Product facts.
Defined and dated.

Counted on 21 September 2026 from the modular-desk build source, revision 52e4a8f4, including Minimum Track Record Length. Machine-readable census.

Catalogue

667 unique family/name entries across 28 families in the shipped-runtime census, including 175 roadmap entries. Of these, 492 are proven by an executable test that names them. The count is read from that census and does not re-execute the full catalogue.

Distinct charts

214+ distinct graphical results, including Minimum Track Record Length. This is a source-reviewed lower bound from build 52e4a8f4: 161 results accessible from the chart picker, 43 additional graph-driven results and 10 explicitly traced node-result figures. A result accessed through several views is counted once. Tables, text diagnostics, configuration readouts and candidates without a per-result rendering witness are excluded.

Examples: candle replay shares the candlestick result; surface, contour and 3D search landscape share one result; repeated prop-firm cash paths and distributions are deduplicated. Different calculations with their own plotted results remain distinct. Data requirements vary by chart. This census is not a fresh execution of the entire library or a competitor ranking. Dated inventory of the 214 counted results and collision decisions.

Code exports

6 registered targets: Python, Pine Script, MQL5, FASTEXPR, C++ and Rust. The translator covers 18 kernels. Target-specific coverage and execution semantics are documented in the generated file and export reference. FASTEXPR represents a continuous score. An exported file requires review and target-platform validation before live use. A MQL5 file does not imply a complete Expert Advisor matching every graph component.

Engine performance

66,000 bar-configuration evaluations per second, rounded down from a median of 66,238. Three first runs in fresh packaged-engine processes; EMA crossover, 52,565 hourly BTCUSDT bars, 12 configurations, one worker, Intel Core i9-12900K, 64 GB RAM, Windows 11. Timing includes orchestration and result retrieval; process startup and cached repeats are excluded. Measured on 21 September 2026. Workload and timings · Machine-readable evidence. No competitor speed comparison is claimed.

The exact input file and runner are not public yet. The recorded arithmetic can be checked, but the elapsed times cannot currently be reproduced independently. This measurement applies to the identified sidecar build; a later build needs a new measurement.

MCP walkthrough source

The landing animation replays selected fields from a real MCP development session recorded on 21 September 2026, source revision 52e4a8f4. Animation timing is editorial. The browser runs no backtest or MCP command. The source session used a protocol client without a language model.

1. compose_graph returns one runnable configuration.
2. backtest_graph starts run rg_0446867e.
3. backtest_status returns done, cached: false and verdict: NO-GO.

Displayed fields and trace fingerprint. No production-build licence setting was changed.

Brand use & attribution

Client names describe MCP-capable software requiring configuration. They do not represent a partnership or endorsement. End-to-end compatibility with every client version has not been certified.

• OpenAI: Codex is named as a client. No OpenAI or Codex logo is displayed.
• Anthropic: Claude Code is a factual client reference. The graphical logo is withheld pending approval.
• Gemini CLI is a Google product. Its graphical logo is not displayed.
• Google: the Google “G” logo appears unchanged on the “Continue with Google” sign-in button, as its sign-in branding guidelines require. Trademark of Google LLC.
• GitHub: the GitHub mark appears unchanged on the “Continue with GitHub” sign-in button, to identify the sign-in provider. Trademark of GitHub, Inc.
• Python and the Python logos are trademarks or registered trademarks of the Python Software Foundation, used by Stochastly under the Foundation's permission for nominative compatibility uses. Official logo unchanged.
• Rust logo, Rust Foundation. CC BY 4.0, unchanged. Trademark policy. Used to identify the Rust export target.
• C++ logo by Jeremy Kratz, licensed by the Standard C++ Foundation. Unchanged, used to identify C++ code exports. Usage terms and official assets.
• Pine Script is a TradingView trademark; MQL5 and MetaTrader are MetaQuotes trademarks. Stochastly is independent of these companies. Platform logos are withheld. Existing legal notice.
• The GitHub, LinkedIn, X and YouTube marks belong to their owners. The site footer uses them as links to the Stochastly accounts and repository. Stochastly is developed independently of these companies.

Pascal, modulo two

The pixel motif is constructed from the odd coefficients of Pascal's triangle: C(n,k) is odd when (k & n) = k. It produces the Sierpinski triangle pattern. Original vector construction; no third-party painting or photograph is reproduced.

Back to Stochastly