NAV and shows how that premium/discount behaves around volatility spikes. Pick a class and a date range to zoom in; every chart and table follows the same From / To range.

IBKR

Parse an Interactive Brokers HTML Activity Statement locally in your browser and summarize expenses, market-data subscriptions, pricing-plan clues, and selected French tax figures. Non-EUR rows are converted with Banque de France ECB exchange rates, and only the currencies and dates needed for the selected year are loaded into the report view.

Box

Price short box spreads for SPX, ESTX50, Z / FTSE 100, N225, and SMI. Enter either credit points or rate, and the tab solves the other side for the selected index contract.

Default strikes are refreshed around the latest index level and sized to approximately USD 50,000 in each settlement currency. Shared or manually entered strikes always take precedence. Z credits must use 0.5-point increments; rate-derived Z credits are rounded to that valid tick.

Output: implied rate, equivalent credit, notional, and a comparison against the matching 1-year interest-rate-swap rate (the term borrowing benchmark). The Close 10% OTM card beside DTE shows a suggested positive debit close limit set 10% below the current gross credit and rounded down to the valid credit tick. It does not construct another order. The rate chart plots each currency's 1-year swap level.

Put

Plan buying shares by selling a cash-secured in-the-money put instead of buying outright. Selling the put commits you to buy 100 shares per contract at the strike if assigned, in exchange for the premium, so your effective entry is strike minus premium. Enter one ask and the tab lays out a ladder of integer strikes; type the chain premium per strike to compare.

Output: per integer strike (from the at-the-money/in-the-money strike up to the configured ITM cap), the breakeven premium (strike minus ask, the premium at which the effective entry equals the ask), and once you enter a premium, the effective entry, the entry delta versus buying at the ask (dollars and percent), the premium credit, the final cost, and the total delta. Dollar totals scale with Quantity. Negative deltas mean a lower entry than the ask and are green; positive deltas mean the effective entry lands above the ask and are red.

Forex

View major policy rates and FX pairs without setting up a portfolio. The From/To fields also act as zoom controls.

Assets

Browse assets by return driver, then inspect one asset at a time. Use it to confirm available history, live date, maintenance margin, and basic performance before adding an asset to a portfolio.

Output: asset classes, ticker, currency, date range, live date, maintenance margin, CAGR, annualized volatility, Sharpe ratio, maximum drawdown, and related dataset links when variants exist.

Upload CSV

Add your own price history from the header Upload CSV button. Uploaded assets appear in autocomplete and can be used across tabs.

MCP endpoint

AI assistants can query the same market data the app uses, directly, over the Model Context Protocol. Point any MCP client at /mcp on this host (for example https://<host>/mcp); it speaks streamable HTTP. The endpoint is read-only — it exposes data, never trades or computations.

Docs

This page is the user guide for the app. Use it to choose the right tab, understand the main inputs, and interpret the major outputs without leaving the app.