# Kalshi Politics — Mispricing Audit (Anchor, NO-side structural) — 2026-09-24

Date: 2026-09-24
Source: https://kalshi-audits.pages.dev/2026-09-24-kalshi-politics-mispricing-audit-anchor

---
Kalshi · Politics · Anchor methodology (strategy=Anchor, audit_version=anchor) · report date 2026-09-24 · horizon 45 days · capital $2,500

# Kalshi Politics — Mispricing Audit (Anchor, NO-side structural)

0 picks · 100% cash
Every Stage-0 survivor failed the Anchor entry band. Five NO legs sit at 90–95¢ (deep-favorite band) and the two Bank of Canada legs sit at 54–57¢ (below band). Zero is a normal, correct Anchor result.
$0allocated of $2,5000picks11logged rejects (6 Stage-2, 5 Stage-0)7Stage-0 survivors from 872 open markets2precedent rows written

## 1. Portfolio & position sizes

| Ticker | Side / limit | Contracts | Cost | Conviction |
|---|---|---|---|---|
| No positions. Reserve held: $2,500 (100%). Anchor minimum reserve 10% is satisfied trivially. |  |  |  |  |

## 2. Stage-0 universe funnel

| Step | Count |
|---|---|
| Allowlisted series in targets.json (deduped) | 115 |
| Open markets in those series (public API) | 872 |
| … closing inside 45-day horizon | 91 |
| … and expiring ≥7 days out | 60 |
| … and vol24 ≥ 500 (live) | 15 |
| … and spread ≤5¢, 5¢ < YES < 95¢, close ≤ expiration+1h (Stage-0 survivors) | 7 |
| Stage-2 diligence completed | 6 |
| Passed Anchor entry band (60–89¢ effective) | 0 |
| Final picks | 0 |

Universe = `/tmp/out/targets.json` (generated 2026-09-24T16:36Z, 120 rows / 115 unique series, tiers 1–2 only; no Tier 3 rows exist in the file). Live data from the Kalshi public API using `*_dollars`/`*_fp` fields; candlesticks from the series/markets candlesticks endpoint. Horizon cut: close_time ≤ 2026-11-08; 7-day cut on expected_expiration_time ≥ 2026-10-01. Mentions markets and cumulative-count markets were removed by hand where they slipped the classifier.

## 3. Stage-2 candidates — all rejected

status-quoscheduled-processNO @ 54¢

### Bank of Canada hikes 25bps at Oct 28, 2026 meeting?

KXCBDECISIONCANADA-26OCT-H25 · closes 2026-10-28 · cluster boc-oct-2026-decision

**REJECT — below band: NO entry 54¢ (YES 46/48) is under the 60¢ floor; model 40% pre-market → 43% final sits inside the OIS ≈ 50% coin-flip, and the loss tail (oil stays ~$100 and Sep CPI on ~Oct 20 prints hot) is a continuation tail**

**YES condition.** For YES to pay, the Bank of Canada must raise the overnight target from 2.25% to 2.50% at its fixed-date Oct 28, 2026 announcement, resolved by Kalshi on the BoC press release.

**Prob (for YES):** 40% pre-market → 43% final. **Book at signal:** YES 46/48, NO 52/54; 671 YES-bid contracts in top 3 levels, 1,312 NO-bid; vol24 2,733; OI 29,865.

**Vehicle / calendar.** Vehicle exists and is dated: BoC fixed announcement Oct 28, 2026 with a Monetary Policy Report (BoC Sep 2 press release). Calendar before it: Sep jobs (~Oct 9), Business Outlook Survey (~Oct 19), Sep CPI (~Oct 20).

- Sep 2, 2026: BoC held at 2.25% for the 7th straight decision; statement: 'excluding gasoline, inflation was 2.2% and measures of core inflation remained close to 2% in July' and Council 'is prepared to adjust monetary policy as needed' (bankofcanada.ca).
- Sep 14, 2026 StatCan: August CPI 3.0% y/y (unch), CPI-trim/median average 2.0% (unch), ex-gasoline 2.4%. The Bank's own core trigger (trim/median above 2%) is still unmet.
- Sep 21, 2026 Halifax: Macklem — 'nor do we want to be too slow to respond if inflationary pressures are becoming more persistent' (Bloomberg/CMT). Globe and Mail Sep 18: markets a 'coin flip', 'narrowly in favour of a hike'; RBC, Desjardins, Capital Economics call hold; money.ca reports Scotiabank/National/RBC calling a hike — sell-side is split.
- Price path: H25 15¢ on Sep 10 → 32¢ Sep 15 → 43¢ Sep 21 → 46/48 today; +4¢ in 48h. Hold leg 84¢ → 57¢ over the same fortnight.

