How we verify election information
Strata-AF™ Elections keeps one verified record for each contest. The calendar, dossiers, updates and results all read from that record, and every fact in it carries its own source and the time it was last checked.
What each label means
| Official | Published by the legally responsible body. |
| Verified | Corroborated by two independent sources. |
| Expected | Inferred from the constitutional timetable. |
| Disputed | Credible sources conflict. |
| Claimed | Asserted by an interested party, not verified. |
| Provisional | Announced, but not yet legally final. |
| Final | Certified by the legally competent body. |
Verified means two independent sources. Two outlets repeating the same wire story count as one.
Source hierarchy
| Tier | Source | Examples | Used for |
|---|---|---|---|
| 1 | National authority | Electoral commission, constitutional court, official gazette, parliament, presidency | Authoritative for dates, eligibility, candidate lists and certified results |
| 2 | Continental and regional bodies | African Union, ECOWAS, SADC, EAC, IGAD observation missions | Process assessments and observation findings |
| 3 | Specialist datasets | EISA, IFES ElectionGuide, International IDEA, ACE Electoral Knowledge Network | Discovery, history and electoral-system data |
| 4 | Independent reporting | Credible national media, civil-society monitors, international reporting | Context and dispute flags; never silently overrides tier 1 |
| 5 | Interested parties | Candidate posts, party statements, eyewitness material | Treated as claims until independently verified |
Election listing sites help us discover what is coming. Nothing is published from them automatically; an editor confirms each fact against a primary source.
Candidates' and parties' own channels
We also read the official websites, channels and X profiles of candidates, their running mates and parties, but only ones two editors have checked are genuine. A running mate's posts are filed under their ticket. What they publish is the campaign's own account, so we treat it as a tier 5 claim:
- It can appear on the campaign trail, labelled as the campaign's claim, and only after an editor approves it. STRATA-AF™ Intelligence (sAI) never publishes it on its own.
- It never moves the projection on its own. A campaign's post about a rally, defection or endorsement is held as an early lead. If an independent outlet reports the same event within seven days, the outlet's report becomes the evidence and the post is credited as the first report. If no outlet does, the lead expires. The database refuses to accept a claim as evidence.
- It never counts as a second source when we check whether something was independently reported.
- If a candidate or party deletes a post on X, we remove its text within a week. A published entry keeps our own summary, with a note that the original post was deleted.
How often we re-check
| Stage | Re-verified |
|---|---|
| Expected | Every 30 days |
| Announced, campaign | Every 7 days; every 3 days in the final month |
| Voting, counting | Every 4–6 hours |
| Contested | Every 24 hours |
| Certified, transition | Every 14 days |
When a dossier passes its check window, it shows a Check overdue label until an editor confirms the record again. We show this on purpose. Cards on the home page and the tracker always carry the date of the last check.
Two editors for anything final
Any change that makes a value official or final, names a winner, or changes a result figure needs a second editor’s approval. The database enforces this rule; no single editor can bypass it.
What we do not do
- We do not give elections a “free and fair” score. We document indicators, such as candidate exclusions, internet disruptions and observer access, and we attribute every assessment.
- We do not publish unofficial tallies. Party and candidate totals appear only as attributed claims. Projections are kept apart from results and appear only on the Projection tab.
- We do not run a minute-by-minute live blog. During voting we publish a slower, verified update stream.
The Projection tab
For selected contests (Nigeria 2027, Cabo Verde 2026, The Gambia 2026, Kenya 2027 and Angola 2027), a Projection tab sits between Campaign and Vote & count. Each is built from that country's last two presidential elections, region by region, and uses that country's own winning rule. It is kept apart from results and is always shown with its uncertainty.
- The starting record. Every region's results in the last two presidential elections (states in Nigeria, counties in Kenya, provinces in Angola, administrative areas in The Gambia, electoral circles in Cabo Verde), with registration and turnout, each figure linked to the electoral authority's declaration or, where that was not available, the best published compilation, which the tab names.
