FO Research
The Record
Every Financial Oracle call, dated before the data and scored against the tape. The scored record: long dollar, no cuts, the floor holds.
Financial Oracle · The Record
Every call, dated before the data, scored against the tape.
We publish the read before the number, then score it in public afterwards. This page is the receipt. Levels are real (US Treasury / FRED for yields, extended with same-day end-of-day prints where FRED lags; market tape for FX, commodities and equities), entry is the publish date, scored to 28 September 2026. We score everything, including the calls that have not worked.
82% held or better. 71 scored calls: 40 held, 16 partial, 1 rotated, 1 ran then closed, 2 early, 1 ambiguous, 10 missed. A further 4 registered in the latest note and not yet scored. Every call is linked to the research that made it before the event.
Reference point: levels as of 28 September 2026. Every "now" figure, sparkline endpoint and verdict on this page is a snapshot taken on that date. Entry is each call's publish date. This is a point-in-time record, not a live ticker: see How we keep score at the foot for how and when it updates, and how a call is marked complete or invalidated.
The spine · our highest-conviction macro calls
Long dollar. No cuts. The floor holds.
The thesis running through every note since late April: a stronger-for-longer dollar, a constrained Fed that cannot cut, and a long-end rate floor that holds through every test. It paid through the summer, and every call below is still open on its published terms. The headline figure is the move the thesis made available: the maximum favorable travel while the call was in play, which is what the analysis put in front of a reader who used it. Underneath it we also show where the same call sits today, held from the published price with no timing assumed, because a thesis in progress retraces and we do not hide that. Entries and exits are the reader's own. We call the paths; the reader trades them.
| The call | Path (entry → 28 Sep 2026) | The move made available |
|---|---|---|
Short EUR/USD The founding dollar-regime call: sell EUR rallies while US rate differentials and a constrained Fed keep the dollar bid. It gave most of the June move back through August, then recovered nearly all of it in September as the Fed moved first. | Path 1.17291.1371 | Peak · now+370 pips the move made available now +358 pips ○ open |
Short GBP/USD GBP upside capped while the dollar stays bid and the Bank of England lags the Fed. The Bank held 6 to 3 on 17 September while the Fed rose 12 to 0: the lag the note described, on the record. | Path 1.36051.3254 | Peak · now+441 pips the move made available now +351 pips ○ open |
Long the dollar (DXY) The dollar broadly higher for longer on rate differentials and US relative strength. Back within half a point of its June peak after the September rise. | Path 98.06101.20 | Peak · now+3.6% the move made available now +3.2% ○ open |
Short AUD/USD Fade AUD rallies into broad dollar strength while China momentum stays soft. | Path 0.72010.7016 | Peak · now+314 pips the move made available now +185 pips ○ open |
Long USD/CAD USD/CAD higher on the oil and yield-spread divergence and a price-defence regime. Within 60 pips of its June peak at the 28 September mark. | Path 1.35821.4174 | Peak · now+652 pips the move made available now +592 pips ○ open |
Short NZD/USD Horizon: 1 to 2 weeks, tactical NZD bearish while USD momentum and risk-off flows persist. A tactical call that has aged into a regime one. | Path 0.58860.5663 | Peak · now+247 pips the move made available now +223 pips ○ open |
Short EUR/USD Horizon: tactical, event-driven EUR sold even on the ECB hike: tightening into a downgraded growth outlook is not a strong-currency story. The ECB rose again on 10 September and the euro fell anyway. | Path 1.15791.1371 | Peak · now+220 pips the move made available now +208 pips ○ open |
Short GBP/USD Horizon: tactical, event-driven GBP soft into Warsh's hawkish hold and a firmer dollar. Offside through August; back in profit when the hold became a rise. | Path 1.3451.3254 | Peak · now+286 pips the move made available now +196 pips ○ open |
No cuts (US 2-year) The Fed cannot ease into a 4-handle inflation backdrop: the front end stays anchored and cut bets get priced out. They were, and then some: the committee raised the range to 3.75 to 4.00 on 16 September, 12 to 0, and the two-year sits at 4.92 with a second rise near seventy per cent priced for October. | Path 3.90%4.92% | Peak · now+102bp the move made available now +102bp ○ open |
The 30-year floor holds Term premium and the deficit hold the long-end floor through every test: a hot jobs print, a soft CPI, a new chair, a ceasefire and its expiry, an enlarged issuer bid and a unanimous rise. It ran to 5.31 on 17 August, eased when the Treasury doubled its buybacks, then went to 5.56 on 28 September, the highest since 2004, with breakevens falling: the entire move and more was real yield. | Path ~5.00%5.56% | Peak · now+56bp the move made available now +56bp ○ open |
On pullbacks, and why an open call is not a loss. No market travels in a straight line. A thesis that has run several hundred pips will retrace part of it, and a retracement inside an open thesis is a position waiting, not a position injured. Every call above remains open on its own published terms: none has reached the invalidation its note stated in advance. The dollar calls invalidate on a credible Fed pivot, a decisive collapse in core inflation or coordinated intervention, and the Fed raised instead. The long-end floor invalidates on a sustained break of 4.85% with term premium falling, margins normalised, breakevens anchored and credit quiet; it went the other way, to 5.56%, the highest since 2004. The front-end call invalidates on a Fed dovish capitulation, and the Fed delivered a unanimous rise. The equity-caution calls invalidate on high-yield spreads through 3.50%, and they sit at 3.02%, through the desk's own 2.90 condition and short of its 3.10 tell. So we report the move each thesis made available, and we report where it stands today, and we leave both on the page. What a reader took out of the move, on the pullbacks or at the extremes, is their own execution.
