8Factor Report #8: The Senate Seat Market Finally Adds Up. That's the Problem.
A month ago, the Senate control market didn't match its own race prices. Now the seat-count market matches Alaska and Mississippi almost to the point. Kalshi prices 56+ Democratic seats at about 67%. That number is only as good as the two uncalled races under it, and we think both are rich.
*Predictagon 8Factor Report · No. 8 · November 4, 2026 · Read-only analysis. Not financial advice.*
The market at a glance
Kalshi, "How many Senate seats will Democrats hold after the 2026 election?":
| Contract | Bid | Ask | Last | Contracts traded (24h) |
|---|---|---|---|---|
| At least 54 | 99¢ | 100¢ | 99¢ | ~99K |
| At least 55 | 93¢ | 96¢ | 97¢ | ~119K |
| At least 56 | 62¢ | 73¢ | 64¢ | ~150K |
| At least 57 | 2¢ | 4¢ | 5¢ | ~65K |
| At least 58 | 0¢ | 1¢ | 1¢ | ~73K |
The 8Factor scorecard
| Factor | What it asks | Grade |
|---|---|---|
| 1 · Compare | Do the seat prices match the race prices? | C |
| 2 · Fair value | What's the real probability? | C+ |
| 3 · Net edge | Is anything tradable after costs? | B− |
| 4 · Timing | When does it close? | B |
| Red team | Inherited risk · Correlation · Seat definitions · Resolution date | 4 live risks |
Verdict: internally consistent, but built on two rich legs. Unlike Report #3's control market, this ladder agrees with its own race markets: 56+ seats at about 67% is almost exactly Alaska (~85%) × Mississippi (~80%). The disagreement is ours. We price those races at 68% and 62% (Reports #5 and #6), which puts 56+ seats at about 42%.
First, the scorecard on Report #3
On October 5, Report #3 argued that a 66% Democratic Senate control price was about 8 points rich against the race-by-race markets, with a fair value of 58%. Democrats won control. With 54 seats called, they cleared the 51 they needed with room to spare.
We were wrong on the direction. The race markets were too pessimistic, not the control market. Texas, Ohio, Iowa, Kansas, Maine and Michigan all went Democratic. Red 2 in that report ("waves make races move together") was the risk that came true. We said a high generic-ballot lead could mean race prices were lagging the wave, and they were.
What we take forward: in a wave year, the sum-of-races approach understates correlated outcomes. Now that every race but two is called, correlation hardly matters, and the sum-of-races check is the right tool again.
Why this market, why now
As of NBC's 5:58 AM ET update, Democrats have 54 seats and Republicans 44, with two races uncalled: Alaska and Mississippi. Every rung of the seat ladder now depends only on those two races:
| Outcome | Democratic seats |
|---|---|
| Republicans win both | 54 |
| Democrats win one | 55 |
| Democrats win both | 56 |
| Anything above 56 | requires a called race to flip |
The race markets (Kalshi / Polymarket mids):
| Race | Democratic price |
|---|---|
| Alaska (Peltola) | 86.5¢ / 84.45¢ → ~85.5% |
| Mississippi (Colom) | 80.5¢ / 79.85¢ → ~80% |
Factor 1 · Compare: Grade C
Question: does the seat ladder match the race prices?
Yes, closely. The two counts are separate tabulations in separate states, so treat them as roughly independent:
| Rung | Implied by race markets | Seat market (mid) | Gap |
|---|---|---|---|
| 55+ | 1 − (0.145 × 0.20) = 97.1% | 94.5% | −2.6 |
| 56+ | 0.855 × 0.80 = 68.4% | 67.5% | −0.9 |
| 57+ | ~1% (a called race flips) | 3% | +2 |
The ladder is within about two or three points of its components on every rung. Only the 57+ tail looks a touch rich, and at 2/4¢ there's nothing to do with it after fees. No internal edge. The Grade C is because Kalshi is the only venue with this ladder in our pull, so there's no cross-venue comparison.
Factor 2 · Fair value: Grade C+
Question: what's the real probability Democrats hold 56+ seats?
