Key takeaways
- The source behind a commodity contract can change while its familiar name stays the same. Kalshi's weekly natural-gas series says markets settling from 4 September 2026 use Pyth's Natural Gas 24/7 index, replacing the earlier front-month futures settlement methodology. Polymarket International's selected October natural-gas contract instead references Pyth candles for an active-month ICE Henry LD1 future. “Pyth” identifies the distributor in both cases; it does not establish one underlying instrument.12
- Eight same-date observations show why the distinction matters. Across eight consecutive Fridays from 4 September through 23 October, Kalshi's reported gas fixing exceeded the EIA's Henry Hub daily spot observation by between $0.019 and $0.267 per MMBtu. These are differences between two references and observation methods, not executable price discrepancies.34
- A touch and a closing value answer different questions even when their source is identical. A gas price can cross a threshold during a week and finish below it. Contract creation time also matters: Polymarket International's selected rules count qualifying prices only after the market was created.2
- An active-month series can change because it switches delivery months. In our illustration, replacing an unchanged $2.90 contract with an unchanged $3.20 contract moves the selected reference by 10.3%. The rule selecting the month is part of the exposure, alongside the price itself.
- Getting the benchmark right still leaves the payout to match. A fixed binary payment and a volume-dependent gas bill respond differently to the same price. The useful institutional record therefore connects the commercial exposure to the benchmark, the observation rule and the payout, with each relationship reviewed separately.
1. Two contracts, one commodity name
On 30 October 2026, Kalshi finalized a weekly natural-gas contract with a reported fixing of $2.883 per million British thermal units, or MMBtu. Its question was whether the specified one-minute closing value at 5 p.m. Eastern time would be above $2.999. The result was NO. Polymarket International's contract asking whether natural gas would reach a low of $2.90 during the week of 26 October had resolved YES.52
Those outcomes are compatible. A price below $2.90 is also below $2.999. The interesting finding is what sits behind the shared natural-gas label: the contracts use different reference instruments, observe different parts of the price path and apply different thresholds. They cannot be treated as two votes on one precisely defined event.
For an energy desk, this is a familiar problem in a new wrapper. A commodity price always has a specification. It concerns a product, a location, a delivery period and a way of measuring transactions or quotations. Event contracts add another step: they turn the specified observation into a payment when a condition is met.
The result is a chain. A customer has an economic exposure; a reference price represents some part of it; a rule samples that reference; and a contract pays according to the sample. Every link can be correct on its own while the whole chain answers a different question from the one the customer began with.
Our first paper, Same Question, Different Contract, showed why settlement terms belong beside prices. Commodity markets make the next step concrete. Before comparing two event-contract prices, a reader needs to know which commodity price each contract actually transforms into an event.
2. Read through the data provider to the instrument
The current Kalshi weekly natural-gas series carries a dated notice: markets settling from 4 September 2026 use the Pyth Natural Gas 24/7 index, identified as Index.NATGAS/USD. Earlier markets retain the front-month methodology. The series ticker is unchanged.1
That is a meaningful boundary in a historical dataset. Joining all observations under “Natural Gas Weekly” would join different methodologies unless the source version were retained. A series name can be stable while the object it measures changes.
Pyth's own page classifies the current NATGAS feed as an index, with continuous coverage. It describes the feed as a proprietary Douro Labs product. That identity is distinct from an individual, dated Henry LD1 future.6
The selected Polymarket International contract names the latter. Its rules specify the active month of ICE Futures U.S. Henry LD1 Fixed Price futures, distributed through Pyth's HN feeds. ICE describes Henry LD1 as a monthly, cash-settled futures contract whose final reference is the NYMEX Henry Hub natural-gas futures settlement for the corresponding delivery period.27
This distinction explains why recording “source: Pyth” is insufficient. A distributor can publish an index, a dated future and a continuous sequence of futures. The distributor's name answers who supplies the observation; the instrument identifier answers what is being observed.