- Nigeria: the 2027 tickets. Since 4 November 2026, Nigeria's starting record is re-cut for who is running in 2027. Each party's 2023 vote is split, state by state, across the 2027 candidates its voters are expected to follow: most of the APC vote stays with Tinubu, with some going to Atiku in the North and to Obi in the cities; about three-quarters of the northern PDP vote stays with Atiku, and elsewhere the PDP vote splits with Tinubu where PDP governors joined the APC, and with Makinde in the South West; Kwankwaso's NNPP vote mostly follows him to Obi, his running mate; and Labour's vote mostly stays with Obi. The splits are editorial judgements drawn from defections, primaries and polls, not measurements. Before 4 November each candidate started from their own party's 2023 result (“2023 party lines”); the published history was rebuilt from the new start so it reads as one series. Running mates count too, once: where a ticket changed its running mate since 2023, the new mate's home state gains a small pull (half a candidate's own home-state pull) and the old mate's loses it. Atiku's running mate moved from Okowa (Delta) to Amaechi (Rivers); Obi's from Baba-Ahmed (Kaduna) to Kwankwaso, whose Kano vote is already carried by the NNPP split above, so it is not added again; Tinubu kept Shettima. Our tests on past elections found this pull too small to measure, so it moves the numbers only a little.
- Win chance. “Win” means winning outright in the first round. Where the rule has a run-off, a candidate with a low outright chance can still reach the run-off; the tab shows run-off chances and pairings separately.
- Evidence. Editors review tagged evidence, such as defections, rallies, endorsements, observer reports and economic shocks. Each piece is recorded as one of six forces (strength, conversion, exasperation, reassurance, resilience, apathy) for a named candidate in a named state, with a source. A campaign's own channel is never the source: what a candidate or party says about itself is a claim.
- The Run. The projection is computed every morning, and on demand when an editor asks. Each run builds on the last published run and applies every piece of evidence accepted since that run closed. The tab says when it was updated (“Projection · updated Mon 2 Nov”); a run with no new evidence says so and carries the numbers forward unchanged.
- Caps. No single piece of evidence, and no single candidate, can move a state by more than a set amount in any 7 days. The cap is a 7-day budget, not a calendar week: a run on Tuesday still counts what the runs since last Tuesday spent.
- Rallies. A rally, town hall or march is sized by the crowd outlets report, not by its format: medium when they report a large crowd, small when they report none, never larger than medium. A report of a candidate's strength mostly protects the support they have. A large rally also brings out some people who would otherwise stay home, and a large rally reported by two or more outlets brings out twice as many, within the same caps. Hover or tap any run to see the reports that moved it.
- Where a piece applies. A report counts in the state it is about. A national report counts a third as much in every state, or only in the regions it concerns: a northern figure's move, for example, counts in the North-Central, North-East and North-West states and nowhere else.
- The books must balance. Every point a candidate loses in a state goes to a named rival or to voters who will not vote. If a state does not balance, that run is held.
- Where support moved. What moved since the last run, where, and why, with the evidence behind each move. It makes no win claim.
- The forecast. Every run simulates the election 10,000 times under the country's real rule. Nigeria: the most votes and at least 25% in 25 of the 37 jurisdictions (we also test the reading 24 of 36 states and the FCT). Kenya: more than half the vote and at least 25% in 24 of the 47 counties. Cabo Verde: more than half the valid votes. The Gambia: the most votes, one round. Angola: the party list with the most votes. We show each chance with an 80% range, the chance of a run-off where the rule has one, and each condition separately.
- Faith, community and home zone in Nigeria (since 12 November 2026). A candidate's pull in a state now changes with how much of that state shares their faith, their community or their zone, compared with the candidate whose votes they inherit there. Only the change counts, so what the last election already shows is not counted twice. This brings the home-zone adjustment back for Nigeria, with two more, on new evidence: we tested them on 2015–2019 and 2019–2023, scoring each election with weights chosen on the other, and the projected shares came closer to the real result in both. They made the run-off a little more likely. Each state's make-up comes from Afrobarometer surveys of 7,199 Nigerians between 2014 and 2024. We project states, never groups of people. Every earlier run was recomputed the same way.
