Paper Portfolio
$1,000 theoretical portfolio — Kelly-sized. Live YES prices from the Polymarket API.
Total P&L (realized + unrealized)
+$743.25
unrealized loads live
Unrealized P&L (open)
—
1 open · $100 staked · Live prices
Realized P&L · Win rate
+$743.25
86% win rate (19/22 correct)
Open Positions1
| Market | Side | Entry YES | Live YES | Stake | P&L | Notes |
|---|---|---|---|---|---|---|
| US Unemployment 2026 >= 5.0% | NO | 25% | — | $100 | — | Sep 9: YES 9.75% mid LIVE (bid 9.7 / ask 9.8), FLAT for a THIRD consecutive session and still 2.20 above where it sat before the August employment report. FV unchanged at 6 on the YES. SIXTH SESSION WITHOUT AN EXPLANATION AND WE ARE STILL NOT INVENTING ONE. Payrolls printed +162,000 against a 53-60,000 consensus, July was revised from -23,000 to +23,000, and the unemployment rate HELD at 4.1% when consensus had it at 4.2% - and this leg has gone 7.55 to 9.25 to 9.75 and stayed there. The candidates remain unverified: thin liquidity, a hiking Fed fattening the 2027 policy-error tail, internals narrower than the headline. We log the disagreement rather than pick the flattering one. THE POSITION SURVIVES BECAUSE IT IS STILL CHEAP ON OUR OWN NUMBERS AT THE BID. Fair value 6 on the YES makes the NO worth 94 against a 90.2c size-aware bid - 3.8 points cheap, the same test that closed pos-013 and pos-010, and the same test that condemned pos-011 this morning. 133.3333 shares clear 70 at 90.2c and 63.3333 at 90.1c for $120.20, up $0.17 on depth alone from yesterday's $120.03. Getting 4.1% to 5.0% by December still needs roughly 1.4 million additional unemployed in under four months, and the August report made that harder. THIS IS NOW THE ENTIRE BOOK, and the uncomfortable corollary is published with it: the fair value of 6 is a CARRIED number, ours, unmeasured and unaudited - which is precisely the property that killed three models in three weeks, including one this morning. Under today's new standing rule it does at least name its exogenous input: the 1.4 million-unemployed arithmetic comes from BLS levels, not from this contract's price. GATES, evaluated on the DAILY SETTLE, BOTH UNCHANGED AND NEITHER TRIGGERED: the review line at YES below 8c is NOT active, the leg is 1.75 above it; the hard auto-close at YES below 5c is untriggered and carries no discretion. NEITHER GATE DEPENDS ON A US RELEASE LATER TODAY - the only US datum on the calendar is July wholesale inventories at 10:00 ET against a -0.1% consensus, which touches nothing here - so nothing is deferred. Next real test is Friday's August CPI at 08:30 ET, then the September employment report in early October. |
Closed Positions22
| Market | Side | Entry YES | Exit YES | Stake | P&L | Outcome | Note |
|---|---|---|---|---|---|---|---|
| Fed Rate End 2026 = 4.0% | YES | 34% | 40.11% | $25 | +$4.49 | Correct | SOLD 2026-09-09 at a volume-weighted 40.11c. NOT ON A GATE - the 50c efficiency review was 8.1 points away and untriggered. We closed it because we went hunting a better input and found a broken method instead. THE METHOD, WRITTEN OUT AS ALGEBRA: the fair value we published every day for weeks was FV = market_4.0 x (our P(at least one hike in 2026) / the market's own hike mass). The price of the thing being valued appears on both sides. If our P(>=1 hike) ever equalled the market's implied one, the model would return the market price exactly. It carried NO information about the question that actually decides this contract, which is not whether the Fed hikes but whether it hikes ONCE or TWICE. Feed it P(>=1)=0.658 and it says 38.9; feed it 0.890 and it says 52.7 - fourteen points of swing from a single scalar, with zero shape content. THE REPLACEMENT PRICES THE SHAPE, AND ITS RESOLUTION RULE WAS VERIFIED IN THIS SESSION rather than assumed: the market description settles on the upper bound AFTER the December 8-9 FOMC, so Sep 16, Oct 28 and Dec 9 all count - the endDate field reads 2026-12-09T00:00:00Z and would have implied the December meeting does not, which is the kind of error