Research note

Strategy Lab: unknown runs, indeterminate SSO

No page-wide verdict. SSO remains INDETERMINATE and definition-only; the ONDS reference PASS belongs to a distinct implementation.

5 min read

Verdict: INDETERMINATE for the SSO recent-window replication; PASS for the ONDS reference script only. There is no single page-wide verdict. The distinct in-app ONDS variant and its ex-ONDS generalization do not inherit that PASS; other parameterized Lab runs have no frozen result recorded in the inventory.

Open this backtest in Backtesting

The rule

The Lab is a family of experiments: catalog strategies, condition DSL, code, variants, entry/exit grids and descriptive analog paths. A rule is identified by ticker, strategy or AST, parameters, interval, lookback, fill settings and costs; there is no universal Lab entry or exit. The catalog's opening settings are 1,800 days, 5 bps round-trip transaction cost plus 2 bps slippage, no stop or target, long-only, unit position size and benchmark enabled. These are configuration defaults, not performance evidence. Each strategy's warmup, filters, exits and short support remain its own schema.

SSO is a separate saved definition: long-only regular-session 15-minute bars; SMA(50) above SMA(200) and ADX(10) above 20. Setup A breaks the prior 14-bar high with volume at least twice its 20-bar average. Setup B crosses session VWAP with RSI(14) above 50 and volume at least 150% of that average. One position fills next regular-session open. Exits are a 3.5% stop, 15% target, 7% highest-close trail, a loss after 25 candles, or less than 1% profit after 60 candles. The replication charges 10 bps round-trip. This definition is display-only and user-added under override; it is voteless and never runnable in the Lab.

Methodology

Evidence as of: the INVENTORY review dated November 5, 2026; this is the consolidation planning baseline, not a new observation window. Publication does not extend any study's sample.

Individual Lab honesty, walk-forward and trade-umpire checks belong to the captured request. They cannot certify the whole catalog. Paths are descriptive analogs, not a trade simulation. No study was run for this article.

SSO used Yahoo 15-minute history: 1,066 bars across 41 sessions, eight completed trades split into five in-sample and three validation. The full data span is September 8–November 3, 2026. In-sample is September 8–October 9, 2026 (24 sessions). The validation window is October 12–November 3, 2026 (17 validation sessions). Its sample is too small for a reliable conclusion despite the negative mean and t statistic. The source's May 2022–November 2026 claim of 153% versus 111% buy-and-hold was NOT reproduced. Those claimed returns are not measured results here.

ONDS has separate reference-script, in-app and ex-ONDS generalization records. The inventory does not record comparable costs, exact windows or drawdowns for these summaries; check the packaged ONDS spec/report and DSL description before making a further comparison. Membership, selection and repeated testing limit generalization; the reference PASS cannot validate a changed implementation.

Results

Study / variantInventory evidenceWindow / basisStrategy return / CAGRBenchmarkMax drawdownClosed trades / observation NSignificance / ICNative verdict
SSO validationD01 / H54October 12–November 3, 2026 (17 validation sessions); Yahoo 15mmean net −1.28%; CAGR not recordednot recordednot recorded3 validation tradest −2.54INDETERMINATE
ONDS reference scriptD01 / H55reference implementation; exact window not recorded+894.1%; mean +5.48%+688.4% B&Hnot recorded56 tradest 2.24PASS — reference only
ONDS in-app variantD01 / H55distinct implementation; exact window not recordedmean +4.80%; total not recordedtrails B&H; value not recordednot recorded57 tradest 1.99no transferred PASS
Ex-ONDS generalizationD01 / H55other symbols; exact window not recordednot recordednot recordednot recorded378 tradest 1.42failure / decay caveats
Other Lab requestsD01 / F01–F10 / H01–H13not recordednot recordednot recordednot recordednot recordednot recordedUnknown — no frozen result

The linked workspace retains the complete native output and notes when available. Missing evidence is not a zero return, a successful empty result or a failed strategy.

Verdict and limitations

SSO stays INDETERMINATE, not FAIL or a near-pass. Its protected definition labels remain: “SSO Opus 5.5 Swing 15m (TrendSpider, user-added under override)”; “Recent-window replication INDETERMINATE — validation 3 trades, mean net -1.28%, Sep 8–Nov 3 2026 (Yahoo 15m)”; “TrendSpider full May 2022–Nov 2026 claim (153% vs 111% buy-and-hold) NOT reproduced — Yahoo 15m is capped at ~60 days”; “Intraday rule — no daily-consensus vote”; and “Not runnable in Strategy Lab: the Lab offers 1D, 1H, 30M and 4H intervals (no 15M), and its generic engine cannot express this rule's two time exits plus its 7% trailing stop. The exact rule is implemented and tested in backend/app/scoring/technical/sso_opus.py and is deliberately NOT registered in the Lab strategy catalog.” The frozen headline's Sep 8–Nov 3 wording names the full data span; its three-trade validation result belongs only to the October 12–November 3 split above. The interval explanation is frozen source wording; current interval support does not grant execution.

The ONDS reference PASS is confined to that script/sample. The in-app variant and failed/decaying generalization remain separate. Other runs are Unknown — no result/verdict recorded in this inventory. None of these editorial links creates a signal, adds a vote or executes a definition.

Sources and live backtest

INVENTORY D01, F01–F11, H01–H13, H54/H55 and §7.4 are the claim register. Implementations: frontend/components/StrategyLabPanel.tsx, StrategyLabHonestyPanel.tsx, frontend/lib/sso-opus-definition.ts, frontend/lib/strategy-lab-dsl-view.mjs, and the deliberately unregistered backend/app/scoring/technical/sso_opus.py.

Read the original SSO report, SSO stats, SSO definition, and ONDS reference report. Original SSO raw bars remain at /home/hatch/workspace/sso-opus-swing/klines.json, digest-only in the packaging manifest; no public raw-bar download is offered. Unknown Lab requests require their exact captured request/native result, not a rerun manufactured for this post.

Open this backtest in Backtesting

Not financial advice

Built for education. Backtests are historical; they do not predict future returns and are not a recommendation to buy or sell any security.

connecting…