**Cleanest loss tail.** Oil stays near $100, September CPI (Oct 20) prints hot, and the Bank follows through on the Sep 21 warning. That is the current trend continuing, not a new discrete event — a continuation tail Anchor rejects even if the price were in band.

**Resolver.** Kalshi on the BoC press release; 12 precedent rows show the resolver reads 'Hike of 0bps' as hold and pays each rung mechanically. Resolver deductions: 0 pts.

**Precedents consulted.** KXCBDECISIONCANADA-26JUL-H25 (no): 'Two-sided guidance … keeps both middle rungs alive in the narrative while neither gets priced'; KXCBDECISIONCANADA-26SEP-H25 (edge note, Sep 22): 'The NO-hike leg only becomes an Anchor-shaped trade when it is inside the 60–84¢ band with the last pre-meeting CPI already printed; at parity with OIS before that print there is no edge either side.' Verified against live rules: the leg pays only on an actual 25bp increase.

status-quoscheduled-processYES @ 57¢

### Bank of Canada holds (0bps) at Oct 28, 2026 meeting?

KXCBDECISIONCANADA-26OCT-H0 · closes 2026-10-28 · cluster boc-oct-2026-decision

**REJECT — below band: YES 57¢ < 60¢ floor; also cluster runner-up to the H25 leg (same decision), and a hold-YES is a forecast of the decision's content, which is a status-quo thesis and not a process-in-motion step**

**YES condition.** For YES to pay, the Bank of Canada must leave the overnight target at 2.25% at its Oct 28, 2026 announcement, resolved by Kalshi on the BoC press release.

**Prob (for YES):** 57% pre-market → 55% final. **Book at signal:** YES 56/57, NO 43/44; 3,040 YES-bid contracts in top 3 levels; vol24 584; OI 28,938.

**Vehicle / calendar.** Same vehicle and calendar as the H25 leg (fixed-date decision Oct 28 with MPR).

- Complement of the hike leg: hold ≈ 100 − 43 (hike) − ~2 (cut/other) ≈ 55%.
- Precedent KXCBDECISIONCANADA-26SEP-H0: 'The 0bps hike leg is the hold leg and Kalshi settles it YES on an unchanged target rate; seven straight holds confirm the resolver reads Hike of 0bps literally as no change.'

**Cleanest loss tail.** Same as H25 inverted: a hike on Oct 28 after hot September CPI — the trend continuing.

**Resolver.** As H25. Deductions 0 pts.

**Precedents consulted.** KXCBDECISIONCANADA-26JUN-H0 (yes), -26JUL-H0 (yes), -26SEP-H0 (yes): hold leg settles YES on unchanged target; 'size it as yield, not as a view'.

status-quodeparture/ousterNO @ 95¢

### Kash Patel out as FBI Director before Nov 1, 2026?

KXKASHOUT-26JUL-26NOV01 · closes 2026-11-01 · cluster patel-fbi-exit

**REJECT — deep-favorite band: NO ask 95¢ (YES 5/8); max win ≤5% of risk and the 90¢+ cohort showed no edge over the market**

**YES condition.** For YES to pay, Patel must actually vacate the FBI Director role (resignation, firing, or other cessation — not a leave of absence) with a departure date before Nov 1, 2026, resolved by Kalshi per the rules_secondary departure clause.

**Prob (for YES):** 5% pre-market → 5% final. **Book at signal:** YES 5/8 (3¢ spread), NO 92/95; 1,223 NO-bid contracts in top 3 levels; vol24 1,057; OI 36,492.

**Vehicle / calendar.** No vehicle: no White House signal, no named successor, no acting designation. Sep 15, 2026 Senate Judiciary oversight hearing (~5 hours; WaPo, CBS, NBC) produced criticism but no departure step.

