Inspect portfolio candidates and export a trading configuration

Select a run and a candidate#

Open Runs in Portfolio Input. The list supports text/status filtering and day groups. A row menu can open the result, pin it in a result-context tab, load its request into Builder, download CSV/JSON, copy its ID or show its results folder where supported.

Related -r<N> runs can appear as a suite. A suite is a read-only aggregate of separately completed runs. Its candidates retain source-run identity; a candidate ID alone may not be unique across the suite. Download a specific run from Runs when suite download is unavailable.

View full-size screenshot ↗

The result-context chips are separate from workspace tabs. Summary · best follows the selected/best candidate, while a candidate chip pins one exact source candidate. Check both identities before comparing numbers.

Read the result views#

Normal application builds provide nine views. The public website's capture client can omit AMM vs cTrader from its illustrated strip; the underlying application feature remains separate from that screenshot configuration.

Daily Sharpe, Sortino and Calmar are based on the composed daily curve. They are not tick-level portfolio measurements. Profit factor/win rate use validated trade-ledger information under the model's scaling rules. Read the artifact-contract label and feasibility reason alongside the headline return.

Inspect feasibility and concentration#

The Candidates table includes rank, fitness, return, daily drawdown, time-weighted profit, profit factor, win rate, trades, generation, feasibility, candidate ID and active algorithms. Select a row to obtain its full detail; table pages themselves are compact summaries.

Inspect Constraint Violations and the objective/penalty breakdown. The highest raw return need not be the highest feasible fitness. Weight concentration and dependence between source strategies can affect the allocation question even when each strategy was attractive alone.

Candidates are loaded in pages. Scrolling can fetch more; opening one candidate does not fetch every candidate's full equity curve. In particular, missing Curve grid curves or max-jump-filter data in a compact result should not be read as flat performance or zero jump. Open the exact candidate in Equity and inspect its loaded detail. Preserve that distinction when reviewing older/large results.

Inspect instruments and executions#

Instrument chart shows one exact instrumentKey at a time. Equal symbol text on different providers/accounts, source resolutions or provenance remains separate. Choose the intended entry and verify its metadata.

The chart pages bars in windows. Trades and Orders have their own paging, and markers apply to the current bar window. A per-strategy trade filter helps inspect one source contribution. Indicators use that strategy's saved parameters. Chart type/timeframe change the displayed historical view; they do not rerun the portfolio search.

Use colour links for supported symbol/timeframe selection. Receiving a link selects a matching available instrument; it cannot expand the saved candidate's instrument set or start work. When a row seems absent, inspect the current filter, selected candidate and source-run identity.

Cross-check and analyse risk#

cTrader check opens producer, data mode, account and timeout settings. Choose exact native/Docker producer identity or supported existing backtests, and review the resulting source comparison. Within tolerance is a numeric comparison label, not a blanket strategy qualification or proof that two execution systems are identical.

The ninth Risk view hosts Factor model | Liquidity. Factor model can analyse the saved candidate, current holdings or a research allocation. Liquidity needs executable inventory quantities/weights and supported traded-volume history; strategy return sleeves alone do not establish that inventory. Follow the separate risk guides before interpreting either result.

Export the selected candidate#

Choose Export to live portfolio… from the run bar/Overview or the candidate menu. The dialog loads supported live routes for every active sleeve. Enter a name, choose an account per sleeve and resolve any no live route blocker. Review the DEMO/LIVE destination for each choice.

A LIVE route requires typed LIVE confirmation for export. Export saves a portfolio definition on the server; it does not start trading. Open Strategies afterwards, review sizing, actual broker funds, exact role symbols and the full account graph, then complete the account start review.

The route dialog's Export JSON saves the chosen route mapping. The candidate table's Export JSON saves the authoritative research candidate JSON. They serve different purposes; a research candidate file is not an approved runnable broker configuration.

Recover and retain evidence#

A failed estimate/run keeps its relevant error distinct from history-load errors. Use Load request into Builder to revise a failure without overwriting the original run. If candidate detail fails, check server/source identity and reload rather than treating absent data as zero.

Delete run… is disabled for active jobs and asks for confirmation; its server action archives history rather than erasing the result folder. Retain run IDs and exports for decisions that need later review. A remote API or browser cannot open the server's local folder directly; use the reported path/download instead.