+2,517
Peltola’s first-choice lead
106,321 votes for Mary Peltola; 103,804 for incumbent Sen. Dan S. Sullivan.
01 / The starting point
Peltola leads.
The race is unfinished.
~69,300
Estimated votes still out
FairVote’s 76%-counted assumption implies 69,266 remaining, about 24% of a projected 288,608 total.
Nov 18
Official ranked-choice rounds
If nobody wins a first-choice majority, tabulation follows the last scheduled ballot count.
| Candidate | Votes | Share including write-ins |
|---|---|---|
| Mary Peltola · Democrat | 106,321 | 48.47% |
| Dan S. Sullivan · Republican incumbent | 103,804 | 47.33% |
| Gerald Heikes · Republican | 4,865 | 2.22% |
| Daniel J. Sullivan Jr. · Independent | 3,651 | 1.66% |
| Write-ins | 701 | 0.32% |
Total: 219,342 valid first choices, including write-ins. Some dashboards show 48.6% for Peltola because they omit write-ins from their displayed denominator. This page uses the full total consistently. Statewide ballots cast, valid votes in this race, and scanned ballot records are different totals. Full candidate count ↗ · Alaska Beacon’s update ↗
What the outstanding estimate does—and does not—say. About 69,300 is a projection of additional valid Senate votes. It is not a verified count of mail envelopes waiting to be opened. “All precincts reporting” can coexist with uncounted early, absentee and questioned ballots.
What do we know about the mail backlog?
The division’s November 9 ballot statistics list 42,435 by-mail ballots issued and 30,595 received: 11,840 issued ballots were not yet returned, about 27.9% of those issued. Some may never return, arrive too late, or fail eligibility review. Received does not mean already included in results.
Across all early, absentee and questioned categories, 126,088 had been received. Alaska News estimated about 62,700 were not yet reflected in results, including roughly 20,000 early and 7,589 questioned ballots. That backlog is broader than mail, and is not a Senate-only valid-vote tally.
Peltola had won approximately 61% of counted absentee votes and 57% of early votes. Those observed shares support an optimistic scenario for her, but do not establish how every outstanding group will vote.
Official ballot-category statistics ↗ · Backlog and vote-type analysis ↗
02 / Interactive scenario calculator
New votes needed.
Final share. Different numbers.
Change the remaining vote and transfer assumptions. The model adds those votes to the reported count, then follows the rankings. Its output is a conditional calculation, not a forecast.
The default is an estimate, not a confirmed backlog.
Of all new valid votes, before transfers. The other minor candidates keep their combined 4.20% share; the rest goes to Sen. Sullivan.
Adjust the transfer assumptions
Among his ballots transferring directly to Peltola or the senator. The shares going to Heikes and becoming inactive stay fixed.
Among ballots transferring to either finalist. The observed inactive share stays fixed. This includes ballots Heikes inherited earlier.
Model order: unresolved write-ins → Daniel Jr. → Heikes. The ranking file covers a smaller set than the full first-choice count. Applying its patterns statewide is an assumption.
A 47% new-vote target is not a 47% final result. Peltola’s existing lead and transfers may allow her to win with less than half of the new first choices. Under repeated current vote shares and the default transfers, the model finishes near 50.4%–49.6%. Change the controls to test when that changes.
| Round | Peltola | Sen. Sullivan | Other continuing votes | Inactive | Peltola’s active share |
|---|
Reproduce the model and the vote targets
For each additional valid vote, the model reserves the current proportions for Heikes (4,865 / 219,342), Daniel Jr. (3,651 / 219,342) and write-ins (701 / 219,342). The Peltola slider assigns her first choices; Sen. Sullivan gets the remainder. Reported votes already counted do not change.
Default transfers use these observed transitions from the preliminary RCVis table. They are round-specific, not final-destination percentages for every original voter:
| Eliminated pool | To Peltola | To Sen. Sullivan | To Heikes | Inactive |
|---|---|---|---|---|
| 660 write-in votes | 121 | 96 | 98 | 290 |
| 3,508 Daniel Jr. votes | 1,013 | 565 | 639 | 1,291 |
| 5,101 Heikes votes | 899 | 2,199 | — | 2,003 |
The remaining 55 of the write-in pool go to Daniel Jr. The model therefore carries them into his elimination. Inactive totals accumulate across rounds. The toy example below is separate from this dataset.