| Field | Kalshi weekly contract | Polymarket International weekly contract |
|---|---|---|
| Selected contract | KXNATGASW-26OCT3017-T2.999 | Market 6097457 |
| Underlying reference | Pyth Natural Gas 24/7 index | Pyth HN, active-month ICE Henry LD1 future |
| Test | One-minute closing value above $2.999 | Qualifying one-minute low at or below $2.90 |
| Observation period | Close at 5 p.m. EDT on 30 October | Applicable sessions in the stated week, after creation |
| Rounding | Three decimal places | Published prices without rounding |
| Reported outcome at capture | NO; fixing $2.883 | YES; resolved outcome prices 1 / 0 |
Sources: public Kalshi market and series records and Polymarket International market description, captured 31 October 2026. These contracts differ in both direction and threshold; their outcomes are not evidence of contradictory settlement.512
Commodity names also conceal differences inside futures markets. ICE's WTI crude future is cash-settled against the penultimate NYMEX settlement for the corresponding production month. Its natural-gas product uses the specified final NYMEX settlement instead. Even two ICE futures that refer onward to NYMEX have different final-reference conventions.87
An event contract can observe either future before that final settlement. A one-minute high in an active contract is an observation of trading along the way. It is not interchangeable with the reference used to settle the underlying future at expiry. The complete description needs both layers.
3. Eight Fridays, two different reference prices
A small historical comparison shows the practical scale of the distinction without pretending to estimate a universal relationship.
We took eight consecutive Fridays beginning with Kalshi's stated new-methodology boundary on 4 September and ending on 23 October 2026. For each date, we retrieved the weekly event's nested market records and verified that every returned market was finalized or settled and that the reported expiration value agreed across the event. We then paired that value with the EIA's Henry Hub daily spot observation carrying the same calendar date.34
The EIA series is a spot-price series. Kalshi's values are the event's reported index fixings at 5 p.m. Eastern time. Pairing by date makes the comparison easy to inspect; it does not synchronize the underlying observations.
| Friday in 2026 | Kalshi reported fixing | EIA Henry Hub daily spot | Fixing minus spot |
|---|---|---|---|
| 4 September | $2.939 | $2.92 | +$0.019 |
| 11 September | $2.918 | $2.71 | +$0.208 |
| 18 September | $3.032 | $2.97 | +$0.062 |
| 25 September | $3.252 | $3.21 | +$0.042 |
| 2 October | $3.039 | $2.90 | +$0.139 |
| 9 October | $3.348 | $3.10 | +$0.248 |
| 16 October | $3.375 | $3.22 | +$0.155 |
| 23 October | $3.047 | $2.78 | +$0.267 |
All prices are dollars per MMBtu. Difference is the first price minus the second. Sources: Kalshi public event records and EIA's daily series, release dated 28 October 2026. This is a descriptive comparison of distinct references and sampling methods, with no assumption that either was an executable quotation.34
The largest difference in these eight rows was 26.7 cents per MMBtu on 23 October. The same day's values also sat on opposite sides of $3: the event fixing was $3.047 and the published spot observation was $2.78. A test based on one series would therefore have answered a different $3 threshold question from a test based on the other.
That is the institutional implication. A reference can be a sound measure of its own defined object and still be the wrong input for another question. Showing an energy-market event price beside a physical benchmark is useful precisely when the display keeps that distinction visible.
The term basis needs equal care. In its conventional cash-versus-futures use, it describes the difference between a physical price and a futures price, usually cash minus futures. Regional gas markets also use it for the difference between a local price and Henry Hub. Our exhibit reports an index-fixing-minus-spot difference; it is not a reconstructed, tradable cash-futures basis.910
4. A price path is different from its last point
Source identity is only the first layer. Once two contracts refer to the same price, their observation rules still have to match.
A closing-value contract asks where a reference finishes. A touch contract asks whether the path reaches a threshold during an eligible window. To isolate that distinction, consider two hypothetical contracts on one common reference and the same week: one pays if the final value is above $3; the other pays if any eligible observation is above $3. The examples avoid equality at the threshold.
| Illustrative path | Highest eligible price | Final price | Finish above $3 | Touch above $3 |
|---|---|---|---|---|
| Rises, then retreats | $3.20 | $2.90 | NO | YES |
| Rises and holds | $3.20 | $3.10 | YES | YES |
| Stays below | $2.95 | $2.90 | NO | NO |
Author calculations on hypothetical paths. Both tests use identical observations, source, window and strict threshold. These are not reconstructed paths for the October contracts.
Under those assumptions, finishing above the threshold implies having touched above it. The reverse does not follow. A higher price for the touch claim can therefore reflect the larger set of qualifying paths. Interpreting that difference as disagreement about the final price would mix two events.
