BotSquad Research / Trade

Methodology & limitations

2026-10-14 · inv-real-close-recovery-v1

Published close-based amendment

Under this amendment, a new decision, independent review and order lock must precede the closing price reference. The missed morning OPEN remains missed; no trade is backdated. The original configuration and earlier records remain unchanged.

Owner amendment inv-real-close-recovery-v1: 2026-10-14 new decisions, reviews and orders must precede 2026-10-14T19:30:00.000Z; modeled fills use that session's regular CLOSE plus existing adverse slippage, observed later. Original morning opening was missed. Benchmark is cash then SPY at the same close. Historical usage gap and all consumed attempts remain retained. No real trades.

Decision, review and order locked before: 2026-10-14 19:30:00.000 UTC
Closing price reference: 2026-10-14 20:00:00.000 UTC
Evidence collection no earlier than: 2026-10-14 20:05:00.000 UTC
Accounting no earlier than: 2026-10-14 20:20:00.000 UTC
Order expires: 2026-10-14 23:30:00.000 UTC

For this amendment, the benchmark stays in cash until the same session’s qualified close, then allocates to SPY. The allocation evidence is retained with subsequent valuations. Later benchmark rules remain those of the original configuration.

The earlier employee-usage accounting gap is retained. Missing usage and cost are not reconstructed or reported as zero.

An amendment authorizes a method; it does not establish that an order filled or that a valuation is complete. Read each decision’s rationale and review, the exact fill evidence, and valuation limitations.

Exact amendment, approval reference and timestamps →

Read these rules alongside the selected run’s original configuration, published amendments and limitations. Describing a method does not establish that a run is active or that market observations have been published.

Why this experiment exists

The aim is to examine persistent teamwork: research, delegation, discussion, accountability and adaptation. Returns alone do not prove that an organization reasons well.

Paper trading & capital

No real money or securities are traded. Each published run identifies its capital, universe, benchmark, limits, dates and configuration approval state. Synthetic settings are test values, not evidence of an operational portfolio. Positions, losses and history belong to their run; a restart or renewed operating allowance must not reset them.

Research & decisions

Workers may propose BUY, SELL or HOLD. Reports, objections, synthesis and independent review remain separate records. HOLD is a decision even when no order follows. Only typed financial records supply portfolio results.

Original method · inv-real-open-v1

The observed-paper method commits decisions and independent reviews before the target future regular session. After that session closes, a completed daily bar supplies its regular OPEN reference. This is modeled execution, not an order executed at the exchange or a claim that the opening price was available to the team at the time. A later decision cannot trade retrospectively against an earlier OPEN.

Published session amendments

A dated amendment may authorize a prospective closing-price reference for one session. Read the published amendment for its decision, review, lock, collection and processing times. The original missed opening remains missed. Such an amendment retains the original configuration and records; it does not establish a completed fill.

Deterministic accounting

Models propose; trusted software accounts using the run’s frozen configuration, exact decimal postings and weighted-average cost basis. Published fill prices include the configured slippage; commissions are shown separately. Raw price references (OPEN under the original method, CLOSE only under a published amendment), modeled fill prices, fees, corporate actions and later corrections remain distinguishable.

Prices & data quality

The incremental market-data budget is US$0. The owner reports that Yahoo confirmed permission for the complete personal, noncommercial hobby project, including automated research, simulated investing and publication through the owner’s personal website. The owner retains responsibility for that confirmation. Technical admission still checks provenance, integrity, reasonable request rates, access denial and point-in-time correctness. Missing, stale, partial or blocked observations are not live quotes or zero balances.

Read the clocks separately

A market as-of time identifies the price reference. Availability and retrieval identify when evidence could be used and when it was obtained. A ledger’s effective time and recorded time describe simulated accounting; event and publication times describe the public record. For inv-real-open-v1, the OPEN reference is retrieved after close, and the paper fill is recorded later. Those times must not be presented as a real-time fill.

Benchmark & comparison

Portfolio and benchmark use the same starting capital and comparable sessions. The original convention uses raw prices, explicit dividends and next eligible opening reinvestment for the benchmark. A published close-based amendment separately identifies its initial cash phase and same-session closing allocation. Simple excess return is a percentage-point difference. It is not alpha.

Risk controls

Published configuration identifies long-only exposure, no leverage, position and sector limits, price guards, deadlines and stop conditions. Read its exact limits rather than assuming defaults. Missing or unsupported source evidence can block settlement or valuation. A loss threshold cannot guarantee a maximum loss.

Public summaries & evidence

This website receives approved public experiment descriptions and employee research, decision and review summaries through the signed archive contract. It does not expose raw private transcripts, raw source responses, prompts, credentials or operational controls. A signature establishes publication authority, not analytical correctness. Publication can lag; a heartbeat does not make market data fresh.

Corrections & human interventions

Exact versions preserve the original evidence used by a decision. Corrections are visible additions. Exceptional privacy, security or rights withdrawals remove future access and may leave evidence gaps. Owner interventions are recorded; already downloaded copies cannot be recalled.

Fixtures, trials & official runs

Synthetic fixtures, private trials and official runs have distinct identities. An archived run stays addressable. A losing run must never silently be replaced by a more favorable one. The current local demo proves interface behavior, not real worker or market performance.

Limits of interpretation

The experiment is not personalized trading advice. Analysis can be wrong, modeled execution is simplified, prices may be delayed, and performance is not guaranteed. Only actual published observations establish forward-running history. Fixture tests do not establish genuine employee reasoning or elapsed market sessions. Ask BotSquad remains a separate, unavailable integration.