The market prices an October dip in the 7-year Treasury yield at 97%. Our path model says 69%.
Across the Treasury contracts PropBetEdge tracks, the market is more confident than the model that yields will touch levels near the month's extremes. What has to happen, what the model assumes, and how it has done on eight years of holdout data.
Original illustration © PropBetEdge Predictions (code-generated editorial art)
Outcome: 7-year par yield prints below 5.12% in Oct 2026
How low will the 7-year US Treasury yield get in Oct 2026?
5.11% or below · RESOLVED
- UPDATE · Oct 5 19:50Z PBE 62% pbe-rates-path@1.0.0 · data cutoff Oct 5 19:50Z
- UPDATE · Oct 6 19:54Z PBE 73% pbe-rates-path@1.0.0 · data cutoff Oct 6 19:54Z
This contract has resolved
Venue settlement: YES (5.11). Official check: YES · 5.11 percent · source. The analysis above is unchanged; scores for the designated snapshots are on the event record.
Quick read
- The 7-year closed at 5.17% on its latest print and has a month low of 5.12%; the contract needs a published value below 5.12% within 19 business days.
- Market 97% (bid 94 / ask 99); PBE path model 69%.
- Across 41 two-sided October Treasury quotes, the market priced higher than the model 27 times, by 15 points on average near the month's extremes.
- The model is a random walk with 5.4 bp/day volatility for the 7-year; it has no data-release calendar.
- Backtest (2018–2026, n=13,123): in the 60–90% range the model was slightly overconfident, not underconfident.
What has to happen
The contract asks whether the U.S. Treasury's published 7-year par yield prints below 5.12% on any business day from Oct. 1 to Oct. 30 — Kalshi labels it “5.11% or below.” Only the Treasury's first published daily value counts; intraday trading does not. The latest official print in the record is 5.17% (Oct 2, 2026), and the lowest print so far this month is 5.12% — one basis point short. The yield needs to close 6 basis points lower than its latest print on any one of the 19 business days left.
The market treats that as all but done: a 97% mid (bid 94%, ask 99%). PropBetEdge's path model says 69%.
The whole curve, not one strike
Kalshi lists a ladder of nested thresholds for the same month, and the model prices every one of them from the same simulation. Plotting them together shows where the two views actually part.
Levels at or above the month's low (5.12% or below, 5.13% or below, 5.14% or below) are already decided by the published path; the model does not forecast decided contracts. A wide bid–ask range is a market that has not agreed on a price, not a probability.
Data table
| Level | PBE % | Bid % | Ask % | Mid % |
|---|---|---|---|---|
| 4.85% or below | 14 | 5 | 23 | — |
| 4.86% or below | 15 | 6 | 27 | — |
| 4.87% or below | 16 | 6 | 32 | — |
| 4.88% or below | 18 | 2 | 37 | — |
| 4.89% or below | 19 | 7 | 41 | — |
| 4.9% or below | 20 | 9 | 74 | — |
| 4.91% or below | 22 | 7 | 72 | — |
| 4.92% or below | 23 | 4 | 70 | — |
| 4.93% or below | 25 | 9 | 72 | — |
| 4.94% or below | 26 | 8 | 72 | — |
| 4.95% or below | 28 | 8 | 73 | — |
| 4.96% or below | 31 | 5 | 90 | — |
| 4.97% or below | 33 | 6 | 89 | — |
| 4.98% or below | 35 | 9 | 91 | — |
| 4.99% or below | 37 | 8 | 92 | — |
| 5% or below | 39 | 9 | 91 | — |
| 5.01% or below | 41 | 9 | 91 | — |
| 5.02% or below | 43 | 26 | 91 | — |
| 5.03% or below | 46 | 29 | 91 | — |
| 5.04% or below | 48 | 25 | 90 | — |
| 5.05% or below | 51 | 29 | 91 | — |
| 5.06% or below | 54 | 30 | 91 | — |
| 5.07% or below | 56 | 28 | 92 | — |
| 5.08% or below | 60 | 65 | 99 | — |
| 5.09% or below | 63 | 76 | 83 | 80 |
| 5.1% or below | 66 | 75 | 95 | — |
| 5.11% or below | 69 | 94 | 99 | 97 |
| 5.12% or below | decided/none | — | — | — |
| 5.13% or below | decided/none | — | — | — |
| 5.14% or below | decided/none | — | — | — |
Far from the current yield, the market is mostly a spread rather than a price: bids and asks tens of points apart. The only two-sided quotes sit next to the month's low, and that is where the gap opens — 5.09% or below at 80% against the model's 63%; 5.11% or below at 97% against the model's 69%.