- Two starting points for Nigeria (since 12 November 2026). Atiku Abubakar left the PDP after 2023, and how much of that party's 2023 vote goes with him is the least certain part of Nigeria's starting point. Our main start keeps a little over half of it with him, a judgement from who has defected and where. The 2023 House of Representatives vote, cast the same day, points lower: he ran behind his party across the South, and only slightly ahead of it in the North. Neither reading can be tested on a past election. So three simulations in four now start from the main reading and one in four from a reading halfway to the House figures, and the published chances carry that doubt. This made an outright Tinubu win a little more likely and Atiku a little weaker. Every earlier run was recomputed the same way.
- Room for the rare big miss (since 12 November 2026). Elections sometimes move much further than usual: a candidate collapses, or a new party breaks through. Our simulations used to treat such moves as almost impossible. They now give them a small but real chance, in every country, while the usual size of a move is unchanged. We tested this on fourteen past elections in eleven countries, scoring each one with settings chosen without it: the projections were a little closer overall, and much less often badly surprised by who won. The published chances moved by one to three points. Every earlier run was recomputed the same way.
- How we check the model. Every test on a past election is also scored against simple guesses: each region voting exactly as last time, and the same swing everywhere. A model that cannot beat them adds nothing, and we say so on the contest's record. Before any 2027 result, we committed to scoring every published Nigeria run against those guesses and against the shared settings with no campaign evidence at all.
- The run-off, played out. Where the rule has a run-off, each simulation that needs one plays it: the other candidates' voters split between the two finalists by each finalist's strength in their region, with a random national tilt for how losing camps move. In Nigeria the second finalist is the candidate with a majority in the most states, as section 134(4) of the Constitution sets out. This adds who wins a run-off; it never changes the first-round chances.
- Ballot what-ifs. When a court case puts a candidate's place on the ballot in doubt, the headline keeps every candidate INEC lists until a final ruling, and each run also simulates the race without them, shown beside the headline and marked as a what-if. Their support in each state is shared among the others in proportion to their support there, and turnout stays the same. Nigeria: from 5 to 17 November we showed the race without Atiku, whose ADC nomination a Federal High Court voided. It ended on 18 November, when INEC's final list of candidates named him; the ADC's appeal is still pending, and the what-if returns if it fails.
- Tested and left out. We tested a blend of starting points (the ticket splits as set, half as strong, and half as strong again) and turnout that rises and falls with each candidate's supporters. On the 2019–2023 test neither improved the projection, so neither is used. Both stay in the model, ready if a later test shows they help. The two starting points for Nigeria described above are a different blend, built on a measurement rather than on stronger and weaker versions of one judgement.
- Polls. Graded polls can pull the simulated national vote shares towards their average, but in Nigeria this is switched off. We tested it on 2023 with the four best-graded polls published before the vote: at every strength it called fewer states and made the vote-share error larger, because those polls overstated Peter Obi by about 20 points. Polls still count towards the confidence label, and we show every poll we hold with its grade.
- When we hold back. If uncertainty is too high, a state is shown as too uncertain to call, and the national headline is withheld. We name a likely winner only when one candidate wins at least 60% of simulations.
- Change, two ways. The tab shows the change since the last run and the change on a week ago, and the win chance run by run.
- Our own readers do not count. Strata-AF™ readers' weekly ballot is shown as readers, not voters, and never moves the forecast.
- Before launch. No forecast is published until the model has been tested against the country's two previous elections (run from the earlier one, scored against the later one) and an admin has signed off that test. The result of each test is shown on the tab. Until a contest passes, its forecast is published with an “Experimental model” label, its backtest score beside the headline and Low confidence. A failed test usually means the two elections tested were a realignment (Kenya 2017 to 2022, when Uhuru Kenyatta backed Raila Odinga) or had no like-for-like pair (Cabo Verde 2016, with no PAICV candidate), which no model can call from no evidence. Once an admin signs off a failing test as experimental, sAI can publish those runs under the same guardrails, still labelled Experimental and Low. Either way, each run is published either by an editor or by sAI when every guardrail passes (see Run by sAI below).