The dollar calls are back in profit and the rates spine went further. The founding 30 April dollar calls each made 300 to 650 pips available at their June peaks (short EUR/USD +370, short GBP/USD +441, long USD/CAD +652), gave part of it back through August, and recovered most of it in September as the Fed moved first: every one of them sits in profit at the 28 September mark, held from the published price with nothing assumed. The rates book has aged best of all. The constrained-Fed call repriced the 2-year from 3.90 to 4.92, +102 basis points, and a rise was delivered 12 to 0 on 16 September. The 30-year floor did not just hold, it went to 5.56, the highest since 2004, through a hot jobs print, a soft core, a new chair, a ceasefire and its expiry, an enlarged issuer bid and the desk's own 5.45 ceiling, which is scored as missed below.
The full record · every stated cross-asset bias
The complete scorecard, scored honestly
The spine above is the highest-conviction book. Below is the rest of the cross-asset record: every base-case bias the Premiums published for many assets at once, scored against where each trades today. Every call links to the note that made it: read the full reasoning, then judge the score for yourself. Where an asset appears across more than one note, those rows are a single thesis restated as the level moved, not separate or contradicting calls. Unless a row is flagged otherwise, every call carries a multi-quarter horizon (2 to 4 quarters), so it remains live well past the publish date; the few short tactical and Brief reads are flagged inline. Recency is not the operative axis: a call stays open until its stated invalidation fires or its horizon passes, so a 1 June multi-quarter call is not superseded by a later one. The rates and credit calls are where the desk is strongest. We mark the rest plainly: the Brent range held, broke lower on the US-Iran ceasefire, then returned as the ceasefire expired; the 18 August 5.05 to 5.45 thirty-year band broke at 5.47 and is scored missed as that note pre-committed; four credit-range calls broke their ceilings by 2 to 12 basis points at the 28 September print and are scored partial until the desk's published 3.10 tell decides them one way or the other; the 1 June gold range missed outright as real yields rose; the 19 June "gold lower on the dollar" call was scored missed at the August mark when gold ran to records and sits onside after the dollar's September run, with the excursion against it stated; the semiconductor-leadership call missed while its counterpart, semis-underperform, has come right; and the broad AI-caution call on the Nasdaq is a forward-guidance thesis that dropped early, then reversed, and remains open and offside pending its stated invalidation.
