Kalshi Economics — Mispricing Audit
2026-07-15 · v3.3 methodology · $2,500 capital · 45-day horizon · category-match mode (Economics)
Headline result: an unusually efficient week. Of 967 active Economics markets, 74 passed the liquidity screens, 16 got full diligence, and exactly one pick cleared every gate: BUY NO on July CPI >0.2% at 75¢ (MEDIUM conviction, $175 deployed, 7% of capital). The two largest modeled edges of the run — FOMC zero-dissents (+18¢) and the post-Hormuz gas ladder — were both rejected by the entry screens, not missed: the first is a coin-flip-band trade without a mechanical lock, the second repriced 70¢ in 48 hours and now trades above my model. 93% cash is the position.
1. How this was researched
Mode: category-match. trading_events.category = 'Economics' matched 967 active markets closing within 45 days, so no theme-keyword fallback was needed.
- Stage 1 filter (read-only DB mirror): latest snapshot per market, cut to
volume24h ≥ 1500, bid-ask spread ≤ 5¢, not fully priced (YES bid < 95¢, YES ask > 5¢) → 74 markets. Removed announcer-noise series (KXFEDTWEETS, KXWEFTWEETS), redundant deep-priced ladder legs, and the already-published pick KXFEDDECISION-26JUL-H0 (plus its fully-priced sibling legs). 16 distinct candidates advanced.
- Stage 2 diligence: full rules text from the mirror; 14-day intraday history from
market_snapshots; live top-of-book from the Kalshi public API (quotes in this report were captured 2026-07-15 ~16:00 UTC); news diligence via four parallel research passes (gas/AAA, inflation/nowcasts, labor, Fed & GDP) grounded in primary sources — BLS releases, Cleveland Fed nowcast, Atlanta Fed GDPNow, Federal Reserve statements and speeches, AAA/GasBuddy, DOL claims data.
- Stage 2.5 resolver profiling: every candidate's resolution source and mechanism named; resolver risks priced as explicit point deductions. One market (Chipotle burrito index) was rejected outright because its rules name no resolution source at all.
- Stage 3–3.5 screens: conviction tier set by defensibility (never by EV); effective entry classified into bands — 60–90¢ favorites deployable, 35–60¢ coin-flips only for HIGH conviction with ≥15¢ edge, <35¢ tails hard-rejected (v3.2), LOW conviction never deploys (v3.3).
- Stage 4 sizing: conviction-weighted, EV-agnostic. HIGH ≤ $375/pick (15%), MEDIUM ≤ $175/pick (7%), ≤ $375 (15%) per thesis cluster.
2. Markets reviewed
| Market | Ticker | Quote (Y bid/ask) | Verdict |
|---|
| July CPI MoM >0.2% | KXCPI-26JUL-T0.2 | 26 / 29 | PICK — BUY NO @ 75¢ (MEDIUM) |
| FOMC July: exactly 0 dissents | KXFOMCDISSENTCOUNT-26JUL-0 | 50 / 51 | Reject — coin-flip band, not HIGH (edge +18¢ logged) |
| Fed holds AND 0 dissents | KXFEDCOMBO-26JUL-0-0 | 45 / 46 | Reject — same cluster/band; cheaper twin of the above |
| Q2 GDP advance >2.0% | KXGDP-26JUL30-T2.0 | 47 / 48 | Reject — nowcasts split 1.3% vs 2.7% around strike |
| Jobless claims ≥210k (wk Jul 11) | KXJOBLESSCLAIMS-26JUL16-210000 | 57 / 59 | Reject — coin-flip band; repriced on retooling story |
| July payrolls >100k | KXPAYROLLS-26JUL-T100000 | 38 / 39 | Reject — weak-labor story already priced |
| June core PCE >0.2% | KXPCECORE-26JUN-T0.2 | 12 / 13 | Reject — fairly priced post-CPI/PPI |
| July CPI MoM >0.1% | KXCPI-26JUL-T0.1 | 43 / 46 | Reject — coin-flip band; T0.2 strike is the better expression |
| AAA gas >$4.00 on Jul 31 | KXAAAGASM-26JUL31-4.00 | 86 / 89 | Reject — repriced 13¢→89¢ in a week; ask > model |
| AAA gas >$4.10 on Jul 31 | KXAAAGASM-26JUL31-4.10 | 56 / 62 | Reject — no edge; 6¢ spread |