that ends a position badly. With the target band at 3.50-3.75% and EFFR 3.63%, the 4.0% leg means exactly ONE net hike across three meetings. A fed-funds-futures path read this morning gives per-meeting hike probabilities of 62.8 / 23.6 / 61.2, cumulative 1.48 hikes. Counting paths at zero correlation: P(exactly one) = 0.628x0.764x0.388 + 0.372x0.236x0.388 + 0.372x0.764x0.612 = 39.4%. Fed meetings are positively correlated, which moves mass out of the middle into the tails, so 39.4 is a CEILING: a Gaussian one-factor model with marginals held fixed gives 35.0 at rho 0.2, 30.5 at 0.4, 25.4 at 0.6, 18.8 at 0.8. Against the 40.2c size-aware bid the edge runs -0.8 / -5.2 / -9.7 / -14.8 / -21.4pp. WHY WE ACTED WHERE ON MONDAY WE DID NOT: on Sep 8 we refused the CPI trade because the SIGN flipped on an input we could not measure. Here it does not flip. Substituting Polymarket's own 52.5c September price for the futures 62.8% produces exactly one positive cell in the entire grid - +0.2pp on the mid at zero correlation - and negative everywhere else. There was no reading on which we were holding something cheap. EXECUTION, PUBLISHED AS A LADDER SO IT CAN BE AUDITED: 73.5294 shares from a $25 stake at 34c. The book paid 35.49 shares at 40.2c, 10 at 40.1c and the remaining 28.0394 at 40.0c. Proceeds $29.49, volume-weighted 40.11c, realised +$4.49 on $25 - an 18% return and a small winner. Record goes to 19/22. WHAT THIS EXIT COST, PUBLISHED WITH IT AND NOT AFTER IT: we acted on a model built this morning off a THIRD-PARTY futures dashboard we could not corroborate - CME FedWatch itself was unreachable this session. And the two markets still disagree by half a hike that we cannot explain: the Polymarket ladder renormalises to 0.97 net hikes by year-end against the futures path's 1.48. We closed a position on an anchor that is in open conflict with the market it was pricing, and a reader is entitled to weigh that against the $4.49. We also crossed a spread to act on a 2.5-point view at a size our own Sep 8 finding says cannot carry execution costs. SECONDARY BENEFIT, NAMED SO IT IS NOT MISTAKEN FOR THE REASON: this removes a leg from the hawkish-Fed cluster the correlation cap has flagged for weeks. The book is now a single position. |
| Fed Rate Hike in 2026 | YES | 55% | 70.0% | $12.5 | +$3.41 | Correct | CLOSED 2026-09-07 at the 70c bid, and the reason is consistency rather than a gate. NO WRITTEN TRIGGER COVERED THIS. The 72c trim is disarmed - it re-arms only below 68c and the leg has not been there since Friday - and the 48c floor is twenty-two points away. What changed is our own number. FAIR VALUE CUT 70 -> 65, BUILT TWO WAYS SO IT CAN BE AUDITED. First, the meeting route: Polymarket has a September hike at 48.5c and CME FedWatch at 49.4%, blending to 48.6; conditional on no September move, 30% across the October and December meetings gives 0.486 + 0.514 x 0.30 = 0.640. Second, the ladder route: the end-2026 rate ladder prices 4.0% at 41.20, 4.25% at 19.05 and 4.5%-or-more at 4.85, a hike mass of 65.10 - which is the market's own direct answer to 'does the rate end above 3.75%', the same question this contract asks. Sixty-four and 65.1, from two independent constructions, so we publish 65. The leg trades 70.50 mid, 70.0 bid. WE ARE MARKED 5.5 POINTS RICH. AND HERE IS WHY WE CANNOT HOLD IT. We closed pos-013 this morning for being ONE point above its fair value. Holding pos-010 at five and a half would mean applying a stricter test to the position we wanted to sell than to the one we wanted to keep, which is not a method, it is a preference with arithmetic attached. So both go, on the same test, in the same session. EXECUTION: 22.7273 shares into a 70c bid, proceeds $15.91 on $12.50 of stake, realised +$3.41. The honest counterweight, published because it argues against us: this stub still had real convexity - $6.82 of upside against $15.91 at risk, 2.3 to 1, far better shape than pos-013's 12.9 to 1 - and a book that sold it anyway on a consistency argument should say so. WHAT THE WHOLE POSITION DID: entered at 55c on $200 in June, exited across four trims and a close at 66.5, 70, 71, 71 and 70. Realised +$48.87 in total. The trims that felt like discipline sold two thirds of it between 70 and 71 while fair value ran from 67 to 73 and back to 65, which is a gate doing its job in a range and looking foolish at both ends. |
