Nigeria’s tightest governorship votes again, with a paper trail we already hold

Ademola AdelekeAdemola Adeleke (A) won this seat in 2022 as the PDPPDP candidate. In November 2025 he left a fracturing PDP and took the ruling machinery of the state (the Speaker, two dozen state lawmakers, six federal Reps) into the small Accord party, while endorsing President Tinubu for 2027. The APCAPC, fresh off a heavy win in neighbouring Ekiti, runs Bola OyebamijiBola Oyebamiji (APC), read across Nigerian press coverage as the candidate of Gboyega Oyetola, the former governor Adeleke unseated. The PDPPDP itself, which swept every Osun federal seat in 2023, held a primary but failed to complete its nomination, and is not on the ballot; neither is LPLP. The ADCADC fields ex-Speaker Najeem SalaamNajeem Salaam (ADC), with former governor Rauf Aregbesola, now the ADC’s national secretary, personally leading the campaign.

The administration of the vote has its own storyline: a court order that briefly threatened to deregister both Accord and the ADC was unanimously overturned by the Court of Appeal on 28 July; the resident electoral commissioner was swapped in four months before the poll; and campaigning was set to end at midnight on Thursday 13 August. The late SDP addition has costs: reprinted ballots and result sheets were due in the state barely a day before the poll, and Yiaga Africa (which is fielding 332 observers, one per ward) warned publicly that the compressed logistics create avoidable risks.

Fifteen candidates are running: fourteen cleared in March, plus the SDPSDP’s Francis Olugbenga Ajala added by INEC on 12 August, three days before the poll, in compliance with a Federal High Court judgment; ballot papers and result sheets are being reprinted to carry the extra line. Three define the race: Ademola AdelekeAdemola Adeleke (A), the incumbent defending on a new party; Bola OyebamijiBola Oyebamiji (APC), carrying full federal backing; and Najeem SalaamNajeem Salaam (ADC). The practical consequence of the party shuffle: no polling unit’s 2022 colour is a safe guide to Saturday. The votes Adeleke won as a PDPPDP candidate must transfer to a new platform; the opposition vote that split three ways in 2022 is arranged differently now.

INEC’s register lists 2,339,233 voters across 332 wards and 30 local governments. That is growth of +384,433 over the 2023 general-election register, roughly thirteen times the margin that decided the last race; 365,356 of the new names registered for the first time. The growth alone is large enough to decide this election, whichever way it breaks, or doesn’t turn out at all.

We hold all 3,763 polling-unit result sheets from Osun 2022, photographed, extracted and mapped, from Osogbo (PDP +11.8)(PDP +11.8)’s 282 units down to Ola-Oluwa (APC +12.0)(APC +12.0)’s 62. Each unit reporting on Saturday will be compared against its own 2022 sheet, not against a model. It is the map behind this page: open it fully here.

In July we reconciled that archive against the declared result, LGA by LGA, using the published 30-LGA collation table (archived, with both party columns summing exactly to the declared totals). Every LGA matched, and nothing suggested totals changed between the polling units and the collation table. That is what we call a collation-shift flag; we found zero. The archive is also re-read continuously: as this page renders, 2,887 of the 3,763 sheets have cleared both machine checks and human review (the only totals this site publishes), and the rest sit in a re-reading queue. On the cleared units our sums read PDP 292,236PDP 292,236 and APC 272,861APC 272,861, set against INEC’s declared 403,371 and 375,027: the difference is review coverage in progress, not disagreement. Same winner, same geography, no flip signal.

Two LGAs deserve a flag for Saturday: Odo-Otin (PDP +1.1)(PDP +1.1) and Isokan (APC +0.7)(APC +0.7) were so close in 2022 that our sheet sums and the declared result sit on opposite sides of the line: margins of a few dozen votes, smaller than the read-errors in our own archive. In a state with Osun’s history, individual LGAs can be that close. Watch them counted unit by unit, not called.

And the 2022 result itself was nearly undone by exactly this kind of arithmetic. An election tribunal voided 181,540 votes across 744 units on an over-voting theory built from a BVAS accreditation report. The ruling was reversed on appeal, and finally at the Supreme Court, because that report was never demonstrated in open court; no machine was ever produced. We hold the full 744-unit schedule from the judgment, transcribed and arithmetic-checked against its own totals. That is the lesson this archive is built on: the last Osun governorship was decided by accreditation-versus-votes reconciliation, and this time every such check can be made in public, sheet by sheet.

Every polling unit counts its ballots in public and writes the totals on Form EC8A. Under the Electoral Act 2026, the sheet is photographed at the unit and transmission to INEC’s IReV portal is mandatory. INEC has billed Osun as the first full test of that requirement, with all 3,763 units activating together at 08:30. From the first upload onward, we archive each photo, read it, and test it against its own arithmetic. A sheet that fails (more votes than accredited voters, totals that don’t add) is held out of every figure on the site until human review concludes, and the sheet itself stays open to inspect.

Two consequences follow, and both are deliberate. First, our total will trail the declared one all night, because we only count what we have verified. Second, we will never call the race: we publish verified sheet totals and how much of the state has actually reported, and the declaration belongs to INEC. None of this is improvised: our anomaly thresholds and decision rules were fixed in July and committed to our repository before polling day, with the commit timestamp as the public record, and every pre-registered metric will be reported whatever the result shows. Any number on the site can be checked back to its sheet.

The clock. Polls close at 14:30. Because all 3,763 units open together at 08:30, the first sheet photographs should reach the portal in one wave from mid-afternoon, and our first verified totals follow as each sheet clears its arithmetic checks and human review. From then on the count moves in bursts, not a stream; a quiet half hour usually means sheets are queuing, not missing. Our total trails the declared one all night by design.

The reading order. Coverage first: what share of units is verified, and from where: an early lead built on urban Osogbo (PDP +11.8)(PDP +11.8), which uploads fast, says little about Ijesaland (PDP +5.6)(PDP +5.6), Ilesa and the five LGAs around it in the state’s east, or the Iwo axis (APC +4.4)(APC +4.4), Iwo with Ayedire and Ola-Oluwa. Then margins. Then swing: this page will show each LGA’s movement against the same units’ 2022 sheets, and those swings do not read equally. Bola OyebamijiBola Oyebamiji (APC) inherits a party line that ran last time, so his swing is apples-to-apples, while the question of the night is how much of Adeleke’s 2022 vote transfers from the PDPPDP line to AA. Watch Odo-Otin (PDP +1.1)(PDP +1.1) and Isokan (APC +0.7)(APC +0.7) late: both were inside a few dozen votes last time.

What winning takes. A governorship is not won by margin alone: the winner needs the most votes plus at least a quarter of the vote in two-thirds of the LGAs (20 of Osun’s 30). And if the final margin is smaller than the registered voters in units where voting was cancelled or never held, INEC declares the election inconclusive and runs a supplementary poll in those units. That is exactly how 2018 came down to a rerun in seven polling units and, a year of litigation later, 482 votes at the Supreme Court. In Osun, the counting is the story.

And one promise about the gap: if our verified count and the declared count diverge beyond what unreported units can explain, that gap itself becomes the story, and every sheet behind our side of it is public.