Seven setups came into today’s close carrying an open verdict — six asset-lean calls from the premarket read and one mean-reversion call raised at midday. All seven resolved against confirmed regular-session closes: four fired, two voided, one split. Before a single outcome was written, every level those setups referenced was checked against the prior session’s settled close, and all six reconciled to the cent.
Three further setups — the weekly ones published Sunday — are deliberately not scored here. Each is conditional on an arming event that has not yet happened, and each runs to Friday. Finalising them tonight would be scoring a thesis that has not been given its window. Their interim state is recorded at the end of this section.
| Setup | Outcome | Evidence | Δ-ATR | Actual |
|---|---|---|---|---|
| AF S USO mm-260803-AF-USO-S | FIRE | The day’s cleanest read. Crude closed -5.46% and never looked back — the session high of 122.86 finished a full seven points under the 129.17 line it needed to reclaim. The brief flagged at emission that every announced halt in this conflict since February has since unravelled, and held conviction at 0.60 for exactly that reason. The move held anyway. | 1.28 | 122.12 |
| AF L FXY mm-260803-AF-FXY-L | FIRE | Closed +1.46%, and the session low of 58.43 never came within three-quarters of a point of the 57.66 invalidation. The rates leg that would have killed it never appeared — long-duration Treasuries finished flat at -0.07%. This is a direct reversal of Friday’s losing call on the same instrument, and the reversal paid. | 2.77 | 58.50 |
| AF L XLF mm-260803-AF-XLF-L | FIRE | Closed +0.77% with a session low of 57.00, just clear of the 56.94 line. The release-gated kill resolved in the setup’s favour: the manufacturing input-price index printed 71.1 against 73.0 prior, the cooling the thesis required. One leg is unresolved — the yield-curve condition could not be confirmed same-day and is logged as no-evidence rather than assumed. | 0.63 | 57.38 |
| AF L XLY mm-260803-AF-XLY-L | FIRE | Closed +1.83%; the low of 117.64 was never near the 116.09 stop-out. Notable because the brief explicitly discarded its own tape evidence here — the premarket print rested on 1,980 shares and was called unreliable in writing — and rested the call on the cheaper-fuel-to-consumer channel instead. That reasoning is what paid. | 1.04 | 118.21 |
| AF S XLK mm-260803-AF-XLK-S | VOID | All three written invalidation conditions triggered, which is a clean kill rather than a vague miss: technology reclaimed 175.35 and closed 178.04, semiconductors turned positive (+0.91%), and the Nasdaq proxy blew through 692.63 to close 700.07. The Chinese frontier-model release simply did not transmit — the two largest weights closed +4.93% and +2.93%. | 0.50 | 178.04 |
| AF S CPER mm-260803-AF-CPER-S | VOID | Voided on the thinnest possible margin — closed 39.64, eight cents above the 39.56 reclaim line, +0.20%. The whole copper complex went against it (miners +1.61% and +1.75%, materials +1.15%). The brief had already named this the weakest of its six calls on both legs, and it was. | 0.11 | 39.64 |
| OMR L XLE md-260803-1217-OMR-XLE-L | MIXED | Missed by three cents. Every structural leg held — no invalidation triggered, the 58.16 session low was never breached after the call went out, and the close sat above the day’s final volume-weighted average of 58.7708 — but the test as written was a close above 58.82, and 58.79 is not that. Against its stated risk unit the trade marks -0.23R, well short of its stop. | 0.03 | 58.79 |
The three weekly setups all remain live and none armed today. The energy short (sun-260802-EXF-XLE-S) needed a break of Friday’s 58.32 low on a reversal bar, plus confirmation that the strait reopening is genuinely proceeding rather than merely announced. Energy did trade through the trigger to 58.16 — and then closed back up at 58.79, which is the opposite of the failure pattern the setup requires. Its invalidation, a close above 60.45, is nowhere near.
The semiconductor long (sun-260802-OMR-SMH-L) cannot arm before Tuesday’s close by its own terms — the catalyst has to be delivered, not anticipated. The group closed 545.46, sixteen points under the 561.44 arming level and comfortably clear of the 503.63 invalidation. Untouched, as designed.
The index short (sun-260802-FBD-SPY-S) is the one that deteriorated. It needed the index to poke above 755.58 and then close back below it. Today delivered the first half — a 758.58 high — and then failed the second, closing 757.67. That counts as one of the two consecutive closes above the line that would kill the setup outright. Another green close tomorrow ends it.
Today’s seven scored at a 0.2240 Brier on a 57.1% hit rate. That is worse than the trailing window on hit rate and better than the lifetime average on Brier — the combination you get when the calls you lose are ones you had already priced cheaply. Net effect on the rolling series: Brier improved from 0.2233 to 0.2189 and the rolling hit rate ticked from 76% to 78%, the best readings the record has carried.
The by-confidence picture is where the honest signal sits, and it has not moved: the record is systematically under-confident in one band and over-confident in the one above it. The 0.55–0.60 band is now 64 calls deep and hitting 78.1% against a stated mean of 0.557 — a +22 point gap that has persisted for weeks. Directly above it, the 0.60+ band is 14 calls deep, hitting 50.0% against a stated 0.634: a -13 point gap and the worst Brier of any bucket at 0.300. The lowest band (under 0.40, 31 calls) hits 41.9%, also too pessimistic. Both of today’s highest-conviction calls sat at 0.60 and one of them — crude — was the day’s best result, which is the first useful data the top band has produced in a while.