| Zero Fed Rate Cuts in 2026 | YES | 77.6% | 92.8% | $41.0 | +$8.03 | Correct | CLOSED 2026-09-07 at the 92.8c bid. THE PRE-COMMITTED REVIEW FIRED AND THIS TIME WE TOOK THE WHOLE THING. On Sunday we trimmed half on the 90c efficiency gate, wrote that the arithmetic said all of it, and said in as many words that we had chosen the trim convention over the pos-002 and pos-003 precedent - convention beating arithmetic rather than analysis. We then pre-committed: a fresh review at the next run with a full exit on the table if the leg was still at or above fair value. It is, so here is the review and here is the exit. THE ARITHMETIC, REBUILT ON TODAY'S NUMBERS RATHER THAN SUNDAY'S. Polymarket's own meeting ladder now prices a cut at 0.50% in September, 3.70% in October and 8.15% in December. Under the perfect-correlation assumption - the most generous one available to us, because it treats a cut at any meeting as one regime shift rather than three independent draws - the ceiling on no-cut-in-2026 is 1 minus 8.15, or 91.85. Under independence it is 88.0. The leg trades 92.85 mid, 92.80 bid. We are marked ONE FULL POINT above the most flattering fair value we know how to construct, and 4.85 above the honest one. THE UPSIDE THAT REMAINED: 52.8351 shares at a 92.8c bid realise $49.03 today against $52.84 at resolution. That is $3.80 over 115 days, on $49.03 of capital carrying a 7.15% chance of going to zero - 12.9 to 1 against, on a position whose own market says the edge is negative. There is no version of this book's method that holds that. EXECUTION: 52.8351 shares into a 92.8c bid carrying 4,969 shares. No walk, no slippage. Proceeds $49.03 on $41 of stake, realised +$8.03. Combined with Sunday's pos-013a trim the position returned +$16.11 on $82 of original stake. THE LESSON WE ARE TAKING, AND IT IS NOT A FLATTERING ONE: the right answer was available on Sunday and we did half of it because halves are what this book does. The convention cost us nothing here - the leg is five hundredths of a cent from where we sold half of it - and that is luck, not vindication. What fixes it is the pre-commitment, which is why it was written. |
| Fed Rate Hike in 2026 (fourth trim) | YES | 55% | 71.0% | $12.5 | +$3.64 | Correct | TRIMMED 2026-09-06 at the 71c bid. THE HYSTERESIS RULE ADOPTED ON WEDNESDAY DID PRECISELY WHAT IT WAS WRITTEN TO DO, AND WE WANT THAT ON THE RECORD BEFORE THE COMPLAINT THAT FOLLOWS IT. On Wednesday the 72c gate had fired twice in 26 hours on half-cent breaches and we called it a liquidation schedule dressed as discipline. The fix was a re-arm condition rather than a wider number: 72c stays, and only re-arms below 68c. Thursday Sep 3 at 21:00 UTC the leg printed 60.50c on Governor Waller saying he would support holding rates if inflation kept converging - an eleven-point drawdown, unmistakably a repricing and not a tick. The gate re-armed. Friday Sep 4 at 15:00 UTC, after payrolls came in at +162,000 against a 53-60,000 consensus, it printed 72.50c and fired. Signal in, noise out. The mechanism separated them. 22.7273 shares - half of 45.4545 - sold into a 71c bid with 1,503 shares behind it; proceeds $16.14 on $12.50 of stake, realised +$3.64. THE COMPLAINT: it fired Friday afternoon and we are executing on Sunday, fifty hours later, at 71c against a 72.5c trigger. That is 1.5 points and $0.34 - trivial money and the FOURTH consecutive gate this book has honoured after the fact rather than at the moment. Two of the four were expensive. We have now published this same structural failure four times without fixing it, and repeating a confession is not the same as correcting a process. The gate is not the problem; the calendar is. ONE THING WENT RIGHT FOR ONCE: fair value fell 73 -> 70 on the fed funds futures repricing to 58% for September, so this trim executed one point ABOVE our own number, after three consecutive trims that executed below it. $12.50 and 22.7273 shares remain. Floor 48c unchanged since June; the 72c gate is disarmed again until the leg prints below 68c. |