- Patel testified before Senate Judiciary on Sep 15, 2026 and defended hiring-policy changes and agent dismissals (Washington Post, Al Jazeera, NBC).
- Rules_secondary: 'must have an actual departure date and vacate the role within the time period'; leaves of absence and suspensions do not count.
- Price path: 4–6¢ through mid-Sep, 9¢ on Sep 22 (1,092 contracts), 8¢ today; previous_price 18¢ shows a spike that already faded.

**Cleanest loss tail.** A surprise firing by the President inside five weeks — a new discrete event, so the tail shape is acceptable; but at 95¢ the band rule rejects before the tail matters.

**Resolver.** Kalshi multi-news on actual vacancy; precedent shows strict reading. Deductions 0 pts.

**Precedents consulted.** KXKASHOUT-26APR-JUN01 (no): 'nothing short of an actual vacancy pays, and a principal who is publicly fighting back … is signalling he has not been told to go'; KXKASHOUT-26APR-JUL01 (no): far rungs 'systematically overpriced'. Both verified against the live rules_secondary.

status-quodeparture/ousterNO @ 91¢

### Cory Mills leaves the House before the Nov 3, 2026 midterms?

KXLEAVEMILLS-26NOV03 · closes 2026-11-04 · cluster mills-house-exit

**REJECT — deep-favorite band: NO ask 90.5¢ (YES 9.5/11); also thin book (156 NO-bid contracts across the top three levels) and OI under 5,000**

**YES condition.** For YES to pay, Mills must resign, be expelled, or otherwise cease to hold his House seat with an actual departure date before Nov 3, 2026, resolved by Kalshi per the rules_secondary departure clause.

**Prob (for YES):** 7% pre-market → 8% final. **Book at signal:** YES 9.5/11, NO 89/90.5; NO-bid depth 156 contracts in top 3 levels; vol24 680; OI 4,893.

**Vehicle / calendar.** Partial vehicle: H.Res. 1193 (Mace) expulsion resolution is filed and the Ethics Committee investigation is open ('serious and complex', Axios May 11, 2026), but no committee report is out and expulsion needs two-thirds on a floor vote during a pre-election October when the House is largely out.

- Mills lost the FL-7 Republican primary on Aug 18, 2026 (NPR) — his term ends Jan 3, 2027 regardless, so the incentive to resign early is weak.
- He has repeatedly said he will not resign (NBC, Apr 2026; Florida Politics).
- Price fell from 20/24 (Sep 12) to 10/16 (Sep 21) to 9.5/11 today — a drift lower with almost no volume, not a news reprice.

**Cleanest loss tail.** An Ethics report recommending expulsion plus a floor vote before the election, or a sudden resignation — new discrete events, acceptable tail shape; rejected on band, not tail.

**Resolver.** Kalshi multi-news on actual vacancy. Deductions 0 pts.

**Precedents consulted.** No rows for KXLEAVEMILLS. Archetype precedent KXLEAVEWALZ-27JAN01-TWAL-26JUL01 (no): 'resignation demands and filed impeachment articles all leave the incumbent in office — only resignation, completed removal, or death resolves Yes'. Derived partly cold.

no-vehiclecorporate-eventNO @ 92¢

### Tesla/SpaceX announce definitive merger agreement before Nov 1, 2026?

KXCOMPANYACTIONMERGER-27-26NOV01 · closes 2026-11-01 · cluster tesla-spacex-merger

**REJECT — deep-favorite band: NO ask 92¢ (YES 8/10)**

**YES condition.** For YES to pay, Tesla or SpaceX must officially announce a definitive, binding agreement to combine (press release, SEC filing, or unambiguous CEO confirmation of a signed deal) before Nov 1, 2026, resolved by Kalshi per rules_secondary.

**Prob (for YES):** 6% pre-market → 7% final. **Book at signal:** YES 8/10, NO 90/92; 1,472 NO-bid contracts in top 3 levels; vol24 2,084; OI 91,544.

**Vehicle / calendar.** No vehicle: no 8-K, no announced board process, no fairness-opinion timeline. Musk on the All-In podcast (mid-Sep 2026): 'imagine what action one might take' — an expression of intent that rules_secondary explicitly excludes.

- Rules_secondary: a Musk statement counts only as 'an unambiguous, affirmative confirmation of a definitive signed agreement — not an expression of intent, a hypothetical, a negotiation update'.
- Tesla holds an equity stake in SpaceX and a 'framework agreement' on collaboration (Yahoo Finance/Benzinga Sep 2026); a framework agreement is not a definitive combination.
- Price flat 8–12¢ for two weeks; 2,117 contracts on Sep 21 at 10¢ with no reprice.