The record behind each projection
How sure each projection is, how it did against past elections, every correction, and where the past results come from. Open a contest to see it.
Nigeria 2027Confidence: Low20 corrections · 37 sources
How sure is this?
- Track recordPasses a normal election (called 31 of 37 states in 2019 (from 2015 data)), but not yet one where new parties broke through (called 24 of 37 states in 2023 (from 2019 data)).
- Fresh evidenceAbout 7 sourced reports a week in the latest run.
- PollsNo published national poll in the last 90 days.
- RangeThe leader’s likely vote share spans 29 points (23–52%).
Tested against past elections
The latest run · Fri 20 Nov
How we will be scored · committed 6 Nov 2026
Scores appear here once the result is declared: every published run, scored on its own against INEC’s declared state-by-state presidential results.
- ↓Brier scoreHow close each run’s win chances came to what happened, outright winner and run-off included. A coin toss between two outcomes scores 0.50.
- ↓Log scoreHow surprised each run was by the result. Saying it could not happen scores 6.9.
- ↑Units calledUnits where our projected leader really led. Our bar on past elections: 80%.
- ↓Average missAverage gap between each candidate’s projected and real share in a unit. Our bar on past elections: 6 points.
- ◎Ranges that heldHow often the real share fell inside our 80% range. 80% is honest; 70–90% passes.
Corrections · 20
- 20 Nov 2026Clarified
Manual intake 20 Nov 2026: Soludo-led Awka rally for Tinubu
- 19 Nov 2026Clarified
Manual intake 19 Nov 2026: Kano convoy attack (Obi resilience), Edo PANN endorsement (Tinubu conversion)
- 18 Nov 2026Clarified
Rallies are now sized by the crowd outlets report, not by format, for every candidate: Obi's Calabar and Lagos town halls count as medium; five Tinubu meetings and visits filed above that rule were brought down. We recomputed the history with the same evidence.
- 18 Nov 2026Clarified
Evidence filed late while our news reader was paused (17 and 18 November) is now in: Tinubu's Birnin Kebbi and Akure rallies, Wike's Rivers campaign council for Tinubu and Atiku's Bauchi tour, each dated to when it happened. We recomputed the history with the same evidence.
- 16 Nov 2026Clarified
Large rallies reported by two or more outlets now add support, and Soludo's appointment as Tinubu's campaign manager counts as large, not governor-sized. We recomputed the history with the same evidence.
- 15 Nov 2026Clarified
Yunusa Tanko's move to the SDP now counts in the northern states only, as he was Obi's adviser on Northern Nigeria. We recomputed the history with the same evidence; the runs held on 14 and 15 November are now published.
- 12 Nov 2026Clarified
In a third method change today, Tinubu's chance of winning outright moved from 42% to 45% and the run-off chance from 49% to 47%. This is not campaign news: every projection now allows a small chance for a rare, very large move.
- 12 Nov 2026Clarified
We changed how every projection allows for a rare, very large move: such moves used to be treated as almost impossible and now have a small chance (see Methodology). The usual size of a move is unchanged. We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 12 Nov 2026Clarified
In a second method change today, Tinubu's chance of winning outright moved from 40% to 42% and the run-off chance from 51% to 49%. This is not campaign news: the projection now allows for doubt about how much of the PDP's 2023 vote follows Atiku Abubakar out of the party.
- 12 Nov 2026Clarified
We changed how Nigeria's projection starts: how much of the PDP's 2023 vote goes with Atiku Abubakar, who has left the party, is uncertain, so three simulations in four now use our main reading and one in four a lower reading based on the 2023 House of Representatives vote (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 12 Nov 2026Clarified
Nigeria's run-off chance rose from 43% to 51% today because we changed our method, not because of campaign news. A candidate's pull in a state now also reflects how much of that state shares their faith, community or zone.