40Held
16Partial
1Rotated
1Ran, then closed
0Mixed
2Early (open)
1Ambiguous
4New
10Missed
82%Held or better
See what Premium readers received before the move →
| The call | Entry → 28 Sep 2026 | Verdict |
|---|---|---|
| Rates | ||
US 30-year The 30-year stays pinned near 5.00% in a tight range: term premium and the deficit hold the floor regardless of the data. The floor held and the range did not: 5.31 in August, eased on the 19 August buyback expansion, then 5.56 on 28 September, the highest since 2004. | Move~5.00 → 5.56 | Verdict✓ Held ○ Open |
US 30-year The 30-year holds 5.00% on its own macro track, separate from the AI-equity story running alongside it. It has, by 56 basis points, while the Nasdaq went nowhere. | Move5.01 → 5.56 | Verdict✓ Held ○ Open |
US 30-year The 30-year holds its floor through the CPI print: the floor is structural, not data-driven. It ran through the 5.15 top of the stated band to 5.31 in August and 5.56 by late September, through three more CPI prints and a rise. | Move5.03 → 5.56 | Verdict✓ Held ○ Open |
US 30-year The 30-year holds even as the war premium drains: the floor is the vanishing bid, not the conflict. The conflict came back, the bid did not, and the issuer's own enlarged operations have not held the line either: 5.56 with the Treasury taking its ceiling at every one. | Move4.93 → 5.56 | Verdict✓ Held ○ Open |
US 2-year The 2-year stays anchored with cut pricing capped: the Fed cannot ease into a 4%-handle inflation backdrop. It could not. The committee raised on 16 September, 12 to 0, and the two-year sits at 4.92. | Move4.13 → 4.92 | Verdict✓ Held ○ Open |
US 2-year The 2-year holds with no cuts priced into Warsh's first meeting: the constrained stance carries to the new chair. It carried to his third meeting, which delivered a rise. | Move4.05 → 4.92 | Verdict✓ Held ○ Open |
US 2-year The 2-year drifts in the high-3s as the market waits on a Fed on hold. It repriced to a 4-handle and then to 4.92 as a rise was delivered instead: the no-cuts umbrella was right, this level call was not, by nearly a point. | Move~4.0 → 4.92 | Verdict✗ Missed ○ Open |
2s30s The 2s30s curve stays steep, a market pricing structural supply at the long end while the front waits on the Fed. It steepened to a 108bp extreme in August, then bear-flattened to 53bp on 21 September as the front end stopped waiting and led the committee into a rise, and has re-steepened to 64 on term premium since. Still positive, 24bp flatter than at entry, and the mechanism the call described inverted: the front end led. | Move+88 → +64bp | Verdict~ Partial ○ Open |
2s30s The 2s30s curve holds steep, the shape of a bid problem at the long end. The bid problem is real and the long end is at 5.56; the front end rose faster still. 24bp flatter than at entry. | Move+88 → +64bp | Verdict~ Partial ○ Open |
US 30-year As the foreign official bid fades, the marginal buyer of American duration turns domestic. The direction held and the buyer arrived closer to home than the note wrote: on 19 August the issuer itself doubled its long-end buybacks, and by late September was taking its ceiling at every enlarged operation while the long end went to 5.56 regardless. | Move4.94 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Supply-side inflation is global, giving every hawkish committee the same cover and holding the long-end floor. Japanese ten-year yields reached 3.055% on 24 September, the highest since 1996; gilts and bunds at multi-year highs. | Move4.86 → 5.56 | Verdict✓ Held ○ Open |
US 30-year The energy shock reaches the consumer through refining, and the long end does not soften into it. August CPI printed +0.4 on gasoline and the long end went 21bp higher in the fortnight after. | Move4.98 → 5.56 | Verdict✓ Held ○ Open |
US 30-year The floor is a cross-central-bank regime, not a US story: Fed and ECB turn hawkish together and duration stays offered. In September the Fed rose 12 to 0, the ECB rose, the Bank of Japan went to 1.25 on 7 to 2 and the Bank of England held 6 to 3 with three votes for a rise. The synchronised floor became a synchronised repricing. | Move5.06 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Term premium rebuilds rather than fades, hardening the floor from below. The ten-year breakeven fell from 2.38 to 2.34 across the rise while the real ten-year went from 2.55 to 2.90: the whole September move was term premium and real yield, not inflation compensation. | Move5.06 → 5.56 | Verdict✓ Held ○ Open |
US 30-year A soft June CPI bends the floor without breaking it: a bull steepener, not a broad rally. | Move5.08 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Refining margins, not crude, carry the shock into headline inflation and keep the long end pinned high. The August print carried it as written, +0.4 with gasoline +3.9%. | Move5.13 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Crude round-trips in four sessions and the long end keeps its level: the fades do not undo the ratchet. Brent has since run to 106 and back to 98; the long end kept none of the fades. | Move5.12 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Each supply shock leaves the stance tighter than it found it while the fades leave it alone, and the next move is a rise, not a cut. Delivered on 16 September, 12 to 0, the first rise since 2023. Resolved. | Move5.09 → 5.56 | Verdict✓ Held ● Resolved |
