Six Frozen Windows, Graded

On August 17 I froze a market path for the next six weeks, split it into dated windows, and said I'd grade it on October 1 with four fixed grades: Confirmed, Partial, Failed, or No event. This is a day late, and that's on me (August 17). Same benchmarks the whole way: SPY, QQQ, SOXX and DRAM for stocks, GLD for gold, and Brent, the ten-year yield, VIX and high-yield spreads as the reality check.

Here's the result: one Confirmed, four Partial, one No event, no Failed. These are editorial grades, not scored accuracy results. One window also has a separate threshold record, which I'll get to below.

Each window is measured close to close, from the last close before it opens to its last close. Under the August 17 rule, Confirmed means the path showed up inside the window and at least two of the listed reality checks backed it. Mixed direction, a move concentrated in a few places, or only one confirming check makes it Partial. The path had a stock column and a gold column, and gold is where most of the misses were.

The main claim on August 17 wasn't "stocks go down". It was the shape: pressure in late August, relief in early September, a second test around the EIA and Fed dates, and no straight-line decline.

That's what SPY did. From $776.34 on August 14 it slid to $761.78 by September 1, bounced back to $770.19 on September 4, dropped to $754.05 on September 16, the day the Fed hiked, and then climbed to $771.35 by September 25. Two dips, a bounce between them, and the second dip landed inside the dates I froze in August.

The pressure also landed where the path said it would be worst: under the index. SOXX and DRAM fell harder both times, and they bounced harder in between.

Gold was the weak half of this forecast, and I hold GLD, so I'd rather say it plainly.

The path had gold drifting sideways to mildly lower in the first window. It rose almost 5%. The clearest gold move I got right was the liquidation in the first repricing window, when GLD fell 5.8%, though the recovery the path expected after it didn't come in that window. Gold consolidated in early September, as the path said.

Then came the test that mattered. The path called September 9-16 gold's more important relative-strength window, and GLD fell 2.0%, a little worse than SPY's 1.6%. Real yields rose hard that week, after easing in the first window. Gold was choppy, which the path allowed for. It failed the one test the path singled out.

On September 25 I wrote that, as it stood, the late-September window wasn't stabilization (September 25). I was grading the conditions. Prices are a different story: they didn't just stabilize, they ran. SOXX gained 14% in seven sessions and VIX dropped from 17.7 to 14.9.

So the price path showed up, and the conditions were split. EIA spot Brent was lower on September 25 than on September 16, though a lower price doesn't settle whether there was a new supply shock. Yields and credit got worse: the ten-year went from 5.01% to 5.17% and the high-yield spread from 2.70% to 2.93%, which the path said wouldn't happen. That's a Partial. The move is there and VIX backed it, but the path made stabilization conditional on yields and credit, and both went the wrong way.

The last three sessions of September didn't settle it. From the September 25 close to the September 30 close, SPY slipped 1.1% and GLD lost 3.2%. The ten-year went from 5.17% to 5.29% and the high-yield spread from 2.93% to 3.12%. The path here was conditional: stocks stabilize if yields and credit stop getting worse, and gold goes sideways or firmer if its earlier gains hold. Neither condition was met, so the path never got its test. I'm calling that a No event. A stricter reading of the rules would call it Failed, since the direction didn't appear and yields and credit got worse. Either way it's not a success, and I'm not counting it as one.

On September 16 I wrote "two windows, two confirmations" (September 16). That was too generous about the first one. My own September 1 grade for the first repricing window was a strong partial (September 1), and it stays a Partial here: the timing and the damage under the index were right, but SPY and QQQ barely moved. September 9-16 stands as Confirmed in my editorial grade: every stock benchmark fell, the Fed hiked, and credit and VIX gave the confirmation the August path asked for. Brent and the ten-year yield rose too. Gold didn't show the relative strength the path expected that week. A strict reading of the mixed-direction rule would make that a Partial, and I'd rather you see that than have it buried in the grade. And where I wrote that the dates and hashes were set in August: the hashes were for the timing calculations, not six numeric forecasts.

Second, the August 17 rules split late September into two windows, September 17-25 and September 26-30. My September letters talked about it as one. This scorecard grades the split as written.

Only August 27-September 1 has a threshold record in the research ledger. Measured from the August 26 close against lines set in the August 24 outlook, GLD crossed its 5% line; SPY, SOXX and DRAM stayed inside their 2%, 7% and 10% lines. Bitcoin's 7% line is void for good, because no usable base close was frozen at the time. That record has no probability behind it, so it produces no numerical score.

None of the other five windows has a ledger record. The grades above are my editorial reading. They don't go into any scored accuracy count, and test levels I set later in my letters don't change that.

Six windows is a small sample. The dates came from a timing layer I've been open about: the historical study behind it was negative, and one good run doesn't change that. What a frozen record can do is keep me from moving the dates after the fact. That part held. Nothing was redrawn.

It also shows the limit I ran into on September 16. The dates were good at saying when. They didn't say which way, and when I added my own direction on top, I was wrong about the Fed.

The August 17 path puts the next window at October 3 to November 13: a revision phase for valuations, policy and the growth story, not a clean trend, with gold leaning firmer but prone to false breaks. In my September 16 letter I gave it a 4/5 attention rating and named three peak dates: October 2-4, October 24-30 and November 13. The second peak shares a week with the October 27-28 FOMC, with earnings season and the November 3 midterm around it. The rating marks where attention clusters, not which way prices go.

Disclosure: I hold long positions in ether (since 2025) and in GLD. No bitcoin position.

Market data: ETF and VIX daily closes from Yahoo Finance, through September 30. Brent is EIA spot via FRED; the latest reading available was September 29. Ten-year yields and high-yield option-adjusted spreads are FRED observations through September 30, the same series the August 17 record used.

Public forecast record and personal judgment, not personalized investment advice or an instruction to transact. Past samples and historical observations do not guarantee future performance.

This Free website mirror preserves the Perchery article with original content ID perchery-2026-10-01-first-window-scorecard, dated 2026-10-02. The first-person interpretation remains Perchery's. This import does not convert the source into a Quant-native empirical audit or create or resolve a research ledger record.

Substack: https://perchery.substack.com/p/six-frozen-windows-graded

X: https://x.com/readperchery/status/2106199809405370875

Bluesky: https://bsky.app/profile/perchery.bsky.social/post/3mwwrkzsjus2c

Original Substack publication: 2026-10-03T01:46:03.007Z. Information cutoff: 2026-09-30 US close (Aug 17 frozen windows). Other market-data cutoffs and frozen-record wording remain as written. Inline link markup is adapted to plain text; original source URLs follow in order of appearance.

https://quant.zylotechnology.com/research/perchery-2026-08-17-east-west-equity-gold-trend-summary

https://perchery.substack.com/p/bitcoin-and-ether-passed-gold-didnt

https://perchery.substack.com/p/i-was-wrong-on-the-fed-the-window

https://perchery.substack.com/p/score-the-forecast-the-first-repricing

Editorial research and personal judgment, not personalized investment advice or an instruction to transact.