- 12 Nov 2026Clarified
We changed how Nigeria's projection is set: a candidate's pull in a state now also reflects how much of that state shares their faith, community or zone, compared with the candidate whose votes they inherit there (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 10 Nov 2026Clarified
We changed how the projection is set: one set of settings now applies in every country, chosen by testing on past elections in five countries (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced. The ranges are wider than before.
- 7 Nov 2026Clarified
Running-mate backfill: Obi-Kwankwaso Lagos town hall/markets (6 Nov) and Atiku-Amaechi Birnin Kebbi rally (5 Nov) added from outlet reports.
- 6 Nov 2026Clarified
We recomputed the whole projection history with the current model, which since 6 November includes mobilisation (a big, checked show of strength can bring out voters who stayed home, weighted by voter-card holders and party strength). Same starting point, same accepted evidence; earlier figures have been replaced.
- 6 Nov 2026Clarified
Why Tinubu's chance of winning outright fell from 74 to 53 in 100: no new evidence caused it; the model changed on 5 November. Each state is now weighted by the share of registered voters who collected their voter cards (PVCs) in 2023 (34 states; Cross River, Delta and Ondo to follow), and votes a candidate inherits from another party are simulated with wider ranges. Projected vote shares barely moved (Tinubu 36%, Obi 31%, Atiku 21%), but more simulated elections now end in a run-off (20 in 100, from 11) or with Obi winning outright (26, from 14).
- 5 Nov 2026Clarified
We updated the projection model on 5 November 2026 and recomputed the whole history with the same evidence. Votes a candidate inherits from another party are now simulated with wider ranges, because that is where past projections were least certain, and a candidate gains a small pull in states of their own home zone. Earlier figures have been replaced; the methodology page explains both changes.
- 5 Nov 2026Clarified
From 4 November the projection starts from the 2027 tickets, counting each running-mate change once, and the published history was rebuilt on that basis. It now weights states by INEC's preliminary 2027 register (103.4m; 13 states published, the other 24 estimated from 2023 and scaled to INEC's total; final register due 15 December 2026). From the next run, each state is also weighted by the share of voters who collected their voter cards (PVCs) in 2023. These are changes to the model's inputs, not new evidence.
- 4 Nov 2026Clarified
We recomputed the projection to add each candidate's projected share of the national vote. The evidence and the chances are unchanged.
- 4 Nov 2026Clarified
We recomputed the projection to add each candidate's projected share of the national vote. The evidence and the chances are unchanged.
About this data
- 2023 state figures were corrected against INEC; the national APC total from state sums is 10,929 above INEC’s declared figure, unexplained.
Sources for past results · 37
- 2023channelstv.com12345678910111213
- 2023icirnigeria.org123456
- 2023kolaking.substack.com123
- 2023ripplesnigeria.com12
- 2023thecable.ng12
- 2023thisdaylive.com12
- 202321stcenturychronicle.com
- 2023blueprint.ng
- 2023dailytrust.com
- 2023fmino.gov.ng
- 2023newtelegraphng.com
- 2023thesun.ng
- 2023thewillnews.com
- 2023withinnigeria.com
- 2019dailytrust.com
Cabo Verde 2026Confidence: Low5 corrections · 2 sources
How sure is this?
- Track recordFails its test on a past election: called 2 of 10 electoral circles in 2021 (from 2016 data).
- Fresh evidenceNo new sourced reports in the latest run; the numbers were carried forward.
- PollsNo published national poll in the last 90 days.
- RangeThe leader’s likely vote share spans 39 points (30–69%).
Tested against past elections
The latest run · Fri 20 Nov
Corrections · 5
- 15 Nov 2026Clarified
Reports on Joana Gomes Rosa Amado's campaign now count towards her line. Until 15 November they could not reach the model because her line was set up as a party bloc. We recomputed the whole history with the evidence on file, so earlier figures have been replaced.
- 15 Nov 2026Clarified
Joana Gomes Rosa Amado is the MpD's candidate. Until 15 November our projection showed the MpD's 2021 vote as an unnamed "MpD bloc" and listed her under Others, so reports on her campaign were left out of the model. We have corrected this and recomputed the whole history with the same evidence, now counted towards her, so earlier figures have been replaced.