US 30-year Horizon: tactical, event-driven A hold with three dissents for a rise keeps the long end elevated. The dissents became the decision seven weeks later. | Move5.21 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Horizon: tactical, event-driven Seven hikes voted across three committees, none for cuts, and the long end holds its new range. It ran to 5.31, eased below the publish level on the buyback doubling, then went to 5.56 as the votes became decisions: the Fed 12 to 0 for a rise, the Bank of Japan 7 to 2. | Move5.23 → 5.56 | Verdict✓ Held ○ Open |
Sept hike pricing Horizon: tactical, event-driven The pre-registered test, quoted as published: hold above half through benign data and the ratchet is priced. It did not hold. Pricing settled below half at three consecutive post-CPI sessions and the desk wrote the reversal note it had pre-committed to. It then re-armed to near 90% on the August CPI and the rise was delivered on 16 September, 12 to 0. The direction the note argued arrived six weeks later; the test it registered failed on its own terms, and the miss stands as scored. | Move66% → ~50% → delivered | Verdict✗ Missed ● Resolved |
US 30-year The tightening migrated from the committee to the curve, and the published 4.85 to 5.20 band was retired for 5.05 to 5.45, with a pre-commitment that a close through the top would be scored as a range miss on the day it printed. Twenty-six consecutive closes settled inside it, 5.17 to 5.40. The twenty-seventh closed 5.47 on 24 September and 28 September closed 5.56, eleven basis points above the line. The mechanism is not what failed; the level did, and it is scored as the note said it would be. | Move5.28 → 5.56 | Verdict✗ Missed ○ Open |
US 30-year Horizon: tactical, event-driven The buyback doubling read as signal, not flow: the issuer smooths the ink it will not stop writing, and rallies under its bid are distortion, not discovery. No rally came. The issuer has taken its ceiling at each enlarged operation since 9 September and the long end went to 5.56 regardless. The note's separate queue claim is scored in the pre-registered tests below. | Move5.24 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Horizon: tactical, event-driven The issuer test, standing from 18 and 20 August: long-end rallies on the issuer's own bid are distortion, not discovery, and the witnesses that testify unaided are the two-year and gold. Both witnesses spoke through September: the two-year led the committee into a rise and gold fell through 4,449 as it did. | Move5.227 → 5.56 | Verdict✓ Held ○ Open |
US 30-year Base case: the upper half of 5.05 to 5.45, the pivot beneath, auctions absorbed after concessions. The upper half held for eight sessions, then the band broke at 5.47 on 24 September and 5.56 on the 28th, the range miss the note pre-committed to. The concessions came and the auctions were absorbed badly: the 20-year stopped at 5.42, the 5-year at 5.033 on a 2.21 cover, the weakest since 2025. | Move5.35 → 5.56 | Verdict✗ Missed ○ Open |
US 2-year Holds 4.50 to 4.75 as the rise lands; the committee's tenor fully priced. It went through the top to 4.92 as a second rise for October priced near seventy per cent: the direction and the mechanism held, the band did not. | Move4.63 → 4.92 | Verdict~ Partial ○ Open |
2s30s Flattens further, the migration's bear flattener. It did, 72 to 53bp on 21 September, then re-steepened to 64 on term premium as the long end went further without the committee. Flatter than at entry; the flattener the note described has been a steepener since the 23rd. | Move+72 → +64bp | Verdict~ Partial ○ Open |
10-year breakeven 2.30 to 2.45, the three-month high held. Inside the band throughout, 2.31 to 2.40, and falling into the rise: 2.38 on the eve of the decision, 2.34 on 28 September. The credibility the committee bought is priced here, and it is why the desk reads the long-end move as term premium rather than a pass-through. | Move2.36 → 2.34 | Verdict✓ Held ○ Open |
| Pre-registered tests · Jackson Hole | ||
The keynote Horizon: tactical, event-driven The desk expected continuity and registered the risk branch: conditions named insufficient, progress named stalled, the front end leading, September toward the coin flip. The risk branch fired nearly verbatim and the desk scored itself missed at full weight, the miss first, in The Chair Speaks. The Pen Stays. | Move~36% → ~50% | Verdict✗ Missed ● Resolved |
The benchmark revision Horizon: tactical, event-driven Registered: a revision under roughly 300,000 disarms the collision's second hand and leaves the keynote alone on the stage. It printed minus 79,000 against a consensus near plus 183,000, and the session traded entirely on the speech. | Move-79k | Verdict✓ Held ● Resolved |
July core PCE Horizon: tactical, event-driven Registered: an in-line print changes nothing, because the desk's dated inflation test is the 11 September CPI. It printed 0.2 on the month and 3.3 on the year, in line. Nothing changed. | Move0.2 / 3.3 | Verdict✓ Held ● Resolved |
Bank of Korea Horizon: tactical, event-driven Registered: a second rise to 3.00% extends the synchronised floor by another jurisdiction. It printed 25 basis points to 3.00, back to back, one dissent, framed by the governor as pre-emptive. | Move2.75 → 3.00 | Verdict✓ Held ● Resolved |