Market creation introduces another boundary. The selected Polymarket International rules count only qualifying observations after creation. The event record for the week of 26 October contains three separate $3.00 low markets created on 23, 26 and 28 October. Each displayed a YES resolution when captured. Their shared title and result do not erase their different start times.11
For a historical analysis, those timestamps determine which observations were eligible. For a user comparing a newly listed claim with an older claim, they determine whether an earlier threshold crossing counts. A calendar-week label alone cannot supply the answer.
The same logic applies to an average. A utility whose invoice is linked to a month's average price has an exposure to all the observations entering that average and to the volume purchased against them. A single Friday close is a different statistic. Replacing the average with the close changes the question before any probability is calculated.
5. The month-selection rule can move the reference
Futures prices are attached to delivery periods. A front- or active-month series is a rule for choosing among those periods as time passes. It eventually switches from one contract to another.
Polymarket International's selected natural-gas rules make that selection explicit: the next listed future becomes active at the opening of the third-to-last trading session of the nearest contract. ICE separately defines the Henry LD1 last trading day as three business days before the contract period begins.27
The distinction between a fixed delivery month and a changing active month matters even if nothing moves within either contract. Consider two hypothetical adjacent futures. The earlier one trades at $2.90 and the later one at $3.20. At the observation boundary, the selected series switches contracts and the later contract records an eligible observation at its unchanged price.
| Observation | Earlier future | Later future | Selected reference | Above $3? |
|---|---|---|---|---|
| Before the switch | $2.90 | $3.20 | $2.90 | NO |
| After the switch | $2.90 | $3.20 | $3.20 | YES |
Illustrative arithmetic: $3.20 − $2.90 = $0.30; $0.30 / $2.90 = 10.3%. Neither individual future changed price. The example assumes an unadjusted contract-selection series and a qualifying observation after the switch.
This is an effect of selecting a different delivery period, not a 10.3% return on one unchanged futures position. Whether a real event contract recognizes that crossing depends on its rules, the selected feed, adjustment conventions and eligible observations.
The operational record therefore needs the actual month at each observation, not just a label such as “front month.” The selection rule and its calendar should travel with the data. The same applies to a historical chart: a series joined across contracts and a fixed-month series answer different questions about price movement.
EIA explains futures in terms of standardized obligations for future delivery, including a specified location and period. That time dimension is central to the instrument. Removing it from the label removes part of the exposure.12
6. The customer's bill adds location, volume and payout shape
An energy customer usually experiences a delivered cost. Its price may include a regional hub, transportation and an invoice convention; its quantity may change with demand. Henry Hub is an important reference within that system. It is not every customer's delivery point.
EIA documents substantial price variation across North American gas hubs, associated with geography, weather, supplies and infrastructure constraints. Its examples include Waha's production and pipeline conditions and SoCal Citygate's transportation into the Los Angeles distribution system.13
A binary event contract adds a further distinction. It pays a fixed amount when its condition holds. A customer's extra gas cost can keep increasing as price or volume increases.
To separate these effects, suppose an illustrative customer buys 100,000 MMBtu and compares its delivered price with a $3 budget. Suppose also that 100,000 hypothetical binary claims each pay $1 if a specified reference ends above $3. The table compares gross amounts at the same horizon. It excludes premiums, fees, financing and taxes, so the final column is a difference between gross amounts, not profit or the all-in cost of protection.
| Reference price | Delivered price | Extra physical cost above $3 budget | Gross binary payout | Extra cost less payout |
|---|---|---|---|---|
| $2.90 | $3.50 | $50,000 | $0 | $50,000 |
| $3.10 | $3.10 | $10,000 | $100,000 | −$90,000 |
| $4.00 | $4.00 | $100,000 | $100,000 | $0 |
| $4.00 | $5.00 | $200,000 | $100,000 | $100,000 |
Hypothetical arithmetic. Extra physical cost = 100,000 × max(delivered price − $3, 0). Gross payout = $100,000 when the reference exceeds $3 and zero otherwise. A negative last-column value means the gross payout exceeds the modeled overrun; it is not a net trading return.
The first and fourth rows show a difference between the reference and the customer's delivered price. The second and third isolate a separate issue: even when the reference equals the delivered price, a fixed payment matches the extra bill at only particular prices. Raising the quantity changes the relationship again.
