Eod Review

Update all open position prices to closing values.

3. Compare Accounts

trade-journal compare

tos vs ira vs tradier vs SPY benchmark. Compare account performance.

4. Reconcile

trade-journal reconcile

Verify the journal matches actual broker positions (TOS, Fidelity, Tradier). Flag any mismatches (phantom trades, missed fills, stale positions).

5. Risk Status

risk_governor.py status

Current risk utilization: exposure, drawdown, daily P&L, account halts.

6. Level Review

Read today's trading brief (memory/trading-YYYY-MM-DD.md). For each ORDER block and key level:

Fetch closing prices (use tradier for accuracy):

tradier quote SPY QQQ ES=F [+ any symbols from ORDER blocks]

Categorize each level:

• TRIGGERED — price came within 0.5% of the level
• NEAR — price came within 2% of the level
• WATCH — level not approached today

For each triggered level: did the setup work? Did the predicted direction play out? Was acceptance confirmed?

7. Signal Accuracy

Review signals delivered to Simon today:

• Premarket brief trade ideas: which were correct?
• Heartbeat alerts: were they timely and accurate?
• DP VTF signals (if any): how did those ideas perform across accounts?
• Mancini setups (if any): how did those ideas perform across accounts?

8. Score Today's Orders

For each ORDER block from today's brief, score the outcome:

track-record score <order_id> <WIN|LOSS|SCRATCH|OPEN> --pnl <dollars> --notes "..."

Order ID format: source-setup-conviction-ticker-date (e.g., mancini-FB-HIGH-SPY-20260415).

Scoring rules:

• WIN — price hit T1 or better during the session
• LOSS — price hit stop during the session
• SCRATCH — closed near breakeven (<0.5% move)
• OPEN — still active, not yet resolved (carry to tomorrow)

After scoring, run track-record stats to see rolling performance. Include the stats summary in the EOD report.

9. Journal Summary

journal summary

Count: signals received, risk checks run, orders placed, fills, governor blocks. This is the audit trail.

10. Produce EOD Report

One message:

=== EOD Review — YYYY-MM-DD ===

MARKET SUMMARY
[ES close, % change, session character (trend/chop/reversal)]
[Key macro drivers of the day]

ACCOUNT PERFORMANCE
tos ($100K TOS):       $X P&L (+Y%)  [positions: ...]
ira ($100K Fidelity):  $X P&L (+Y%)  [positions: ...]
tradier ($3K Tradier): $X P&L (+Y%)  [positions: ...]
SPY benchmark:         +Z%

LEVEL ACCURACY
[For each ORDER block: TRIGGERED/NEAR/WATCH, outcome if triggered]
Mancini: [N] levels in play, [M] triggered, [accuracy summary]
DP: [N] levels, [M] triggered

SIGNAL REVIEW
[Brief: how many ideas were correct? What was the best/worst call?]

TOMORROW'S SETUP
[Mancini pull running at 2:30 PM — key carryforward levels/positions]
[Open positions carrying overnight with current stops]
[Any strategy halt flags (account down >10%)]

10. Forced Reconciliation (MANDATORY)

The EOD report does not publish until Simon confirms positions. Present the shadow portfolio for each account and ask Simon to confirm or correct:

RECONCILIATION — please confirm positions are correct:

TOS: [list positions from shadow portfolio]
  Correct? (reply "yes" or paste corrections)

IRA: [list positions from shadow portfolio]
  Correct? (reply "yes" or paste corrections)

TRADIER: [list positions from shadow portfolio]
  Correct? (reply "yes" or paste corrections)

If Simon corrects a position, update the shadow portfolio before publishing the EOD report. If Simon doesn't respond within 30 minutes, publish with a warning: "Positions unconfirmed — shadow portfolio may be drifted."

This step exists because the shadow portfolio has no API connection to TOS or Fidelity. Without daily reconciliation, every position sizing calculation drifts from reality.

11. Feed Forward

• Open positions carry into tomorrow's RESEARCH phase
• Lessons → append to memory/trading-YYYY-MM-DD.md notes section
• Strategy concerns → if an account is consistently losing, flag for halt review:
  • Account down >10% from allocation → recommend trade-journal halt <account> "drawdown review"
  • 3+ consecutive losing days → flag pattern
• Mancini pull at 2:30 PM PT begins tomorrow's RESEARCH phase automatically
• Write EOD data to memory/trading-YYYY-MM-DD-eod.md for the portfolio review cron
• Cross-account exposure → log combined exposure per ticker for tomorrow's premarket check

Failure Modes

• trade-journal mark fails → run remaining steps (compare, reconcile, level review). Flag journal_incomplete: mark in report header.
• trade-journal compare fails → skip account comparison section. Flag journal_incomplete: compare.
• trade-journal reconcile fails → skip reconciliation. Flag journal_incomplete: reconcile.
• Quote fetch fails (tradier unreachable) → skip level review and closing prices. Alert Simon: "EOD prices unavailable — level accuracy deferred."
• Daily brief file missing → skip level review entirely. Report account performance only.
• Limitation (Phase 2): TRIGGERED/NEAR/WATCH categorization uses closing prices only (no intraday bars). A level hit and reverted mid-day will be miscategorized as WATCH. Note this in report header until Phase 3 ta-enrichment is available.

Boundaries

• Read-only for positions. This skill reviews and reports. It does NOT place orders, close positions, or modify the portfolio. Exception: it runs trade-journal mark which updates prices (not positions).
• No trade recommendations. The review is backward-looking. Forward-looking ideas are tomorrow's premarket-brief.
• No reconciliation fixes. If trade-journal reconcile finds mismatches, report them to Simon. Don't auto-fix.

References

• knowledge/trading/ — trading wiki (query for methodology, conviction scoring)
• references/TRADING_SOP.md — daily cycle, phases, signal routing
• references/STANDARD_ORDER_FORMAT.md — ORDER block format

Requires

• trade-journal workspace tool (mark, compare, reconcile, summary)
• quote workspace tool (closing prices)
• risk_governor.py (risk status)
• journal.py (journal summary)
• market-calendar (market day check)