| Zero Fed Rate Cuts in 2026 (efficiency trim) | YES | 77.6% | 92.9% | $41.0 | +$8.08 | Correct | TRIMMED 2026-09-06 at the 92.9c bid. THE 90c EFFICIENCY REVIEW FIRED ON FRIDAY AND THIS IS THE FIRST GATE IN THIS BOOK THAT PAID FOR WAITING. It had sat 1.2c away for three sessions. On Thursday evening Waller's dovish remarks knocked it to 88.65c - further from the trigger than it had been all week. On Friday the August employment report printed +162,000 against a 53-60,000 consensus with unemployment holding 4.1%, and the leg gapped to 93.15c at 11:00 UTC. 52.8351 shares - half of 105.6701 - sold into a 92.9c bid carrying 507 shares, so no walk and no slippage. Proceeds $49.08 against $41 of stake, realised +$8.08. WHY THE REVIEW CONCLUDED SELL RATHER THAN HOLD, STATED AS ARITHMETIC: 105.6701 shares marked at $98.17 against $105.67 at resolution is $7.50 of remaining upside over 116 days, on capital that carries a 7.1% chance of going to zero. Polymarket's own meeting ladder prices cuts at 0.5% in September, 4.35% in October and 7.65% in December; under the honest correlation assumption that ceiling is 92.35% for no cut in 2026, so at 92.9 we were marked ABOVE our own fair value of 92. The edge was not thin, it was negative. AND WE ONLY TOOK HALF, WHICH IS THE PART WE ARE LEAST SURE OF. pos-002 and pos-003 were both closed ENTIRELY on this exact arithmetic in July. The consistent answer was the whole position. We took half because every trim this book has executed has been a half, and because fully liquidating the anchor holding on a Sunday - with no letter published to explain it to readers who own the reasoning as much as we do - is a bigger discretionary act than a written trim rule authorises. Readers should weigh that we chose convention over the arithmetic and judge it accordingly. The remainder is pre-committed to a fresh review with a full exit on the table at the next run if the leg is still at or above fair value. $41 and 52.8351 shares remain. |
| Fed Rate Hike in 2026 (third trim) | YES | 55% | 71.0% | $25.0 | +$7.27 | Correct | TRIMMED 2026-09-02 at the 71c bid. THE 72c GATE FIRED A SECOND TIME, at 72.5c at 06:00 UTC this morning, and this time we saw it within two hours rather than fifteen. 45.4545 shares - half of the remaining 90.9091 - sold into a 71c bid carrying 9,092 shares, so no walk and no slippage. Proceeds $32.27 against $25 of stake, realised +$7.27. THE PROBLEM WITH HONOURING THIS ONE IS THAT IT IS THE SAME GATE AT THE SAME LEVEL TWO DAYS RUNNING, and a trigger with no hysteresis on a one-cent-wide market is not a gate, it is a liquidation schedule dressed as discipline. The breach was half a cent. We honoured it anyway, because refusing a trigger the day it becomes inconvenient is the failure mode we have spent two weeks writing rules against - but we are naming the design flaw in the same breath rather than waiting for it to cost something. FIX ADOPTED TODAY, AND IT IS A HYSTERESIS RULE RATHER THAN A MOVED TRIGGER: the 72c level stays exactly where it is, and it re-arms only after the leg prints below 68c. A gate that can fire twice in 26 hours on half-cent breaches would grind this position to nothing on noise, and we would rather fix the mechanism than quietly widen the number. $25 of stake and 45.4545 shares remain. Floor 48c unchanged. |