This is also where event contracts can contribute something useful. A threshold claim describes a specific contingency in a simple form. It can make a discontinuous exposure legible—for example, an explicitly modeled cost that begins once a defined condition is met. The work is to connect the claim to the actual cost structure, rather than treating the commodity name as that connection.
The same mathematics explains why a full range of compatible thresholds contains more information than one threshold. A rising cost above a budget level can be decomposed into increments associated with successively higher price levels. That interpretation requires the increments to share their reference, horizon, units and settlement rules. A collection of superficially similar contracts cannot supply it automatically.
7. A record that starts with the exposure
Our earlier paper, The Missing Layer, proposed a record separating an event from the venue contracts referring to it. For commodity applications, a second relationship is needed: how that defined event relates to the commercial exposure being examined.
These are different review questions. Two venue contracts can be equivalent to each other while both describe a price different from the customer's invoice. Conversely, two distinct contracts can illuminate separate parts of one exposure without being interchangeable.
| Record | Minimum useful content | Question it answers |
|---|---|---|
| Commercial exposure | Product, location, quantity, delivery period, invoice formula | What changes the customer's cost or revenue? |
| Reference instrument | Administrator, feed, venue, instrument, delivery month or month-selection rule | Which price is being observed? |
| Observation rule | Eligible sessions, start time, fixing, average or touch, rounding, fallback | Which values count toward the event? |
| Contract payoff | Threshold, direction, payment amount, settlement process and terms version | What does the claim pay in each relevant state? |
Proposed research record. The fields organize the distinctions established by the source comparison and worked examples; they are not a statement that a particular platform automates every review.
A useful comparison can then state its conclusion precisely: same commodity but different reference; same reference but different observation window; same event but different payment; or a reviewed match on the relevant dimensions. The reader learns what can be combined and what must remain separate.
That is the practical relevance to Predict Here. Its public methodology separates venue contracts from canonical events and gives semantic relationships explicit review states. Commodity research extends the same discipline upstream to the price reference and downstream to the exposure. The purpose is to preserve the economic meaning through the comparison.14
8. Our take
The benchmark belongs in the headline description. “Natural gas above $3” is a topic. A defined instrument, observation rule and horizon make it a usable proposition. The index-versus-future distinction in the October contracts should be visible before a reader compares their prices.
Historical continuity must be earned. Kalshi's stated September source boundary means the weekly ticker alone is insufficient to describe a continuous methodology. Active-month contracts add recurring boundaries of their own. Version and month-selection information belong beside the observations.
A price comparison is only one part of an exposure comparison. The eight Friday observations show how different references can sit on different sides of a threshold. The budget example shows why even an exact reference match does not establish an exact payout match. Institutional use requires both questions to be answered.
Event contracts can make commodity contingencies easier to express and compare. Their usefulness increases when the reference price and the economic exposure remain visible all the way through to settlement.
Methodology and limitations
Primary contract records. We captured public Kalshi and Polymarket International API records on 31 October 2026. The comparison uses the exact instruments and rules identified in the notes. We inspected the current series notice; we did not establish when it was first published or reconstruct every historical version. Polymarket International outcomes are reported from the venue's resolved API state, not from an independent oracle adjudication. Polymarket US is outside this sample.
Eight-Friday comparison. The window begins at the notice's 4 September effective boundary. We used every Friday through 23 October; the EIA release available at capture contained no 30 October spot observation. Across the eight events, all 349 returned Kalshi market records were finalized or settled, with one common nonempty expiration value per event. These 349 contracts supply eight unique fixing observations, not 349 independent price observations. We transcribed the eight EIA Friday cells from the retained daily table and checked them against its HTML. Subtraction uses decimal arithmetic.
The comparison measures differences between a 5 p.m. index fixing and a daily spot observation. It does not isolate which component—instrument construction, timing or another feature—caused each difference. We did not recover the proprietary NATGAS index formula, synchronize source ticks, estimate a hedge ratio, test predictive performance or establish executable relative value. The eight positive differences are a description of this window, not evidence that the relationship is permanently positive.
Illustrations. Exhibits 3–5 are authored examples, separate from the recorded market observations. Exhibit 4 assumes an unadjusted switch and an eligible post-switch observation; it does not reconstruct the selected contracts' actual roll path. Exhibit 5 fixes volume and horizon and omits all acquisition and carrying costs. It evaluates gross payout shape, not suitability for a particular customer. Source and contract rights must be checked for any proposed commercial use or redistribution.