- 12 Nov 2026Clarified
Neves's chance of winning outright moved from 50% to 51% today, the MpD side's from 33% to 30%, and the run-off chance from 17% to 19%. This is a method change made in every country on 12 November, not a reaction to the campaign: every projection now allows a small chance for a rare, very large move.
- 12 Nov 2026Clarified
We changed how every projection allows for a rare, very large move: such moves used to be treated as almost impossible and now have a small chance (see Methodology). This is a change of method made in every country on 12 November, not a reaction to the campaign in Cabo Verde. We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 10 Nov 2026Clarified
We changed how the projection is set: one set of settings now applies in every country, chosen by testing on past elections in five countries (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced. The ranges are wider than before.
About this data
- The official results (Boletim Oficial) are published by electoral circle, not by municipality: five single-municipality circles, five multi-municipality circles and three diaspora circles. The map shows circles.
- Diaspora circles count towards the national total but are not drawn on the map.
- Joana Gomes Rosa Amado (MpD) starts from the MpD-backed 2021 vote (Carlos Veiga). Corrected on 15 Nov 2026: earlier runs showed this line as an unnamed “MpD bloc” and listed her in Others.
- Antero Tavares and Loisito Sousa have no past vote to start from; they sit in Others until evidence moves them.
Sources for past results · 2
The Gambia 2026Confidence: Low3 corrections · 2 sources
How sure is this?
- Track recordFails its test on a past election: called 7 of 7 administrative areas in 2021 (from 2016 data).
- Fresh evidenceNo new sourced reports in the latest run; the numbers were carried forward.
- PollsNo published national poll in the last 90 days.
- RangeThe leader’s likely vote share spans 35 points (33–67%).
Tested against past elections
The latest run · Fri 20 Nov
Corrections · 3
- 12 Nov 2026Clarified
Barrow's chance of winning moved from 80% to 82% today, and Darboe's from 20% to 18%. This is not campaign news: every projection now allows a small chance for a rare, very large move.
- 12 Nov 2026Clarified
We changed how every projection allows for a rare, very large move: such moves used to be treated as almost impossible and now have a small chance (see Methodology). The usual size of a move is unchanged. We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 10 Nov 2026Clarified
We changed how the projection is set: one set of settings now applies in every country, chosen by testing on past elections in five countries (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced. The ranges are wider than before.
About this data
- The IEC reports seven areas; Kuntaur and Janjanbureh are reported together as Janjanbureh (Central River Region) and are drawn together on the map.
- The IEC’s 2016 sheet has candidate votes 96 higher than votes cast nationally (Kanifing and Brikama rows); figures are kept as published.
- Bensouda (UNITE) has no past vote of his own to start from; he sits at zero until evidence moves him. Kandeh’s GDC vote (12% in 2021) starts in Others.
Sources for past results · 2
Kenya 2027Confidence: Low3 corrections · 2 sources
How sure is this?
- Track recordFails its test on a past election: called 39 of 47 counties in 2022 (from 2017 data).
- Fresh evidenceNo new sourced reports in the latest run; the numbers were carried forward.
- PollsNo published national poll in the last 90 days.
- RangeThe leader’s likely vote share spans 27 points (36–64%).
Tested against past elections
The latest run · Fri 20 Nov
Corrections · 3
- 12 Nov 2026Clarified
Kenya's projection barely moved today (Ruto's side 51%, the opposition 45%, a run-off 4%) after a method change, not campaign news: every projection now allows a small chance for a rare, very large move.
- 12 Nov 2026Clarified
We changed how every projection allows for a rare, very large move: such moves used to be treated as almost impossible and now have a small chance (see Methodology). The usual size of a move is unchanged. We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 10 Nov 2026Clarified
We changed how the projection is set: one set of settings now applies in every country, chosen by testing on past elections in five countries (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced. The ranges are wider than before.