| Fed Rate Hike in 2026 (second half-trim) | YES | 55% | 70.0% | $50.0 | +$13.64 | Correct | TRIMMED 2026-09-01 at the 70c bid on the trim gate published since June - trim above 72c, floor 48c. THE GATE FIRED WHILE WE WERE NOT WATCHING, AGAIN, AND THIS TIME IT COST US. The hourly series has the leg at 74c at 14:00 UTC Monday, 72.5c at 15:00 and 75.5c at 17:00 - three hours above the trigger during the US session, five hours after Friday's letter published. It is 71.5c now. On Aug 31 we adopted the rule that every price-triggered gate is checked against the hourly series rather than one daily observation, and that a trigger which fires in our absence gets acted on at the next run rather than quietly re-dated. So we are acting: 90.9091 shares, half the $100 remainder, sold at the 70c bid where the book shows 5,681 shares of depth. Proceeds $63.64 against $50 of stake, realised +$13.64. THE HONEST ARITHMETIC OF BEING LATE: the gate fired at 75.5c and we executed at 70c. That is 5.5 points, or $5.00 on this half, surrendered purely to the gap between when the rule triggered and when we were next at the desk. It is the second time in three sessions that publishing at 04:00 ET has cost this book real money, and unlike Friday - where we called a 3.5-cent near-miss luck - this one is a bill. Worse: our own fair value on this position is 71 and we sold at 70, so the gate made us sell one point BELOW fair value. We did it anyway, because a gate you only honour when it agrees with your fair value is not a gate. $50 of stake and 90.9091 shares remain, with the floor at 48c unchanged. |
| 10Y Treasury Touches 4.8% Before 2027 | YES | 16.5% | 71.69% | $25 | +$83.61 | Correct | SOLD 2026-09-01 on the written take-profit gate, at a volume-weighted 71.69c against an 82c top-of-book bid, and the difference between those two numbers is the most useful thing this position ever taught us. THE GATE: sell on a 66c BID, published continuously since Aug 20 and one cent away on Friday, when we wrote that we were not pre-empting it and not moving it. Overnight the bid went to 82. It fired, decisively, and we sold the whole position. THE EXECUTION, PUBLISHED AS A LADDER SO IT CAN BE AUDITED: 151.5152 shares from a $25 stake at 16.5c. The book paid 10 shares at 82c, 10 at 81c, 3.53 at 73c, 14.82 at 71c and the remaining 113.17 at 70c. Proceeds $108.61, volume-weighted 71.69c, realised +$83.61 on $25 - a 334% return and the second-largest single winner this book has produced. EVERY SHARE FILLED ABOVE THE 66c TRIGGER, which is the one thing that makes this a clean execution rather than a story about slippage. WHAT IT COST US TO LEARN: we have been marking this position at the mid all along. At Friday's 87.5c mid the mark said $132.58. The realisable value was $108.61. We carried $23.96 of paper profit that was never available at our size, on a $25 ticket, for weeks - and we only found out because we finally tried to sell. From today this book marks every position at the BID. WHY THE PRICE MOVED: the 10-year settled 4.75% on Aug 31 (Treasury par yield curve), the highest since January 2025, leaving a 5.0bp barrier to the 4.80% touch, while the JGB 10-year hit 3.00% for the first time since 1996 and the Bund reached 3.3% for the first time since May 2011. Renewed US-Iran strikes put Brent above $91. The thesis was right and the market got there before we did, which has been the pattern for a week. SAME MORNING, THE OTHER SHOE: the pre-commitment written on Aug 31 - a fifth consecutive failure of the 5.00% consistency check retires the barrier model - fired today at 11.1pp. The model is retired to diagnostic-only. It never got to price this exit; the gate did. ||| ADDENDUM 2026-09-09 - THE CONTRACT RESOLVED AND THE BILL IS FINAL. The Treasury daily par yield curve prints the 10-year at 4.80% on 09/08/2026 and Polymarket's 4.8%-touch market closed and resolved YES at 22:21:34 UTC on Sep 8. Our 151.5152 shares pay $151.52 at resolution against the $108.61 we took on Sep 1. THE GATE COST $42.91 - 172% of the original $25 stake - and that number is now realised opportunity cost rather than a daily mark. Realised P&L on the trade is UNCHANGED at +$83.61 and the record still counts it as one win; this is not a P&L revision and is not double-counted anywhere. THE VERDICT STANDS: the 66c-bid trigger was published continuously from Aug 20, it fired on its own terms, and all five fills