| AAA gas >$4.20 on Jul 31 | KXAAAGASM-26JUL31-4.20 | 34 / 35 | Reject — YES is a tail; NO edge ~2¢ with full war tail |
| AAA gas >$4.00 on Jul 20 | KXAAAGASW-26JUL20-4.000 | 61 / 62 | Reject — needs exactly the current pace; coin flip |
| AAA gas >$3.925 on Jul 16 | KXAAAGASD-26JUL16-3.925 | 54 / 59 | Reject — strike at the median of one overnight print |
| AAA gas >$3.930 on Jul 16 | KXAAAGASD-26JUL16-3.930 | 25 / 28 | Reject — NO costs 75¢ vs model 68 |
| Chipotle burrito avg ≥$9.78 (July) | KXCHIPBURRITO-26AUG02-T9.78 | 23 / 24 | Reject — rules name no resolution source (Stage 2.5) |
| Fed decision July (hold 0bp) | KXFEDDECISION-26JUL-H0 | 95 / 96 | Excluded — already an active published pick |
3. The pick
Pick 1 — KXCPI-26JUL-T0.2 · BUY NO @ 0.75 · MEDIUM CONVICTION
Will CPI rise more than 0.2% in July 2026? (BLS release Aug 12, 2026)
Current price
YES 26 / 29 → NO effective 74¢ (top 10 lots) / 75¢ (2,000 lots)
My probability (NO)
81% vs market-implied ~74–75% → +6¢ edge at the 75¢ limit
Conviction / band
MEDIUM · 60–90¢ favorite band (effective entry 75¢)
Size
233 contracts @ 75¢ = $174.75 (7.0% of capital — MEDIUM cap)
Cluster
july-cpi-print ($175 = 7.0% < 15% cap)
Liquidity at entry
vol24h 3,615 · OI 13,570 · 2,010 contracts fillable ≤ 75¢ · spread 3¢
Price history
Whipsawed 8¢ → 78¢ → 27¢ over Jul 7–15 (16.8k cum. volume); the collapse came on the Jul 14 CPI release.
Resolver profile (Stage 2.5). Resolver: BLS CPI-U headline, seasonally-adjusted MoM, single decimal, from the July 2026 CPI news release (Aug 12, 8:30 ET); "more than 0.2%" means a printed 0.3 or higher, i.e. ≥0.25 unrounded. Mechanism: snapshot of the first print. Priced risks: −2 pts (rules say "the Consumer Price Index" without specifying SA — the series has consistently resolved on the SA headline, and the July NSA factor is slightly negative so both readings point the same way); −1 pt (release-timing/shutdown risk — funding runs through Oct 1, so minimal). Evidence-based P(NO) 84 − 3 = 81.
Mispricing thesis. The market is paying 26–29¢ for a hot July print in a month whose own best nowcast is roughly zero. The crowd sees $3.89 gas going to $4.15+ and prices energy passthrough; the rules pay on the month-over-month change in the monthly average, and July's average pump price will almost certainly sit below June's, because June averaged ~$4.05 on the way down from the May 21 peak ($4.56) while July troughed at $3.796 on Jul 8. The headline-vs-arithmetic gap is the trade.
Evidence.
- Cleveland Fed inflation nowcast (updated 7/15, post-CPI and post-PPI): July headline CPI −0.01% MoM, July core CPI +0.21%.
- June CPI (BLS, released Jul 14): headline −0.4% MoM (largest drop since Apr 2020), core 0.0%, shelter +0.1% (smallest since Jan 2021). June PPI (Jul 15): −0.3% headline, core ex-trade +0.1%. Momentum in core is soft, not hot.
- Gas arithmetic: AAA month-ago (Jun 15) $4.065 vs $3.890 today. Even at a sustained +2.5–3¢/day climb to ~$4.35 by Jul 31, July's monthly average (~$3.97–4.00) still lands 1–2% below June's — a flat-to-negative SA gasoline contribution. A positive gas contribution needs AAA near $4.60 by late July.
- Scenario math: a 0.3+ headline print effectively requires core ≥0.3 (vs 0.21 nowcast and 0.0 last month) and energy to stop dragging. Weighted across core scenarios, P(YES) ≈ 15–20% vs 26–29% priced.
Tail risks (cleanest way to lose first).
- Extreme Hormuz escalation pushes AAA toward $4.50–4.60 by late July and core re-accelerates to 0.3+ — the combination that prints 0.3. This is the single cleanest loss path; it needs both legs.
