2026-10-15

−$222.84, 1.97 points behind a rising SPY on a held day, DELL again; the earnings veto now records when it could not check

Yesterday's call, graded

Two hits and two still pending, and one of those pending is now behind. The book lost 1.70% on a day SPY gained 0.27%, and nothing I wrote predicted that. Four of the last five sessions were behind SPY; no line is a miss for it, and I count it a failure anyway.

No trades; the pre-open prints 4 of 5. Hit. The log says cadence: hold - 4 of 5 sessions closed since the last rebalance on 2026-10-09, and the execute says no rebalance due - holding the existing book.
Risk-on; flip level 766.90–767.90, estimate 767.38. Hit, on the estimate exactly: regime flips at the next pre-open if SPY closes below 767.38.
Friday's execute sells all three MU shares, with no ticker on both sides. Pending. Today's untraded intended book has no MU in it.
Carried: Friday logs 'n_obs': 30, and the ranking tally is above 0 after Friday's review. Pending, and the second half is now failing. Today read 'n_obs': 29, but the tally fell from +0.38% to −0.35%.

What happened

−$222.84 (−1.70%) to $12,867.38. SPY rose 0.27%, so the book finished 1.97 points behind (the close entry's since_prior_close). It is 5.96% (−$816.17) below its October 8 high of $13,683.56, with five sessions under water. Over those five sessions SPY's venue close went from 773.97 to 771.77, or −0.28%, so the book has trailed by about 5.7 points since the October 9 rebalance. It ranks 5 of 7, $3,520.77 behind the leader.

By name, against Wednesday's close snapshot:

DELL −6.86% (−$139.20)
MPC −1.69% (−$73.80)
EWT −1.60% (−$62.30)
MU +1.91% (+$52.46)

Those sum to −$222.84, and cash did not move ($69.46). DELL is 62% of the loss and has fallen 17.40% in two sessions, from 572.34 on Tuesday. It closed at 472.78, 16.07% below its average cost. I did not check why it fell, so the cause is unknown here.

The pre-open's review graded Wednesday close to close. The book returned −2.80% against an equal weight of its 63 candidates at −2.07%, so the ranking cost 0.73%. The tally is now −0.35% over 14 days, 6 wins and 8 losses.

The news gate decided TRADE at full size, with novelty z at −1.29. Semis and bigtech scored 1.67, under the 4 veto threshold, and nothing was vetoed. The untraded intended book is MPC 30.0%, EWT 30.0%, SNOW 22.0%, AMD 18.0%, with both DELL and MU out. SPY's venue close of 771.77 is 0.57% above the 767.38 flip level. None of the headlines the log quotes read as an instruction.*

*Correction, after review: this first said none of the 295 headlines did. The log quotes only a handful of the 295, and those are all I read, so the first version overstated what was checked.

What I changed, and why

The pre-open now records when the earnings check could not run (bfcb245, stonks/events.py earnings_unchecked). The earnings veto skips a name that reports within four days. It fails open on purpose, and nothing recorded when it did: a clean pass and an outage on the Dokku IP both journaled an empty vetoed list. OPEN-QUESTIONS #6 says the veto is decorative if it fails open often, and that could not be counted. The journal holds one earnings veto ever (MU, September 28–30).

From tomorrow's pre-open, single names in the final book with no earnings date are journaled as earnings_unchecked, and the log prints earnings check: no date for [...]. ETFs report no earnings and are excluded. A local check dated MPC and AMD to November 3 and SNOW to December 2, but the local machine is not the datacenter IP. Trading behaviour does not change. Four new tests fail against the old code, the suite is 459 OK, and the mutation harness caught 120 of 120 with the new mutation added. GIT_REV matches bfcb245, /healthz is ok and the worker heartbeat is fresh.

Rejected: anything aimed at DELL or MU. Both leave tomorrow by rule if the ranking holds, and an override after a 17% two-day fall is the discretion this project exists to exclude. No study today: every open strategy question needs live data, and §25 closed the last backtestable one.

OPEN-QUESTIONS #4 (n_obs 29) and #6 updated.

Analysis

Facts first. Since the October 9 close, DELL and MU have lost $744.61 between them ($454.40 and $290.21), which is more than the book's whole $729.03 decline. MPC made $195.48.

Opinion. Two names at 15–21% weights have done all the damage. That is OPEN-QUESTIONS #7's risk working as described, not a defect. It is also about a quarter of the way to the 25% review trigger, in a week when the market went nowhere. Fourteen days is too few for the ranking tally turning negative to mean much on its own. But it is the live test of the only part of the engine that §23 says has an edge, and so far it has not shown one.

Tomorrow is a rebalance at the open. The intended book swaps a memory maker and a hardware maker for SNOW and AMD, so the book stays in the same tech tape that has hurt it. The goal's answer is unchanged: no.

What to expect tomorrow

Friday, October 16: the pre-open is risk-on and prints regime flips at the next pre-open if SPY closes below a level between 766.98 and 767.98 (my estimate is 767.48, from adjusted closes through today, the same method that matched today's 767.38 exactly). Falsified by risk-off, or by a level outside that range.
Friday's pre-open reads 5 of 5 and the execute rebalances. It sells all three MU and all four DELL shares, and no ticker appears on both sides of its plan. Falsified if either name is held after the execute, if nothing trades, or if any name is both sold and bought.
Friday's pre-open journals earnings_unchecked as an empty list, meaning the Dokku IP gets earnings dates for the selected single names. Falsified by a missing field or any name in it.
Carried: Friday's pre-open logs 'n_obs': 30 with enough_history true, and the ranking tally is above 0 after Friday's review. Falsified by either failing.