The same pattern across the Treasury curve
This is not a 7-year quirk. Across the 41 two-sided quotes on October Treasury paths that had a stored PBE forecast at 21:00 UTC on Oct. 3, the market priced the higher probability 27 times and the lower 11 times. On the 6 contracts within three basis points of the month's high or low so far, the market sat 15 points above the model on average; further out, the average gap was 1 point, with the market below the model on 11 of 35 quotes.
Data table
| Contract | Level | PBE % | Market mid % (bid–ask) | Market − PBE |
|---|---|---|---|---|
| 7Y how low | 5.09% or below | 63 | 80 (76–83) | 17 |
| 7Y how low | 5.11% or below | 69 | 97 (94–99) | 28 |
| 10Y how low | 4.99% or below | 17 | 11 (7–14) | -6 |
| 10Y how low | 5.21% or below | 67 | 65 (60–69) | -2 |
| 10Y how low | 5.23% or below | 73 | 89 (87–91) | 16 |
| 10Y how high | 5.3% or above | 81 | 86 (81–90) | 5 |
| 10Y how high | 5.31% or above | 78 | 87 (84–89) | 9 |
| 10Y how high | 5.33% or above | 71 | 74 (69–79) | 3 |
| 10Y how high | 5.34% or above | 68 | 70 (68–71) | 2 |
| 10Y how high | 5.35% or above | 65 | 69 (65–73) | 4 |
| 10Y how high | 5.36% or above | 62 | 70 (65–75) | 8 |
| 10Y how high | 5.37% or above | 58 | 63 (61–65) | 5 |
| 10Y how high | 5.38% or above | 56 | 57 (56–58) | 1 |
| 10Y how high | 5.39% or above | 53 | 56 (54–57) | 3 |
| 10Y how high | 5.4% or above | 50 | 48 (43–53) | -2 |
| 10Y how high | 5.41% or above | 47 | 47 (43–51) | 0 |
| 10Y how high | 5.42% or above | 45 | 48 (43–53) | 3 |
| 10Y how high | 5.43% or above | 42 | 41 (38–44) | -1 |
| 10Y how high | 5.44% or above | 39 | 36 (31–41) | -3 |
| 10Y how high | 5.45% or above | 37 | 39 (37–41) | 2 |
| 10Y how high | 5.46% or above | 35 | 27 (22–32) | -8 |
| 10Y how high | 5.47% or above | 33 | 28 (25–32) | -5 |
| 10Y how high | 5.48% or above | 31 | 27 (22–32) | -4 |
| 10Y how high | 5.49% or above | 28 | 27 (22–32) | -1 |
| 10Y how high | 5.5% or above | 26 | 27 (24–29) | 1 |
| 10Y how high | 5.51% or above | 25 | 25 (20–29) | 0 |
| 10Y how high | 5.52% or above | 23 | 25 (20–30) | 2 |
| 10Y how high | 5.53% or above | 21 | 25 (20–30) | 4 |
| 10Y how high | 5.54% or above | 20 | 25 (23–26) | 5 |
| 10Y how high | 5.55% or above | 18 | 20 (19–20) | 2 |
| 10Y how high | 5.56% or above | 17 | 18 (15–20) | 1 |
| 10Y how high | 5.57% or above | 15 | 17 (14–19) | 2 |
| 10Y how high | 5.58% or above | 14 | 12 (9–15) | -2 |
| 10Y how high | 5.59% or above | 13 | 9 (5–12) | -4 |
| 5Y how low | 4.76% or below | 19 | 19 (17–21) | 0 |
| 5Y how low | 5% or below | 73 | 88 (84–91) | 15 |
| 5Y how high | 5.11% or above | 72 | 79 (76–82) | 7 |
| 5Y how high | 5.12% or above | 69 | 70 (69–70) | 1 |
| 5Y how high | 5.13% or above | 66 | 68 (63–72) | 2 |
| 30Y how low | 5.34% or below | 10 | 19 (14–24) | 9 |
| 30Y how low | 5.59% or below | 73 | 89 (87–90) | 16 |
One reading is that the market expects more short-horizon movement than the model's volatility estimate — more volatility raises the odds of touching nearby levels in either direction. Another is a directional view the model, by construction, does not hold. The model is a random walk with no drift; it has no calendar of data releases. Neither explanation is established by one afternoon of quotes, and the market's numbers on the edges of the ladder are often one-sided.