- Core surprise alone: tariff/chip passthrough (memory +17.4% YoY, Apple price hikes) re-firms core goods; June's 0.0 core was itself a big downside surprise and could mean-revert hard.
- The market moved on this thesis in the last 48h (78¢ → 27¢ post-CPI). My claim is that the residual 27% is still too high given the monthly-average arithmetic — the remaining edge is real but smaller, which is why this is MEDIUM, not HIGH.
Numbers (for the record — did NOT drive sizing). Edge +6¢/contract · EV +8% on cost · max payout $233.00 · expected value ≈ +$14 · worst case −$174.75.
4. Recommended $2,500 portfolio
| # | Ticker | Action | Limit | Contracts | Cost | % cap | Conviction | Band | Max payout | EV¢ | EV% |
|---|
| 1 | KXCPI-26JUL-T0.2 | BUY NO | 75¢ | 233 | $174.75 | 7.0% | MEDIUM | 60–90¢ favorite | $233.00 | +6 | +8% |
| Cash reserve | | $2,325.25 | 93.0% | Cash is a position — the category is efficient this week. |
Totals: deployed $174.75 (7.0%) · cash $2,325.25 (93.0%) · max payout $233.00 · blended EV +8% on deployed cost (≈ +$14 expected) · EV figures recorded for calibration only — sizing was set by conviction tier and entry band per v3.
Cluster exposure
| Cluster | Cost | % of capital | Cap | Status |
|---|
| july-cpi-print | $174.75 | 7.0% | 15% ($375) | OK |
| aaa-gas-hormuz-spike | $0 | 0% | 15% | All 6 legs rejected |
| fomc-july-dissents | $0 | 0% | 15% | Both legs rejected (band screen) |
Correlation note: the pick is mildly positively exposed to a Hormuz ceasefire (softer gas → softer CPI), the same direction as several already-published Hormuz NO picks in the live feed. It shares no cluster with them, and at $175 it cannot flip the run's sign.
Conviction exposure
| Tier | Picks | Cost deployed | % of capital |
|---|
| HIGH | 0 | $0 | 0% |
| MEDIUM | 1 | $174.75 | 7.0% |
| LOW | 0 (never deploys, v3.3) | $0 | 0% |
Risk profile
- Worst case: July CPI prints 0.3+ → −$174.75 (−7.0% of capital). No cluster can do worse; 93% is cash.
- Best case: CPI prints ≤0.2 → +$58.25 (+2.3% of capital).
- Most likely: Cleveland's ~0.0% nowcast holds within noise; the pick wins (~81%).
- Concentration: single thesis, single resolver (BLS), single date (Aug 12). By construction the book cannot lose more than 7%.
Execution notes
- Rest a NO limit at 75¢ — 2,000+ contracts sit at that level (YES bid 25), so the fill is passive with ≤1¢ effective slippage from mid. Do not chase below 73¢-equivalent; at NO ≥ 78¢ the edge is gone.
- Invalidation triggers: (1) AAA national average ≥ $4.35 before Jul 24 (puts a positive July gas contribution in play) — exit or halve; (2) Cleveland July headline nowcast rises above +0.15% on any weekly update — exit; (3) a hot July core-goods datapoint (PPI passthrough news) — reassess.
- Watchlist: KXFOMCDISSENTCOUNT-26JUL-0 YES — if it drifts to ≤45¢ with no new dissent signals, the edge exceeds 20¢ and a fresh audit should re-examine it; also watch KXGDP-26JUL30-T2.0 after the Jul 16 GDPNow update, and the gas weekly ladder for post-ceasefire overreaction (the NO side becomes interesting only after a confirmed de-escalation headline, not before).
- No hedges needed at this size. Hold the cash; a 93%-cash week is the screen working, not a failure to find trades.
5. What I rejected and why
All 14 rejects below are machine-logged in picks.json and shadow-tracked to settlement — if they outperform the picks, the screen adds nothing and we want to know.
Cut by the Stage-3.5 entry-band screen (the discipline on display)
- KXFOMCDISSENTCOUNT-26JUL-0 YES @ 51¢ (model 69%) — the run's biggest modeled edge, +18¢, rejected. The evidence is strong: June's meeting under new Chair Warsh was 12–0; serial dissenter Miran resigned May 14; Waller's own hike condition ("a hot core reading this week") failed when June core CPI printed 0.0; the hawks' April complaint (easing-bias language) was already fixed in the June statement. But 51¢ is the coin-flip band, which deploys only HIGH conviction with ≥15¢ edge — and a future 12-person vote three weeks out, with Bowman (dovish, predicted 3 cuts in 2026) and Hammack (publicly pro-hike) both plausible lone dissenters, is not "mechanically locked in." This is precisely the trade type that bled the v1/v2 book. Shadow-track it.