**Cleanest loss tail.** A surprise signed merger announcement — a new discrete event, acceptable tail; rejected on band.

**Resolver.** Kalshi on official company channels; rules_secondary is unusually specific. Deductions 0 pts.

**Precedents consulted.** No rows for KXCOMPANYACTIONMERGER. Archetype rows (KXIPO-26-SPACEX, KXIPOANTHROPIC) show corporate-event resolvers read triggers mechanically. Derived cold.

calendar-impossibilitycorporate-eventNO @ 95¢

### Anthropic officially announces (confirms) IPO before Nov 1, 2026?

KXIPOANTHROPIC-DATE-26NOV01 · closes 2026-11-01 · cluster anthropic-ipo-timing

**REJECT — deep-favorite band: NO ask 95¢ (YES 5/6); the Sep 13–20 collapse from 50¢ already priced the WSJ 'staging delayed to November' report**

**YES condition.** For YES to pay, the SEC must declare Anthropic's S-1 effective, or the IPO must price, or an exchange must assign a ticker, before Nov 1, 2026, resolved by Kalshi per rules_secondary.

**Prob (for YES):** 6% pre-market → 6% final. **Book at signal:** YES 5/6, NO 94/95; 1,813 NO-bid contracts in top 3 levels; vol24 1,136; OI 94,967.

**Vehicle / calendar.** Vehicle exists but is dated past the deadline: confidential draft S-1 submitted Jun 1, 2026; WSJ (via Investing.com, ~Sep 19–20, 2026) reports IPO staging pushed to a November debut with marketing mid-October at the earliest. S-1 effectiveness typically lands the day before pricing, so a November pricing puts all three triggers after Nov 1.

- Rules_secondary: confirmed only if S-1 declared effective, IPO priced, or ticker assigned by an exchange.
- Precedent KXIPO-26-SPACEX: a proposed symbol in a public S-1 and a venue-selection headline (Bloomberg Sep 13, 2026: Anthropic picks Nasdaq) are non-events for this trigger.
- Price path: NOV01 rung 70¢ → 50¢ on Sep 13 (18.7k contracts), then to 5/6¢ by Sep 24 on the delay report.

**Cleanest loss tail.** Anthropic accelerates to an end-October pricing after all — a reversal of the reported schedule, a new event; rejected on band.

**Resolver.** Kalshi on SEC/exchange facts. Deductions 0 pts.

**Precedents consulted.** KXIPOANTHROPIC-DATE-26JUN01 (no), -26JUL01 (no): 'Officially announce an IPO' is defined mechanically … a confidential draft S-1 explicitly lacking price and share count leaves all three Kalshi triggers unmet'. Verified against live rules_secondary.

## 4. Stage-0 mechanical rejects (logged)

stage-0 cutconfirmation/enactNO @ 93¢

### Mamdani signs >0 non-emergency executive orders Sep 20–26, 2026?

KXMAMDANIEO-26SEP26-T0 · closes 2026-09-27 · cluster mamdani-eo-count

**REJECT — continuous-number market (weekly count of executive orders) that slipped past the classifier; also decided at close Sep 27, inside 7 days**

stage-0 cutmulti-news-otherNO @ 94¢

### Hegseth out as Secretary of Defense before Nov 1, 2026?

KXHEGSETHOUT-26AUG-26NOV01 · closes 2026-11-01 · cluster hegseth-dod-exit

**REJECT — close after expiration: close_time 2026-11-01T04:59Z is a month after expected_expiration_time 2026-10-01T14:00Z**

stage-0 cutmulti-news-otherNO @ 79¢

### Hegseth out as Secretary of Defense before Dec 1, 2026?

KXHEGSETHOUT-26AUG-26DEC01 · closes 2026-12-01 · cluster hegseth-dod-exit

**REJECT — inside 7d: expected_expiration_time 2026-10-01T14:00Z is under 7 days out (and vol24 356 < 500)**

stage-0 cutmulti-news-otherNO @ 89¢

### Trump publicly criticizes Kevin Warsh before Nov 1, 2026?

KXDJTATTACKWARSH-26SEP-26NOV01 · closes 2026-11-01 · cluster trump-warsh-mentions