The first-choice target is floor((219,342 + additional votes) / 2) + 1 − 106,321. Later targets are found by solving for the smallest Peltola share of new first choices that leaves her above half of active votes at that stage, using the selected transfers. If the current lead already suffices, the target can be zero. If the allowed new-vote allocation cannot get her there, the page says so. The calculator does not estimate the probability that a particular vote share will occur.
Rounding can display 50.00% at an extremely narrow win or loss. The model uses unrounded totals to determine the leader. A numerical tie is identified separately; Alaska resolves an actual final tie by lot.
Inspect the preliminary round table ↗ · Why the ranking file differs from the full count ↗
03 / How Alaska’s ranked-choice system works
One ballot. One active vote.
Your next available choice.
Rankings are instructions for moving a vote if a preferred candidate is eliminated. They are not extra votes counted for every candidate at once.
Check for an outright majority
A candidate with more than half of first-choice votes wins. A temporary percentage above 50% during an incomplete count does not stop eligible ballots from being counted.
Move only affected votes
If there is no first-round majority, eliminate the lowest continuing candidate. Their ballots move to each voter’s next available continuing choice. Votes already with a continuing candidate stay there.
Compare continuing votes
Alaska completes tabulation to two candidates unless someone won in the first round. The finalist with more votes wins. Ballots without a usable remaining choice are inactive.
Division of Elections: how the rounds work ↗ · Tabulation settings and continuing ballots ↗
Try a 20-voter example
This invented teaching example shows how a candidate can start with 40% and finish with a majority. It is not the actual Senate count.
Tile key: P = Peltola, S = Sen. Sullivan, H = Heikes, J = Daniel Jr., — = inactive. Each tile remains one voter, whichever candidate receives it.
Follow an individual ballot through the rounds
Daniel Jr. → Heikes → Peltola: first counts for Daniel Jr., then moves to Heikes, then to Peltola. It contributes one vote in each round.
Heikes only: counts for Heikes until he is eliminated. With no further ranking, it becomes inactive. Its earlier vote still appears in the record; it simply cannot choose between the remaining candidates.
Peltola first: stays with Peltola throughout. Lower rankings are never used while she remains in the race.
Why can the majority threshold shrink?
In the example’s last round, 19 of the 20 ballots still have a usable choice. Peltola’s 10 votes are 52.63% of 19 active votes, although they are exactly 50% of the original 20. The final comparison uses continuing votes, not a permanently fixed election-night denominator.
What about skipped rankings, overvotes and write-ins?
Alaska treats a ballot as inactive when no continuing candidate is ranked, when an overvote blocks the highest continuing ranking, or when two consecutive skipped rankings block the next choice. A single skipped rank can be passed over. These are ballot-interpretation rules, not evidence that all such voters left their second choice blank.
Unresolved write-in names are not treated as separate continuing candidates unless the division’s resolution criteria are met. The preliminary analysis groups write-ins as a pool. Its transfer step is a modeling convenience; the division’s official handling controls the real election.
04 / Why counting takes so long
The wait is mostly for ballots
and review—not arithmetic.
The ranked-choice calculation runs after the eligible ballots are assembled. Alaska’s receipt deadlines, regional counting schedule and verification work explain much of the gap between election night and the official rounds.
1. Ballots may arrive after Election Day
Eligible domestic absentee mail can arrive through November 13; overseas mail through November 18. They must have been postmarked by November 3. A ballot reaching an office later can be lawful without being a vote cast after the election.
2. Early voting has more than one count date
Regional-office early votes through the Thursday before Election Day are counted on election night. Those cast from Friday through Election Day are scheduled for the seven-day count. “Voted early” does not mean “already reported.”
3. Review precedes inclusion
Officials check eligibility and compare records before adding later early and absentee votes. Questioned ballots also need review. The November 9 reporting described ballots held while Election Day registers were cross-checked to prevent double voting.
4. Geography affects delivery
Some communities are reachable only by boat or plane. Weather and distance complicate the movement of ballots and materials. Alaska had extended post-election counts before adopting ranked-choice voting.