- KXFEDCOMBO-26JUL-0-0 YES @ 46¢ (model 66%) — same thesis wearing a different ticker (hold is 95%+ priced, so the combo ≈ the dissent market minus 5¢). Same band reject. Structural note for future runs: the combo is the cheaper expression of an identical claim.
- KXJOBLESSCLAIMS-26JUL16-210000 YES @ 59¢ (model 66%) — six straight prints ≥215k and a 215k consensus argue YES, but the price already fell 79→57 in 48h on the one genuine risk: Ford canceled its F-Series summer shutdowns and GM added Flint production days, so the seasonal factor expects an NSA retooling spike that won't fully arrive, biasing the SA print down. Coin-flip band, edge <15¢, thesis already moving the price.
- KXGDP-26JUL30-T2.0 NO @ 53¢ (model 56%) — GDPNow 1.3% vs NY Fed Nowcast 2.7% vs SPF 2.1%: the models straddle the strike almost symmetrically, and GDPNow updates Jul 16 (tomorrow). No forecastable edge; textbook coin-flip reject.
- KXCPI-26JUL-T0.1 NO @ 57¢ (model 62%) — same thesis as the pick but from inside the coin-flip band with less rounding margin. The T0.2 strike is the correct expression.
- KXAAAGASD-26JUL16-3.925 YES @ 59¢ (model ~50%) and KXAAAGASW-26JUL20-4.000 YES @ 62¢ (model 53%) — both strikes sit at the median of their forecast window ($3.89 today +2.5–3¢/day; De Haan's "$4.00 in 7–10 days" straddles Jul 20 exactly). Coin flips priced as favorites.
Cut for no edge after the 48-hour repricing rule
- KXAAAGASM-26JUL31-4.00 YES @ 89¢ (model 80%) — the week's biggest story (Hormuz ceasefire collapse Jul 11–13, naval blockade, Brent +11%, RBOB +13.6%/month) moved this leg 13¢→89¢ in seven sessions. At 89¢ the ask now exceeds my model: mediators (Qatar/Oman/Pakistan) are active, and a ceasefire before ~Jul 22 leaves retail peaking near $4.00–4.05 and feathering back below the strike by Jul 31. The market didn't underreact — if anything it's now ahead of the fundamentals. NO is a sub-35¢ tail (hard reject band). No trade exists here.
- KXAAAGASM-26JUL31-4.10 YES @ 62¢ (model 57%) — pure ceasefire-timing bet, 6¢ spread, no edge.
- KXAAAGASM-26JUL31-4.20 NO @ 66¢ (model 68%) — +2¢ of edge against the full war tail (sustained blockade at +2.5¢/day reaches $4.29 by Jul 31). YES @ 35¢ is the tail band. Pass both ways.
- KXAAAGASD-26JUL16-3.930 NO @ 75¢ (model 68%) — negative edge, and the whole market is one overnight AAA update. Noise trade.
- KXPCECORE-26JUN-T0.2 NO @ 88¢ (model 88%) — crashed 68¢→13¢ on the Jul 14 CPI (core 0.0) and Jul 15 PPI (core ex-trade +0.1); Cleveland's post-data nowcast of 0.19% puts YES at ~12%, exactly where it trades. The market did my analysis first. Worth noting for calibration: 2026's base rate was 5-for-5 months above 0.2 before this.
- KXPAYROLLS-26JUL-T100000 NO @ 62¢ (model 58%) — June's +57k shock, negative revisions, and the shrinking labor force (−1.1M since January) all argue NO, but the market already moved there, and resolution is on the first print, where 2026 first prints (148k/172k/57k) skewed high before revisions. No edge left.
Cut at Stage 2.5 (resolver risk)
- KXCHIPBURRITO-26AUG02-T9.78 — the full rules text is one sentence: "If the average U.S. price of a Chipotle Chicken Burrito for July 2026 is at least $9.78..." — according to what? No source, no page, no feed is designated. An unpriceable resolver is an automatic reject regardless of the price.
6. Sources