**REJECT — Mentions market ('will X say Y'), skipped whatever the category; also vol24 173 < 500**

stage-0 cutdeparture/ousterNO @ 86¢

### Democratic senator calls for Schumer to step down before Nov 3, 2026?

KXDEMSCHUMER-27-NOV03 · closes 2026-11-03 · cluster schumer-leadership

**REJECT — vol24 0 < 500 (live); 4¢ spread on a dead book**

Liquid within-horizon markets removed by the mechanical screen (vol24 ≥ 500 but failed another cut):

| Ticker | YES bid/ask | vol24 | Cut |
|---|---|---|---|
| `KXRETIREMM-26` | 3/4 | 15,655 | yes_ask<=5 |
| `KXLOCALLAW-CAJEWISH-26OCT01` | 0/4 | 4,790 | yes_ask<=5 |
| `KXUSAIRANAGREEMENT-27-26NOV` | 3/4 | 3,913 | yes_ask<=5 |
| `KXAGICO-COMP-26Q3` | 1/2 | 2,455 | yes_ask<=5 |
| `KXUSAIRANAGREEMENT-27-26OCT` | 0/1 | 1,309 | yes_ask<=5 |
| `KXCOMPANYACTIONMERGER-27-26OCT01` | 0/1 | 1,222 | yes_ask<=5 |
| `KXCBDECISIONMEXICO-26NOV05-HOLD` | 56/86 | 900 | spread>5 |
| `KXIPOOPENAI-26NOV01` | 0/1 | 805 | yes_ask<=5 |

## 5. Reject log (scoreboard fields)

| Ticker | Side | Price | P(YES) final | Cluster | Rule |
|---|---|---|---|---|---|
| `KXCBDECISIONCANADA-26OCT-H25` | NO | 54¢ | 43 | boc-oct-2026-decision | below band |
| `KXCBDECISIONCANADA-26OCT-H0` | YES | 57¢ | 55 | boc-oct-2026-decision | below band |
| `KXKASHOUT-26JUL-26NOV01` | NO | 95¢ | 5 | patel-fbi-exit | deep-favorite band |
| `KXLEAVEMILLS-26NOV03` | NO | 91¢ | 8 | mills-house-exit | deep-favorite band |
| `KXCOMPANYACTIONMERGER-27-26NOV01` | NO | 92¢ | 7 | tesla-spacex-merger | deep-favorite band |
| `KXIPOANTHROPIC-DATE-26NOV01` | NO | 95¢ | 6 | anthropic-ipo-timing | deep-favorite band |
| `KXMAMDANIEO-26SEP26-T0` | NO | 93¢ | — | mamdani-eo-count | continuous-number market (weekly count of executive orders) that slipped past the classifier |
| `KXHEGSETHOUT-26AUG-26NOV01` | NO | 94¢ | — | hegseth-dod-exit | close after expiration |
| `KXHEGSETHOUT-26AUG-26DEC01` | NO | 79¢ | — | hegseth-dod-exit | inside 7d |
| `KXDJTATTACKWARSH-26SEP-26NOV01` | NO | 89¢ | — | trump-warsh-mentions | Mentions market ('will X say Y'), skipped whatever the category |
| `KXDEMSCHUMER-27-NOV03` | NO | 86¢ | — | schumer-leadership | vol24 0 < 500 (live) |

## 6. EV for the record (Stage-2 rejects, not deployed)

| Ticker | Side @ price | P(YES) | Edge (¢) | EV % | Why not deployed |
|---|---|---|---|---|---|
| `KXCBDECISIONCANADA-26OCT-H25` | NO @ 54¢ | 43% | +3¢ | +5.6% | below band |
| `KXCBDECISIONCANADA-26OCT-H0` | YES @ 57¢ | 55% | -2¢ | -3.5% | below band |
| `KXKASHOUT-26JUL-26NOV01` | NO @ 95¢ | 5% | +0¢ | +0.0% | deep-favorite band |
| `KXLEAVEMILLS-26NOV03` | NO @ 91¢ | 8% | +1¢ | +1.1% | deep-favorite band |
| `KXCOMPANYACTIONMERGER-27-26NOV01` | NO @ 92¢ | 7% | +1¢ | +1.1% | deep-favorite band |
| `KXIPOANTHROPIC-DATE-26NOV01` | NO @ 95¢ | 6% | -1¢ | -1.1% | deep-favorite band |

Edge is model fair value for the named side minus the price paid. The deep-favorite legs show 1–3¢ of nominal edge on 90–95¢ risk, which the record says is not distinguishable from zero; the BoC leg shows nominal edge inside a 2¢ spread against OIS parity.