cleared above it. A book that moves its gates when they hurt has no gates. THE UNCOMFORTABLE COROLLARY, PUBLISHED RATHER THAN OMITTED: the barrier model we retired on Sep 1 after five consecutive consistency failures had one live reading left on this contract - 76.8 against a 67.5 market on Aug 31 - and the contract resolved at 100. On the last question it was asked, the model we threw away beat the market, and it was saying hold where the gate said sell. IT STAYS RETIRED: one correct directional call does not undo five straight cross-check failures, and the retirement was pre-committed in writing on Aug 31 before this outcome could be known, which is the only kind of rule worth having. But we will not record the retirement as costless, and we are not quoting that model again - which is why we publish NO fair value for the 5.0% touch leg now trading 40.0c with the 10-year at 4.80%. |
| US GDP Q2 = 1.5-2.0% (advance) | YES | 18% | 100% | $25 | +$113.89 | Correct | WON. BEA Q2 advance landed inside 1.5-2.0%, exactly the GDPNow anchor (~1.7). The fully automated Friday-evening entry (18c, condition-gated) returns +$113.89 — the book's biggest single winner and validation of the conditional-entry system: model beat crowd, rule beat discretion. |
| Fed Rate Hike in 2026 (half-trim) | YES | 55% | 66.5% | $100.0 | +$20.91 | Correct | Half of pos-010 sold at 66.5c per the published >64c trim rule — thesis partially realized into elevated hike odds (38%%) as the oil engine cooled. First profit booked on the Fed complex. |
| China GDP Q2 = 4.6-4.9% | YES | 72.5% | 0% | $25 | -$25.00 | Incorrect | NBS printed 4.3% — outside the bracket. The five-session seller was right: Beijing chose candor (slowest growth since Q4 2022, Iran oil shock + weak demand). The print-regularity thesis assumed the statistical regime of 2020-2025; regimes change exactly when stimulus needs a permission slip. -$25, tuition for a regime lesson. |
| June CPI YoY = 3.9% | YES | 15% | 0% | $75 | -$75.00 | Incorrect | June CPI printed 3.5% YoY — below every forecast: our bracket (3.9), the crowd's favorite (3.8), and Cleveland's frozen 3.92. The model we trusted missed by 0.42pp, its worst miss in our sample; energy passthrough ran harder than the nowcast could see and core went flat (0.0% MoM). The winning bracket (≤3.6%) traded at ~2c pre-print — nobody held it. Full $75 stake lost. Second loss of the book; sized at 7.5% of budget for exactly this. Lesson kept: a frozen model is a point estimate, not a distribution — we paid 5:1 odds for its mode and the whole distribution shifted left. |
| Fed Hike at July 29 Meeting | YES | 14.6% | 24.5% | $25 | +$16.95 | Correct | Exited Jul 8 pre-minutes at 24.5c: +68% in 22 hours. The thesis was convergence to FedWatch (~26%), not a view on the decision — converged, so out. Fastest win of the book. |
| Strait of Hormuz Normal by Jul 31 | YES | 55% | 14.5% | $75 | -$55.23 | Incorrect | Exit trigger fired Jul 6: <20c before the Jul 19 checkpoint without confirmed transit recovery. IRGC corridor warnings + 8 weekend U-turns broke the 60-calls/day ramp thesis. First loss of the book. Sized at 7.5% of budget for exactly this scenario. |
| Fed Rate < 3.0% before 2027 | NO | 16% | 5.5% | $125 | +$15.62 | Correct | Early profit-take Jul 6 at 5.5c. The <3.0% path was dead (Fed at 3.50-3.75% with a hike bias). Same capital-efficiency logic as the pos-001 exit. |
| US Unemployment 2026 >= 6.0% | NO | 18% | 6.5% | $100 | +$14.02 | Correct | Early profit-take Jul 6 at 6.5c. Edge exhausted — remaining ~$8 over six months wasn't worth $114 of tied capital. Recycled into the zero-cuts add. |
| Inflation 2026 > 4.5% | NO | 82% | 42% | $125 | +$277.78 | Correct | Early exit Jun 23 at ~42% YES (held NO). Iran MOU + WTI falling to $73 from $92 removes the path to 4.5%+ CPI in any remaining month. May peaked at 4.2% — the high water mark. Energy deflation now dominant. Extracted ~40pp from 82% YES entry. +$277.78 realized (+222% on stake). |