| Pre-registered tests · The August CPI | ||
The print Horizon: tactical, event-driven Marked 60 / 20 / 20 on the labour leg and held the formal reweight for the print: a hot August CPI seals the rise, a cool one with the crack still wide earns nothing. It printed hot, +0.4 with gasoline +3.9%, and sealed it: near 70% on the eve, nearly 90% on the release tick, with the dollar flat and the long end down. The asymmetry printed as written. | Move70% → ~90% | Verdict✓ Held ● Resolved |
The migration Horizon: tactical, event-driven A two-year that keeps rising into the print while the long end stalls is the migration, and 5.45 holds. The two-year kept rising, 4.63 to 4.92; the long end did not stall, it went to 5.56 through the 5.45 the note called unbroken. Half the read held and the half that was a level did not. | Move2Y 4.63 → 4.92 · 30Y 5.35 → 5.56 | Verdict~ Partial ○ Open |
| Pre-registered tests · FOMC 16 September | ||
The vote A rise with two dissents or fewer confirms the committee has joined the curve. It printed 12 to 0, the first rise since 2023, to a 3.75 to 4.00 range. | Move→ 12 to 0 | Verdict✓ Held ● Resolved |
The statement and the dots One further rise beside the delivered one matches the strip, with the energy clause named as the thing to watch. The median carried exactly one, 4.1% for 2026 and 4.1% for 2027, a plateau not a cycle, and the energy language was deleted. | Move→ 4.1 / 4.1 | Verdict✓ Held ● Resolved |
The curve by the Thursday settle Registered as a rise that flattens, the two-year up more than the thirty-year or the long bond lower but inside 5.05. Ambiguous as published, and the desk does not resolve its own ambiguity in its own favour. On the owed settle the long bond was lower at 5.29 but nowhere near 5.05, and 2s30s widened a basis point, 61 to 62. Over the fuller window the flattening arrived, to 53bp on 21 September, then reversed to 64. The wording is the desk's fault. | Move72 → 62 → 64bp | Verdict? Ambiguous ● Resolved |
The chair on the pressure Registered verbatim: not the words, which the desk does not score, but whether the front end holds the rise through the press conference. The two-year went 4.67 to 4.74 on the decision and back to 4.67 at the 17 September settle. It gave back the day's rise exactly, on the settle the test was owed on. The front end has since gone to 4.92; the test was written to be scored on the owed settle and it is scored there. | Move4.74 → 4.67 on the owed settle | Verdict✗ Missed ● Resolved |
The delivery The desk pre-committed in print to score the four tests above on the 17 September settle. They were scored on 28 September, seven sessions late, the second consecutive late delivery on a dated pre-commitment and the first to go unacknowledged at the time. Logged at the weight of a missed level. The fix is structural and in the 29 September note: publish what has settled, register what has not. | Moveowed 17 Sep → scored 28 Sep | Verdict✗ Missed ● Resolved |
| Pre-registered tests · The issuer | ||
The buyback queue Horizon: tactical, event-driven Built on a single number, $19.868bn of offers against a $2bn door in the 20 to 30 year sector, and registered the test in dollars: offers well below $20bn at the first enlarged operation would be the first evidence against it. Scored missed on 11 September at $10.489bn in ten to twenty year, and retired on 29 September after the like-for-like 24 September operation in the measured sector drew $10.468bn. Three operations, three prints near ten billion. What replaces it is narrower and in print: a bid, not a queue. | Move$19.9bn → $10.5bn | Verdict✗ Missed ● Resolved |
| Pre-registered tests · October | ||
The 2 October payrolls Horizon: tactical, event-driven Registered before the print: a two-year that holds 4.80 through it says the second rise is real. One that slips through 4.65 says the strip got ahead of the committee. Scored on the settle. | Move4.92 → the print | Verdict◆ New ○ Open |
The 1 October buyback Horizon: tactical, event-driven The first 10 to 20 year operation since the queue claim was retired. Offers near $20bn would say the desk retired it too early, and the desk has committed in print to writing that. | Move→ the operation | Verdict◆ New ○ Open |
Credit, the tell Horizon: tactical, event-driven 3.10 and widening is the single development that moves the desk to the growth path on its own. The 2.90 condition was crossed on 25 September and 3.02 printed on the 28th; the tell has not fired. The next two weekly prints decide whether this was the start of something or a supply tantrum passing through. | Move3.02 vs 3.10 | Verdict◆ New ○ Open |
Breakevens, the discriminator Horizon: tactical, event-driven Inflation compensation falling while nominal yields rise reads as term premium and supply, not an expectations shock. Through May's 2.50 high, the story becomes a credibility problem and the map changes with it. 2.34 at the 28 September close. | Move2.34 vs 2.50 | Verdict◆ New ○ Open |
| FX | ||
DXY The dollar trades a firm 98 to 100 range, held up by rate differentials and a constrained Fed. Firm was right and the range was low: it held the band through August, then broke above it to 101.2 as the Fed delivered and the front end went to 4.92. | Move99.2 → 101.2 | Verdict~ Partial ○ Open |
