ai-stock-agentPer-provider paper track
gemini
- Runs
- 1
- Latest equity
- $100,000.00
- Cash
- $69,583.33
- Positions
- 3
Latest run
20260529-034133 · 2026-05-29T03:41:33.465075+00:00
Action breakdown
Decisions
- NVDABUYconv 0.70
- MSFTBUYconv 0.70
- GOOGLBUYconv 0.75
Research readiness
Can we actually test whether this framework has edge? An honest read on the data — not a backtest.
Data on hand
- Runs
- 1
- Days
- 1
- Decisions
- 3
- Sell decisions
- 0
- Filled / orders
- 3/3
- Equity marks
- 0
- Provider
- gemini ×1
Known gaps
- Too little history — walk-forward needs months of trading days.(have 1 / need 60)
- No sell decisions yet — the sell framework cannot be validated without sell events.(have 0 / need 20)
- Too few closed outcome windows to measure edge.(have 3 / need 100)
- Too few equity marks for a meaningful equity curve.(have 0 / need 30)
Signal attribution (schema)
Which signal carried the edge? Decided per signal once outcomes close — unknown until then.
| Signal | Supported | Later outcome | Status |
|---|---|---|---|
| Bullish equity research analyst | unknown | unknown | Unknown — insufficient history |
| Bearish equity research analyst | unknown | unknown | Unknown — insufficient history |
| Technical / price-action analyst | unknown | unknown | Unknown — insufficient history |
| Fundamental analyst | unknown | unknown | Unknown — insufficient history |
| Risk manager (stop / take-profit / halt) | unknown | unknown | Unknown — insufficient history |
No backtest results are claimed. Attribution stays “unknown” until enough out-of-sample outcome windows close.
Read-only research audit. No backtest has been run and no profitability is claimed.
Research outcomes — paper replay
For each past paper decision, what did price do afterward — measured only from later trading days.
- Evaluated
- 3
- Pending
- 0
- Partial
- 0
- Closed
- 3
- No data
- 0
Price data as of: 2026-06-26
Raw close-to-close price movement after the decision, before fees and slippage. Not P&L, not direction-adjusted, not a backtest.
Read-only paper replay. No profitability is claimed; outcomes only mature into evidence once enough windows close.
Signal attribution
Which evidence lanes supported or opposed each decision — recorded now, scored only once outcomes close.
Coverage
- Decisions
- 3
- With attribution
- 3
- Ready to score
- 3
Lanes represented (decisions with a known stance)
- Agent decision3
- Technical3
- Risk3
- Event awarenessunknown
- Thesisunknown
- World Monitorunknown
Stances recorded
- Supports buy
- 5
- Supports hold
- 1
- Supports sell
- 0
- Caution
- 3
Each decision links to its Research Outcomes window; stances mature into evidence only as those windows close.
Attribution is not causation: a lane supporting a decision does not mean it caused the result.
Read-only attribution coverage. No edge or profitability is claimed.
Post-mortem loop
Reviewing decisions after their outcome windows close — to learn, not to score. Nothing is reviewed until a window closes.
- Closed
- 3
- Pending
- 0
- Review queue
- 3
Status: Ready for review
What will be reviewed once windows close
- Each closed decision against its t+1/t+3/t+7d outcome (later prices only).
- Whether each evidence lane's stance lined up with the realized move.
- Whether the red-team memo's failure modes actually mattered.
- A short human-readable lesson per decision (no automatic grade).
Future signal grades (not applied yet)
- Correct warning
- False alarm
- Missed risk
- Useful support
- Noise
Read-only. No win rate, profitability, or edge is computed; review begins only once outcome windows close.
Other tracks
Read-only per-provider track. Paper only — no real orders, no edge or profitability claimed. Per-provider validation panels (outcomes / attribution / sizing / red-team / post-mortem) are the next slice.