| US Recession 2026 | NO | 28% | 10% | $200 | +$50.00 | Correct | Exited Jun 22 at 10% YES (held NO). Hit stated exit target. FV ~8% — 2pp gap remaining not worth PCE Jun 25 two-way risk. Extracted ~18pp from 28% YES entry. |
| CPI May MoM = 0.6% | NO | 40% | 0% | $25 | +$16.67 | Correct | May CPI printed 0.5% MoM (below 0.6% threshold). Resolved NO on Jun 10. |
| CPI May YoY = 4.3% | NO | 40% | 0% | $150 | +$100.00 | Correct | May CPI printed 4.2% YoY (below 4.3% threshold). Resolved NO on Jun 10. |
| CPI May YoY >= 4.4% | NO | 38% | 0% | $150 | +$91.94 | Correct | May CPI printed 4.2% YoY (below 4.4% threshold). Resolved NO on Jun 10. |
| May Unemployment Rate = 4.3% | YES | 34% | 100% | $25 | +$48.53 | Correct | BLS confirmed May U/E = 4.3% on Jun 5 NFP report. Resolved YES. |
Position sizingas of 2026-09-09
Equity
+$1,763.45
Idle cash
+$1,663.45 (94.3%)
Entry bar
10pp
Max single position
20.0% of equity
Stake ladder $50 · $100 · $125 · $175 · $225 · $275 · $350
= 2.5% / 5.0% / 7.5% / 10.0% / 12.5% / 15.0% / 20.0% of equity, recomputed at month end
SIZING RESCALED TO EQUITY, EFFECTIVE TODAY, AND THE REASON IS A NUMBER WE HAD STOPPED LOOKING AT. This book runs a published $1,000 bankroll. Equity is now $1,763.04 - bankroll plus $738.76 realised plus $24.28 unrealised - and $1,613.76 of that, 91.5%, is sitting in cash. Only $125 is at stake, and no NEW position has been opened since July: August and September are both zero, and everything in between was trims and exits. THE STAKE LADDER WAS ALWAYS PERCENTAGES AND WE JUST STOPPED UPDATING THE BASE. The old $25 / $50 / $75 / $100 / $125 / $150 / $200 steps are 2.5% / 5% / 7.5% / 10% / 12.5% / 15% / 20% of a $1,000 book. Applied to today's equity and rounded to the $25 step they become $50 / $100 / $125 / $175 / $225 / $275 / $350, with a single-position cap of 20% of equity. From today the ladder is defined as those percentages and recomputed at each month end, so it never needs a manual rescale again. WHAT THIS DOES NOT DO, AND WE WANT THIS UNAMBIGUOUS: it does not touch the 10-point entry bar, the half-Kelly x confidence x liquidity formula, or the correlation cap. Nothing gets bought that would not have been bought yesterday. A qualifying 15-point edge at 0.7 confidence and 0.7 liquidity now deploys $125 instead of $75; a 20-point edge at full confidence deploys $275 instead of $150. THE HONEST RISK, PUBLISHED WITH THE CHANGE: scaling stakes to equity is pro-cyclical. It means the book takes more absolute risk after a winning run, which is exactly when a method is most likely to be overrated by the person running it. Half-Kelly is supposed to absorb that and we are not certain it fully does. We are making the change because a 91.5% cash position on a book whose entire purpose is to test a method is not caution, it is the method not being tested - but a reader is entitled to note that we rescaled upward after a 76% gain and to judge the next drawdown accordingly. SEP 8 UPDATE: the tiered entry bar proposed on Sep 7 is SHELVED unbuilt. Eight days of screening establish that this book's binding constraint is its own size rather than the entry bar - our measured execution costs eat the 5-10pp band the tier was designed to unlock. Low deployment is now published as the finding rather than treated as a problem to engineer around. SEP 9 UPDATE: the book is down to ONE position and 94.3% cash after pos-011 was closed - not on a gate, but because its published fair value turned out to be an algebraic restatement of the market price. Equity is unchanged at $1,763.45 (total P&L identical to yesterday to the cent), so the ladder is unchanged; it is next recomputed at month end as written. A NEW STANDING RULE APPLIES TO EVERY FUTURE SIZING DECISION: no position is sized off a fair value that cannot name at least one input exogenous to the market being priced. Three models have now been retired in three weeks - barrier, CPI bracket, Fed ladder - and all three would have failed that two-minute test on the day they were published.
Cumulative P&L over time — $ (realized; live unrealized loads in browser)