By pattern, the asset-lean family continues to carry the record: 79% across 43 long calls and 67% across 46 shorts. The standing weak points are unchanged — level-rejection shorts at 41% over 34 calls and momentum-scalp longs at 46% over 33. Today added a first data point for the oversold mean-reversion pattern, and it was a miss.
Six of eleven logged macro predictions have resolved, and they hit 2 of 6 — 33.3%. The two that worked were both modest upside calls: the broad index for a +0.5% move (realized +2.29%) and bitcoin for +4.0% (realized +4.51%). The four misses were all downside calls on equity indices and gold, every one of which resolved against the direction predicted — index futures realized +2.06% and +1.20% against short theses, and gold managed +1.44% against a +2.0% target it fell short of.
The remaining five carry no resolvable price and are still pending. Worth flagging plainly: every resolved entry in this log dates to the first ten days of July, so this panel is a stale snapshot rather than a live read — it has not been fed since. Provenance label as returned by the reader, verbatim: est. (thinktank-v2 derived, asOf 2026-08-03). The database was read, never written.
This was the broad rally Friday only pretended to be. The S&P proxy closed +1.42%, the Nasdaq proxy +1.76%, the Dow proxy +1.32% — and critically, small caps closed +1.72% and the equal-weight version of the index +0.98%. Friday’s tell was an index carried by two names while equal-weight went red; today every one of those internals participated. Equal-weight still lagged the cap-weighted index by roughly forty-five basis points, so concentration has not vanished — but this is participation, not a narrow squeeze.
The engine was crude. Oil fell -5.46% after the planned strike on Iran was called off and talks were convened, and that single repricing propagated everywhere: into the consumer via cheaper fuel, into financials via relieved inflation pressure, and out of energy equities, which were the only sector left behind. The yen extended its intervention-driven rally +1.46% while the dollar index closed exactly flat. Duration was inert — long Treasuries -0.07%, high-yield credit -0.21% — and gold went nowhere at +0.05%. Copper edged up +0.20%.
Eight of eleven sectors closed green and only energy fell more than a quarter of a percent — the mirror image of Friday, when seven of eleven were red beneath a rising index. Communication services led at +2.86%, industrials and consumer discretionary both cleared +1.8%, and the defensives that led last week’s tape sat at the bottom without actually breaking: utilities flat, health care and staples down about two tenths.
A note on what is not here: a hard advancer/decliner count was not pulled this run, and none of tonight’s scored setups was gated on a breadth threshold, so none is being scored by inference. The participation read above rests on the equal-weight and small-cap closes, which are confirmed prices.
Mega-cap technology did the heavy lifting. The largest software name closed +4.93% and the large online retailer +4.58% — between them enough to explain most of the Nasdaq proxy’s gain on their own. The dominant accelerator name added +2.93% and the electric-vehicle name +3.49%. That such strength arrived days after a second Chinese frontier-model release is the session’s most quietly interesting fact.
The losers were almost entirely the crude complex. An exploration-and-production name closed -2.80%, a major integrated -1.85%, an oilfield-services name -0.57%, and the largest integrated -0.24% — a notably shallow decline given oil fell more than five percent, which says the equity market is pricing something less than a full de-rating. Outside energy the only meaningful drag was a large restaurant name at -2.00%, closing at its session low. Financials were broadly firm: the money-centre banks ran +0.24% to +1.67%.
The premarket read framed the day around a called-off military strike and the crude repricing that followed, and that framing was right about the mechanism and right about most of its consequences. The crude short, the yen long, the financials long and the consumer long all resolved as written. Where the read went wrong was in assuming the risk-positive impulse would stop at the sectors it had reasoned through.
The technology short is the instructive failure. Its reasoning was sound in isolation — a second Chinese frontier-model release in a fortnight, with the sector still negative on the month — but it was a bearish call placed into a tape that the brief’s own other five calls had just described as broadly risk-positive. The three invalidation conditions it named all triggered within the session. The copper short shares the shape: a bearish call on a day the brief had already argued the reflation channel was opening.
The midday call deserves separate credit for how it was written rather than how it resolved. It disclosed at emission that it was counter-trend, in the day’s worst sector, on a binary diplomatic catalyst, and pointed directly at its own contradiction with the still-live weekly short on the same instrument. It then missed by three cents. A call that documents exactly why it might fail, and fails for exactly that reason, is a process working correctly even when the trade does not.
One setup scored tonight carried an execution block: the midday energy long, entered at 58.82 against a 58.69 stop — a risk unit of thirteen cents. It closed at 58.79, marking -0.23R. The stop was never reached; this was a trade that went quiet rather than wrong.
That thirteen-cent risk unit is worth dwelling on. On a name whose average daily range is roughly 1.15, a stop that tight converts ordinary noise into a binary outcome — the three-cent shortfall that decided this row is around two percent of one day’s normal movement. Across the full record, execution now spans 56 closed setups at a mean of +0.16R, a 58.9% win rate and a 14.3% stop-out rate. Level-rejection trades remain the best-executing family at +0.36R over 26 trades, despite being among the worst-calibrated on direction — a reminder that entry quality and forecast quality are separate skills.