> AI fair value (Democrats ≥56 seats): 30% low · 42% mid · 60% high. Confidence: C+.
This is the product of our two race calls:
56+ = 0.68 × 0.62 ≈ 42%. And for the other rungs:
| Rung | Our fair value | Market | Divergence |
|---|---|---|---|
| 55+ | 1 − (0.32 × 0.38) = 88% | 94.5% | −6.5 |
| 56+ | 42% | 67.5% | −25.5 |
Why it isn't lower: both race markets may be right. If the outstanding ballots are as favorable to Democrats as traders seem to think, the market's 68% stands. Our band's upper end (60%) allows for that.
Why the two legs are close to independent: Alaska and Mississippi count different ballots under different rules. The only shared factor is that both markets may be leaning on the same late-vote assumption. That's a reason to trust both prices less, not a reason the outcomes move together.
Factor 3 · Net edge: Grade B−
Net Edge = divergence × log10(volume + 1), using the 56+ rung:
The trade is selling the 56+ rung (buying NO at 1 − 62¢ bid = 38¢) against a fair NO of 58%:
| Step | Cents |
|---|---|
| Buy "At least 56" NO on Kalshi | 38.0 |
| Fair value of NO | 58.0 |
| Gross edge | +20.0 |
| Kalshi fee (≈ 0.07 × 0.38 × 0.62) | −1.6 |
| Predictagon buffer | −1.0 |
| Net edge | ≈ +17¢ per contract, on our fair value |
That's large on paper and entirely a function of our race calls. If you don't buy Reports #5 and #6, this trade disappears. The book is also wide (62/73¢) and moved about 150,000 contracts in 24 hours, enough for modest size but not much more. B−.
Simpler expressions of the same view: buying Sullivan in Alaska or Hyde-Smith in Mississippi directly gives deeper books. The seat ladder is a combined bet that pays only if both hold.
Factor 4 · Timing: Grade B
Two to three weeks if both counts are clean, possibly much longer if not: B.
The Red Team: four ways this call breaks
Red 1 · Both race markets are right
This ladder inherits every risk in Reports #5 and #6. If the uncounted ballots in rural Alaska and in Hinds County and the Delta are as Democratic as traders think, 68% is fair and our sell is a loss. This is the most likely way our analysis is wrong.
Watch: the next count updates in both states. If Colom's lead grows past a few hundred votes and Peltola closes the first-choice gap, both of our calls are in trouble.
Red 2 · We just underrated correlation once
Report #3 lost on correlation. Here we argue the two counts are close to independent. If something common drives both late counts, such as Democratic-leaning mail arriving late everywhere, our independence assumption understates the chance both go Democratic.
Watch: whether late mail shifts toward Democrats in other states still counting.
Red 3 · A called race isn't final
The 57+ rung prices about a 3% chance that something beyond Alaska and Mississippi changes. A canvass reversal in a called race is rare, but this ladder pays on certified totals, not network projections.
Watch: canvass totals in the closest called races.
Red 4 · The plumbing: what "hold" means and when
The contract counts seats Democrats hold after the election. Independents who caucus with Democrats, vacancies, and the resolution date (certification vs. swearing-in) all matter at the margin.
Watch: Kalshi's exact resolution text on independents and on the resolution date.
Bottom line
| Internal consistency | Good: ladder matches race prices within ~3 pts |
| Our view | 56+ seats rich: ~42% fair vs ~67.5% market |
| Derived from | Report #5 (Alaska 68%) × Report #6 (Mississippi 62%) |
| Clock | Nov 18 (AK tabulation) to late November; longer if litigated |
| Biggest risk | Red 1: both race markets are right about the late vote |
Report #3 was wrong because race markets lagged a national wave. That wave is over and the seats are counted. What's left are two counting fights, and the market is pricing Democrats to win both at roughly 2 to 1. We think it's closer to a coin flip that leans the other way.
New 8Factor Reports publish every week at Predictagon.com/8Factor.
Data notes and methodology
Sources: NBC News: Senate results · CBS News: Senate results live · Kalshi: Senate control · Predictagon 8Factor Report #3