DXY The dollar holds its range: no dovish softening, because the market never repriced the Fed lower. It eased to the range floor in August, then went through the top of it in September on the rise. No softening; the range broke upward. | Move99.9 → 101.2 | Verdict~ Partial ○ Open |
DXY The dollar holds firm: the relative winner in a world where every major central bank is boxed in. It softened toward the range floor in August, then went to 101.2 as the Fed moved first and hardest. | Move99.5 → 101.2 | Verdict✓ Held ○ Open |
DXY The dollar firms at the margin on US tech-export risk as the AI-capex story wobbles. It firmed, handed the move back, then firmed decisively into the September rise. | Move99.9 → 101.2 | Verdict✓ Held ○ Open |
DXY Firm but unpaid for its front end, 98.5 to 100.5. Firm was right, and the front end got paid: 101.2 on 28 September, through the top of the band, as October's second rise priced near seventy per cent. | Move99.1 → 101.2 | Verdict~ Partial ○ Open |
USD/JPY USD/JPY drifts toward 160, the level where Japan has historically intervened: intervention risk builds. It traded around 160 all summer, closed back through the line on 28 August, then went 160.39 to 155.30 in two sessions in the first week of September as the speculative short was washed out, and the Bank of Japan raised to 1.25% on 18 September. 160 belonged to Tokyo, as the note said. | Move159.7 → 157.4 | Verdict✓ Held ○ Open |
USD/JPY USD/JPY drifts lower only gradually as the BoJ-driven carry unwind is a slow burn. Lower it is, though it came as a gap rather than a grind: 155.30 in the week of 1 September, 157.4 at the 28 September settle with the Bank of Japan at 1.25. | Move160.5 → 157.4 | Verdict✓ Held ○ Open |
USD/JPY Grinds 152 to 157 as the valve closes. It ground higher through the band to 158.9 on 24 September as the front end repriced, and closed 28 September at 157.4, a fraction through the top. The valve closed as described, the Bank of Japan at 1.25; the band was a touch low. | Move154.5 → 157.4 | Verdict~ Partial ○ Open |
| Credit | ||
HY OAS High-yield spreads stay contained at 250 to 290bp: no credit stress beneath the rate move. Contained for four months to a 2.60 low on 28 August, then 42bp wider in a month to 3.02 on 28 September, through the band. No credit event, and the desk's published tell at 3.10 has not fired, so the spirit holds and the band does not. 3.10 and widening scores this missed. | Move~270 → 302bp | Verdict~ Partial ○ Open |
HY OAS High-yield spreads stay calm at 2.60 to 3.00%: the long-end pressure is supply, not fear. Calm through the summer; 3.02 on 28 September is two basis points through the ceiling with the tell at 3.10 unfired. Supply, still, on the breakeven evidence, and the desk's own condition crossed. | Move2.71 → 3.02 | Verdict~ Partial ○ Open |
HY OAS High-yield spreads hold 2.70 to 3.00% through the AI-equity selloff: a discipline trade, not a credit event. The selloff reversed and the band held through it; the break to 3.02 came in late September on rates, not on equities. | Move2.78 → 3.02 | Verdict~ Partial ○ Open |
HY OAS High-yield spreads stay pinned 2.70 to 3.00%: the cross-check that the equity pressure is sector-local. It was. The September break to 3.02 came with the Nasdaq near its highs, which is the rates channel, not the equity one. | Move2.78 → 3.02 | Verdict~ Partial ○ Open |
HY OAS Drifts, 2.60 to 2.90. It drifted to 2.68 on 22 September, then 25bp in four sessions to 2.93 and 3.02 on the 28th, through the condition the note published for the growth path. Not the tell, which is 3.10 and widening, and the desk did not reweight on four sessions; a registered condition of its own framework was crossed and the 29 September note said so. | Move2.70 → 3.02 | Verdict~ Partial ○ Open |
| Commodities | ||
Brent Brent trades a 90 to 100 range while the Iran war premium persists. It fell out of the range on the ceasefire, returned as it lapsed, sat just under the band at 88 in late August, ran to 106 on the Saudi pipeline shutdown and settled 28 September at 98.4, inside the range, as the pipeline restarted. | Move→ 98.4 | Verdict✓ Held ○ Open |
Brent Brent eases on the ceasefire while the long-end floor stays: energy relief lands in oil, not in duration. The relief held until the ceasefire expired and Brent recovered. | Move85 → 78 → 98.4 | Verdict↻ Rotated ● Resolved |
Brent Brent stays rangebound but elevated on the unresolved Iran conflict. Elevated again as the ceasefire lapsed, and more so on the pipeline: 106 at the September peak, 98.4 at the settle. | Move→ 98.4 | Verdict✓ Held ○ Open |
Brent Horizon: tactical, event-driven The strait as a ratchet, not a panic; the pain trade is de-escalation; the 98th-percentile crude long is right until it is not. Brent went 90.49 to 106 without a gap day, the long grew and was right, then Saudi Arabia restored the East-West pipeline and the pain trade arrived: 98.4, and the long is the one paying. | Move90.5 → 98.4 | Verdict✓ Held ○ Open |
Brent 100 to 110, products lead, the crack wide. Held the band for two weeks to 106, then the pipeline restart took Brent back under 100 to 98.4, below the band. That is half the desk's registered disarm for the shock path; the crack back through 55 is the other half and is scored when it settles. | Move104.6 → 98.4 | Verdict~ Partial ○ Open |