What the model is — and how it has done
The model simulates 4,000 paths of daily yield changes for the remaining business days. Each day's move is drawn from standardized historical changes (1962–2017), scaled by an exponentially weighted volatility estimate (λ = 0.97): currently 5.4 bp per day for the 7-year. Each simulated day is rounded to the two decimals Treasury publishes, and a path counts only if a published value crosses the line — exactly as the contract settles.
It has a track record before it ever went live. On a chronological holdout from 2018 to 2026 — 13,123 contracts across the 5-, 7-, 10- and 30-year tenors that the model never saw in training — its Brier score was 0.1704 against 0.1741 for a simple Gaussian baseline. The calibration below is the relevant part for today: when it said 60–70%, the event happened 62% of the time (n=1,213); when it said 70–80%, the event happened 73% of the time (n=1,448); when it said 80–90%, the event happened 79% of the time (n=563). On that history the model was, if anything, slightly too confident in this range — not too cautious.
n = 13,123 scored forecasts.
The live record starts with these October contracts; live calibration is published only once at least 30 forecasts have resolved.
Data table
| Bin | n | Mean forecast | Observed |
|---|---|---|---|
| 0-0.1 | 2614 | 0.042 | 0.069 |
| 0.1-0.2 | 1763 | 0.148 | 0.13 |
| 0.2-0.3 | 1534 | 0.249 | 0.223 |
| 0.3-0.4 | 1436 | 0.349 | 0.331 |
| 0.4-0.5 | 1261 | 0.449 | 0.408 |
| 0.5-0.6 | 1291 | 0.551 | 0.543 |
| 0.6-0.7 | 1213 | 0.647 | 0.622 |
| 0.7-0.8 | 1448 | 0.754 | 0.733 |
| 0.8-0.9 | 563 | 0.82 | 0.79 |
What would settle it
This one can settle early: Kalshi may expire the contract as soon as a published print falls below 5.12%. Every new Treasury print becomes a new model input; a changed input creates a new immutable snapshot, and the forecasts that count for scoring were fixed by rule before any of it happened.
Evidence & method
What this story is based on. Every value above was read from these immutable records.
- Model
- pbe-rates-path@1.0.0 · RESEARCH
- Snapshot
- a8d1dbfa-04b0-4369-85c3-a787e7b97411
- Captured
- Oct 3 19:30Z
- Data cutoff
- Oct 2 23:59Z
- Market observation
- 97% · Oct 3 20:00Z
- Source families
- U.S. Department of the Treasury · PBE calibration rates-path-v1 (standardized daily changes 1962-2017)
- Resolves on
- U.S. Department of the Treasury (Daily Treasury Par Yield Curve Rates) — Daily Treasury Par Yield Curve Rate, 7-year tenor, first published value per business day
- Independent check
- FRED DGS7 (H.15 constant maturity, same Treasury curve)
- Measurement
- Daily par yield for the 7-year tenor on each business day 2026-10-01..2026-10-30; intraday values not considered
- Rounding
- Yields as published (two decimals)
- Exceptions
- The market may expire early once the threshold is met · First published value per business day governs unless the Exchange finds a material error
- Random-walk drift assumption; no macro-release calendar effects yet
- Contracts already decided by the published path are not forecast
Rates · Treasury yield paths — inputs: Official Daily Treasury Par Yield Curve (settlement source); EWMA volatility x bootstrapped historical daily changes. This model on the research board →
Market prices are a benchmark only and never enter a PropBetEdge model. Research-stage probabilities; not advice.