Reproducibility. Retained API and reference-page captures, SHA-256 source hashes, the eight-date extraction and a Python calculation script were kept with this research. They support these bounded comparisons.
Notes and references
- Kalshi Trade API v2, KXNATGASW series, captured 31 October 2026 at 09:36:25 UTC. Fields:
product_metadata.important_info.markdownandsettlement_sources. The notice states a 4 September 2026 effective boundary. Its presence at capture and the series update timestamp do not establish the first publication time. See also COMMODITIES terms, which provide the parameterized source and price-type framework. ↩ - Polymarket International Gamma API, market 6097457, and the week-of-26-October event, captured 31 October 2026. The market description identifies the active-month ICE Henry LD1 reference, eligible sessions, creation boundary, no-rounding convention, month switch and fallback.
outcomes,outcomePrices,closedandumaResolutionStatusjointly identify the reported YES resolution. ↩ - Kalshi Trade API v2 event records with nested markets, captured 31 October 2026: 4 September, 11 September, 18 September, 25 September, 2 October, 9 October, 16 October, 23 October. Values are the common
expiration_valueacross each event's returned market records. Date pairing and differences are the authors' calculations. ↩ - U.S. Energy Information Administration, Henry Hub Natural Gas Spot Price, daily, release dated 28 October 2026, captured 31 October 2026. Friday cells in the eight weeks identified in Exhibit 2; units dollars per million Btu. Values are the source's current published historical observations, not recovered first-release vintages. ↩
- Kalshi Trade API v2, KXNATGASW-26OCT3017-T2.999, captured 31 October 2026. Fields:
rules_primary,rules_secondary,close_time,expiration_value,result,status,settlement_ts. The fixing is the preceding one-minute candle's close at 21:00 UTC / 5 p.m. EDT; finalization is recorded at 22:00:43 UTC. The figure $2.883 is the underlying expiration value, not the event-contract price. ↩ - Pyth Terminal, Natural Gas 24/7 index, read 31 October 2026. Feed metadata identifies
Commodities.Index.NATGAS/USD, instrument type Index and continuous coverage; the accompanying notice identifies the proprietary Douro Labs data product. No index-composition formula is inferred from the name. ↩ - ICE Futures U.S., Henry LD1 Fixed Price Future specifications, read 31 October 2026. Relevant sections: Description, Last Trading Day and Reference Price A, including its pricing date and specified settlement price. ↩
- ICE Futures Europe, WTI Crude Futures specifications, read 31 October 2026. Settlement section specifies the penultimate NYMEX settlement for the production month. ↩
- CFTC, Futures Glossary, entries “Basis” and “Basis Risk,” read 31 October 2026. Used for terminology, not a legal conclusion about any product. ↩
- EIA, Natural gas price differentials to Henry Hub narrowed at most hubs in first half of 2020, 8 September 2020. Used for the regional-hub meaning of basis, not evidence about current price levels. ↩
- Polymarket International Gamma API, event query for the week of 26 October 2026, captured 31 October 2026. Selected $3.00 LOW records: 6097456, created 23 October at 22:01:48 UTC; 6256613, created 26 October at 16:39:23 UTC; 6349508, created 28 October at 16:39:25 UTC. Each was closed and resolved with YES / NO outcome prices 1 / 0 at capture. The distinct creation times are observed metadata; no individual qualifying price path is reconstructed here. ↩
- EIA, What is the natural gas futures market?, 29 July 2024. Used for the future-delivery, location and physical-versus-financial settlement distinctions. ↩
- EIA, Market dynamics vary at key natural gas pricing hubs, 23 October 2024. Used for regional-price mechanisms and the named hub examples, not for October 2026 observations. ↩
- Predict Here, Methodology and Governance, read 31 October 2026. The proposed exposure record in this paper is a research recommendation; it does not establish automated coverage of every commodity or customer exposure. ↩
Disclosures
This paper is general research published for information only. It is not investment, legal or tax advice, and it does not recommend any position, venue or strategy. References to other organizations describe their published work and do not imply an endorsement of, or relationship with, Predict Here. Predict Here is operated by Odyssea Research LLC. It does not list markets, hold client funds, take the other side of trades, or execute or route orders.
Persons associated with Predict Here may hold positions in event contracts, including on venues discussed in this report.