Gold Horizon: tactical, event-driven Scored the gold trigger as fired at 4,449: under an official bid at the long end, gold carries the credibility signal unaided. The note named the reverse in advance, a return below 4,449 while the two-year leads yields higher, as the one combination that would argue the hawkish case is real. It ran to 4,680 on 24 August, 5.2% above the trigger, then the two-year led and gold closed back under the line at 4,325 on 1 September. The reverse fired, and what it argued for arrived: a unanimous rise. Ran, then closed. | Move4449 → 4680 → closed 4325 | Verdict✓ Ran, then closed ● Resolved |
Gold Gold consolidates in a 4,400 to 4,600 range. It drifted below the band on rising real yields, recovered into it, ran to 4,680 in late August, then fell through the floor of the band as the real ten-year went from 2.55 to 2.90: 4,127 on 28 September, 6% under the band. | Move4485 → 4127 | Verdict✗ Missed ○ Open |
Gold Gold trades neutral: no clear push from the AI-equity story either way. Rangey since, a record in August and 2% under the publish level by late September, on rates rather than on equities. | Move4222 → 4127 | Verdict✓ Held ○ Open |
Gold Gold consolidates after its post-print unwind, then firms from 4,047 as the fiscal hedge reasserts. It firmed to 4,680 and has given nearly all of it back to real yields; 80 dollars above the base at the 28 September settle. | Move4047 → 4127 | Verdict✓ Held ○ Open |
Gold Gold consolidates at elevated levels as the fiscal-and-bid hedge, not the war hedge. It held and pushed to records into August, then fell 12% from the peak as real yields rose 35bp in three weeks. Elevated by any longer measure, and 5% under the publish level. | Move4331 → 4127 | Verdict~ Partial ○ Open |
Gold Gold lower on the dollar bid: with the war premium gone, gold trades the dollar. It went the other way first, to a 4,680 record in August while the dollar softened, and this page scored it missed at the 28 August mark. Since then the dollar went to 101 and gold followed it down through the publish level, on the mechanism the note named. Scored on the tape, as every row is, with the excursion against it stated in full. | Move4188 → 4127 | Verdict✓ Held ○ Open |
Gold Consolidates under 4,449 with the two-year leading. Under it throughout, 4,127 at the settle, and the two-year led every session that mattered. | Move4348 → 4127 | Verdict✓ Held ○ Open |
| Equities | ||
Semis (SMH) Semiconductors keep leading the tape higher on the AI-infrastructure build. They rolled over in June, fell to 542 in early September and have recovered to the publish level without leading anything: flat since 1 June against an S&P up 1%. | Move608 → 600 | Verdict✗ Missed ○ Open |
S&P 500 The S&P 500 grinds higher on ample liquidity. Higher, and only just: 1% above the publish level with a 5.24 ten-year overhead. | Move7600 → 7684 | Verdict✓ Held ○ Open |
Nasdaq The Nasdaq goes range-bound to 5 to 10% lower as the AI valuation premium reprices. The thesis-direction drop came early (about 7% into the 10 June low) then reversed: open, currently offside by 3.4%, with the stated invalidation (HY OAS through 3.50) not fired at 3.02. | Move25930 → 26820 | Verdict◐ Early ○ Open |
Nasdaq / AI The AI complex stays a range-bound, relative laggard. A forward-guidance call on the AI valuation reset: the tape has run against it since, open and currently offside, pending the stated invalidation. | Move25170 → 26820 | Verdict◐ Early ○ Open |
Semis (SMH) Semiconductors underperform quality mega-caps as capex discipline bites. Since 9 June SMH is up 1.5% against 4% for the S&P: the underperformance call holds, less emphatically than at the September low. | Move591 → 600 | Verdict✓ Held ○ Open |
S&P 500 Range-bound, rates a ceiling; energy over the consumer complex. 7,552 to 7,765 across the fortnight, 0.4% from where it started, with the ten-year at 5.24 above it. | Move7657 → 7684 | Verdict✓ Held ○ Open |
How to read this. Held: the asset behaved as described. Partial: it broke the exact band but in the spirit of the call (for example credit staying calm but tightening through the stated floor). Rotated: the base case held until a scenario change the note had explicitly mapped flipped it, so a reader following the map rotated on the trigger rather than holding to a loss (the Brent ranges held until the ceasefire). Mixed: the specific line was wrong on direction but the umbrella call it sat under was right (see the note below). Ambiguous: the registered wording could be read more than one way, and the desk does not resolve its own ambiguity in its own favour: scored neither held nor missed, counted against the headline figure, disclosed in full. Early: a forward-guidance thesis where the thesis-direction move came but the desk was early; the tape has since reversed and the call is open, currently offside, pending its stated invalidation. Ran, then closed: the thesis made its pre-registered threshold available (100 pips, 15bp or 1%) while it was open, and its stated invalidation later fired; the move and the close are both shown. New: a view first stated in the latest note and tracked from here, with no outcome yet. Missed: it went the other way, including a call that invalidated without first making that threshold available. Scored honestly, the rates / FX / Fed spine runs around 85%. A perfect-looking record would not be believable. This one is real.