5. Updates come in batches
Scheduled post-election updates fall seven, ten and fifteen days after Election Day, with extra updates possible. Officials plan regional work and then publish new totals. An unchanged results screen does not mean no one is working on ballots.
6. Preliminary rankings are not the complete result
The public cast vote record lists scanned ballot rankings; hand-count-only ballots are absent. Analysts can run unofficial rounds, but those may cover fewer votes than the first-choice totals. Official RCV follows the last scheduled count on the fifteenth day.
What the public ranking file contains ↗ · Official sequence ↗
Three milestones, three meanings. A provider’s race call is its judgment that the remaining vote cannot change the winner. The official ranked-choice tabulation produces the state’s round results. Certification follows review and makes the results official. They can occur on different dates.
05 / U.S. Central and GMT+1
When to check.
What each update can tell you.
The dates below are scheduled. No exact release hour has been verified. A planning window is not an appointment or a guarantee that the race will be called.
Tuesday, November 10 • Seven-day count
First choices, plus new ranking data
A substantial early/absentee batch is expected; questioned-ballot counting begins. Updated ranking records may let analysts revise their unofficial RCV calculations.
Planning window: 5–9 p.m. CST, November 10 / 00:00–04:00 GMT+1, November 11.
Friday, November 13 • Ten-day count
Further first choices and domestic-mail deadline
Another scheduled update. This is the final receipt date for eligible domestic absentee mail; review and additional counts can continue.
Planning window: 5–9 p.m. CST, November 13 / 00:00–04:00 GMT+1, November 14.
Wednesday, November 18 • Fifteen-day count
Final regional count, then official ranked-choice rounds
Eligible overseas-mail receipt closes. Official eliminations follow the last counting of ballots. The write-in, Daniel Jr. and Heikes steps are parts of one tabulation process, not separately scheduled election days.
Watch: U.S. evening / overnight into November 19 in GMT+1. Tabulation may run later than the earlier planning windows.
Wednesday, November 25 • Target, not guarantee
Certification
The state’s target date follows review. A media call may come earlier; this page does not promise a call time.
The 5–9 p.m. CST watch windows are this page’s broad planning estimates, assuming an afternoon release in Juneau. They are not announced hours. In November, Juneau is three hours behind CST and ten behind GMT+1.
Convert an Alaska release time
Conversion example only. It does not predict when an update will appear.
06 / Do not mix odds with vote shares
A market estimate.
A vote target. A projected share.
75%
At least 50% first choices
Kalshi’s displayed probability for this specific vote-share contract. Its “at least 50%” threshold differs from the legal requirement of more than half for an outright majority.
Kalshi vote-share market ↗97.4%
Peltola wins overall
Kalshi’s candidate market, including any ranked-choice path. This is not a separate probability of winning after a particular elimination.
Kalshi candidate market ↗98.8%
Peltola wins overall
Polymarket’s displayed candidate estimate. It is a different market, not an independent confirmation that the election is over.
Polymarket ↗Frozen snapshot checked November 10, approximately 4 a.m. CST / 11 a.m. GMT+1. Prices change. These are market-implied probabilities, not official election forecasts or calls. No separately verified probability is supplied for winning immediately after Daniel Jr.’s elimination.
- 54.84% of the remaining first choices
- A calculated target with 69,266 more valid votes. It says what vote share would suffice, not how likely that share is.
- About 50.4% in the final round
- A modeled outcome if current vote shares and default transfer patterns repeat. It is not a 50.4% win probability.
- 97.4% market-implied probability
- A snapshot of traders’ pricing of the overall outcome. It is not a percentage of ballots already won.
07 / Shareable count guide
Take the explanation with you.
The static graphic preserves the November 9 count and November 10 market snapshot. The calculator above can explore other assumptions; the graphic does not change with the controls.
View the graphic
08 / Sources and boundaries
Read the evidence.
Keep the units straight.
Snapshot discipline. This page does not poll result feeds or refresh probabilities automatically. It preserves source dates, distinguishes valid votes from all ballots, and treats missing rankings as uncertainty. The forecast controls cannot establish that a candidate has won.
Calculator design: deterministic arithmetic using the displayed source snapshot and transfer assumptions. Educational example: invented 20-ballot dataset. No individual voter identity is used. All charts and outcome text update locally in your browser.