## 7. Clusters

| Cluster | Members | Deployed leg |
|---|---|---|
| `boc-oct-2026-decision` | `KXCBDECISIONCANADA-26OCT-H25`, `KXCBDECISIONCANADA-26OCT-H0` | none deployed |
| `patel-fbi-exit` | `KXKASHOUT-26JUL-26NOV01` | none deployed |
| `mills-house-exit` | `KXLEAVEMILLS-26NOV03` | none deployed |
| `tesla-spacex-merger` | `KXCOMPANYACTIONMERGER-27-26NOV01` | none deployed |
| `anthropic-ipo-timing` | `KXIPOANTHROPIC-DATE-26NOV01` | none deployed |
| `mamdani-eo-count` | `KXMAMDANIEO-26SEP26-T0` | none deployed |
| `hegseth-dod-exit` | `KXHEGSETHOUT-26AUG-26NOV01`, `KXHEGSETHOUT-26AUG-26DEC01` | none deployed |
| `trump-warsh-mentions` | `KXDJTATTACKWARSH-26SEP-26NOV01` | none deployed |
| `schumer-leadership` | `KXDEMSCHUMER-27-NOV03` | none deployed |

## 8. Risk profile

- **Capital at risk:** $0. Max loss $0. Cash is the position this week.
- **Correlated exposure:** none. The only cluster with two live legs (Bank of Canada October) is a single decision; had one leg qualified, only one would have deployed.
- **What would change the answer:** the BoC NO-hike leg (KXCBDECISIONCANADA-26OCT-H25) becomes Anchor-shaped only if YES trades ≤40¢ (NO ≥60¢) _after_ the September CPI print on ~Oct 20 with trim/median still ≤2.0%. Before that print the tail is a pure continuation tail and the precedent store says there is no edge at OIS parity. Patel/Mills/Tesla-SpaceX/Anthropic NO legs would need YES to trade ≥11¢ to enter the 85–89¢ half-size lane; none has a catalyst that would push it there without also changing the thesis.

## 9. Execution notes

- No orders. Do not chase the 90¢+ NO legs: the depth is real (1,200–1,800 contracts at the top three NO-bid levels on Patel, Tesla/SpaceX, Anthropic) but the 90¢+ cohort has no edge over the market on the record and max win is ≤ 11% of risk.
- Existing live pick KXCRYPTOSTRUCTURE-26JAN-OCT (NO) is outside this run's allowlist and was not touched.
- Re-run trigger: Oct 20–21 (Canada September CPI) for the BoC ladder; otherwise the next scheduled run.

## 10. Precedent writeback

Two rows written to `/tmp/out/precedents.json` for the settled September 2, 2026 Bank of Canada rungs (KXCBDECISIONCANADA-26SEP-H0 → YES, -26SEP-H25 → NO), verified against the public API (status `finalized`) and the BoC press release. Lesson recorded: a hold with conditional two-sided guidance repriced the next ladder from carry to coin flip within three weeks while the Bank's own core trigger stayed unmet.

## 11. Method notes

Anchor rules applied in order: side rule (NO default; YES only for already-happened or process-in-motion), continuous-number thesis, continuation tail, rules-reading edge, then the entry band (60–84¢ core, 85–89¢ half size, ≥90¢ and <60¢ reject), one leg per cluster, flat 10% sizing with a 1/3-of-visible-depth cap. Probabilities were committed before opening the orderbook or price history for each candidate and are shown as pre-market → final on each card. Sources: bankofcanada.ca (Sep 2 press release, Sep 2 opening statement), StatCan Daily Sep 14, Globe and Mail Sep 18, Bloomberg/CMT Sep 21, Washington Post/NBC/CBS Sep 15 (Patel), NPR Aug 18 and NBC (Mills), Benzinga/Yahoo Finance Sep 2026 (Musk), Investing.com/WSJ Sep 2026 (Anthropic), Kalshi rules_secondary on each market.