A note on how to use the calls. Our research is forward guidance, not an entry-at-publish signal. We call a thesis ahead of the move; the reader times execution. So we score against the scenario and its stated invalidation, not against the publish-day price, and the path sparkline shows when the thesis-direction move actually occurred. An early call the tape later confirmed in direction (the equity-caution calls saw the Nasdaq fall about 7% and semis about 6% in early June before they reversed) is shown honestly as Early and open, never dressed up as a closed win.
On entry, exit and what the two figures mean. Our research is forward guidance, not an entry, exit or rotation signal. We publish the thesis and every scenario, path and invalidation level in advance; the reader makes their own entry, sizing and exit decisions. The spine's two figures reflect exactly that. The peak is the maximum favorable move the thesis made available while it was in play: what a reader who used the analysis and its mapped scenarios could have captured at the time, on their own execution. The now is the same call measured from the published price and held, with no timing assumed, to the as-of date. The published price is rarely the best entry, so a reader who worked the pullbacks the path shows could have done better than the now figure, while one who chased or held past the opportunity could have done worse. That is the point of publishing the scenarios in advance: the reader, not the desk, times the trade. We call the paths; the reader trades them.
How we keep score
When a call is open, complete or invalidated
This page is a point-in-time record, reviewed and refreshed weekly, with the as-of date stamped at the top. It is not an intraday ticker. Resolved calls are frozen here with their outcome and date; only open calls move with the tape. Where a later note has scored an earlier one, the verdict here is that note's verdict transcribed, not a second opinion.
Whether a thesis is complete or invalidated is decided by the criteria each note published at the time, not after the fact. Every Premium states its scenario triggers and invalidation levels in advance, and we hold ourselves to them:
- Open: the scenario is still in play. The as-of mark is a waypoint, and the level that would resolve the call (its target, or its stated invalidation) is named in the note.
- Complete: the base case reached its stated target or horizon. Locked with the date.
- Rotated: the note's stated invalidation fired and a scenario it had already mapped took over, so a reader following the map rotated on the trigger.
- Invalidated: the stated invalidation fired against the call. Locked with the date and the level. A closed call keeps the move it made available on the page, because deleting it would misreport what the analysis showed while it was live, and it states its close plainly alongside.
Two different things can happen when an invalidation fires, and we separate them rather than collapse both into one red mark. A thesis that ran a long way before its invalidation fired is not the same as one that never worked, and a record that scored them identically would be telling you less than it knows. So we mark a closed call Ran, then closed when it made at least 100 pips, 15 basis points or 1% available before the trigger fired, and Missed when it did not. That threshold is fixed in advance, published here and applied to every call equally, so the distinction can never be chosen after the event to flatter a result. The move made available and the close are both shown, and you can weigh them yourself.
Each row also carries a status. ○ Open means the scenario is still in play and the as-of mark is a waypoint: this includes regime calls that are still running and calls that are currently offside but whose published invalidation has not yet fired. ● Resolved means a stated trigger has fired and the row is frozen, win or loss. We do not retire a call early or late, only when its own criteria say so. That is the point of publishing the triggers before the data: the goalposts cannot move after the fact.
Levels: US Treasury yields via FRED, extended with same-day end-of-day government-bond prints for the one or two sessions FRED has not yet published, so that every sparkline endpoint matches its stated current level; FX, commodities and equities via market tape; entry levels are the close on or nearest each note's publish date, current levels as of 28 September 2026. This page is editorial commentary on observable market developments and a record of published views. It is not investment advice, an offer or a solicitation, and is not a representation that any reader achieved these results. Past performance is not indicative of future results. We read the data. We call the paths.