About this data
- The United Opposition has pledged one candidate but not named them. Until it does, the model holds two blocs: Ruto on his 2022 county vote, and the declared opposition aspirants (Gachagua, Kalonzo, Matiang’i, Karua, Wamalwa) together on Azimio’s 2022 county vote; evidence about any of them counts towards the opposition bloc.
- 2017 uses the 8 August results (later annulled), the best measure of 2017 county voting. IEBC’s printed 2017 county rows do not sum to its national totals (Odinga −25,294; Kisumu row out by 60,516); figures are kept as printed.
- Diaspora votes count nationally but not towards the 24-county rule.
Sources for past results · 2
- 2022IEBC Kenya
- 2017IEBC Kenya
Angola 2027Confidence: Low3 corrections · 2 sources
How sure is this?
- Track recordFails its test on a past election: called 17 of 18 provinces in 2022 (from 2017 data).
- Fresh evidenceNo new sourced reports in the latest run; the numbers were carried forward.
- PollsNo published national poll in the last 90 days.
- RangeThe leader’s likely vote share spans 36 points (34–70%).
Tested against past elections
The latest run · Thu 19 Nov
Corrections · 3
- 12 Nov 2026Clarified
The MPLA's chance of the most votes moved from 59% to 60% today, and UNITA's from 41% to 40%. This is not campaign news: every projection now allows a small chance for a rare, very large move.
- 12 Nov 2026Clarified
We changed how every projection allows for a rare, very large move: such moves used to be treated as almost impossible and now have a small chance (see Methodology). The usual size of a move is unchanged. We recomputed the whole history with the same evidence, so earlier figures have been replaced.
- 10 Nov 2026Clarified
We changed how the projection is set: one set of settings now applies in every country, chosen by testing on past elections in five countries (see Methodology). We recomputed the whole history with the same evidence, so earlier figures have been replaced. The ranges are wider than before.
About this data
- 2022 provincial figures are the CNE provisional count at 97% (the final provincial breakdown could not be retrieved). The final national totals are higher by MPLA 46,628 and UNITA 28,901.
- For Luanda, Bengo and Cabinda in 2022 only MPLA and UNITA are available; 2022 registered voters by province come from a compilation not checked against a CNE source.
- Angola split into 21 provinces in 2024; results are shown on the 18 provinces used in 2017 and 2022.
- The 2027 MPLA head of list is not yet known (João Lourenço is term-limited); UNITA’s is Adalberto Costa Júnior, re-elected party leader in November 2025.
Sources for past results · 2
Run by STRATA-AF™ Intelligence (sAI)
STRATA-AF™ Intelligence (sAI) is the intelligence behind Strata-AF™ Elections. It covers every presidential contest we follow across Africa: it reads the news, files campaign events, checks each one against its source, and, where a contest has a projection model, turns verified evidence into each morning's run: the movement since the run before, and a fresh forecast. Strata-AF™ editors set the rules it works to and audit its work.
- How sAI checks a claim. It re-reads the source page and asks whether the page reports the claim. It must copy the supporting sentence, and that sentence must really be in the page. It checks that the candidate, the place and the date match, looks for other publishers reporting the same thing, and looks for duplicates already counted. Large claims, such as a governor defecting, need independent reporting.
- What sAI may do on its own. Each contest has its own level, from checking only (editors decide everything) to publishing routine campaign events and the daily run. Defections, endorsements, controversies and anything sAI cannot verify always go to an editor.
- Guardrails. A run is held for an editor if it fails any check: the numbers must balance, enough evidence must be behind it, no single story may drive most of the movement, and no state may swing implausibly. Publishing the forecast also needs a signed-off backtest: signed off as passed, or as experimental, which keeps the Experimental model label and Low confidence.
- Earned trust. sAI publishes nothing on its own until editors have reviewed enough of its calls, over enough weeks, and agreed with almost all of them.
- Who published it. Every run says whether sAI or an editor published it. Corrections and retractions are listed on the Projection tab.
Corrections
No fact is edited in place. Every change is a new, sourced entry, and the previous value stays visible in the corrections log. To report an error, email legal@strata-af.com with the page and the source.