The Case of a Potentially Critical Software Bug That Inflated Senatorial Overvotes in the 2025 Midterm Elections And Why the Total Overvotes Per Precinct Are Suspiciously Divisible by 12

The 2025 midterm elections may have been marred by an overlooked systemic software error, one that inflated overvote figures by 1,200% and potentially misrepresented the voice of millions of Filipino voters.

This is not a baseless claim, nor is it hypothetical. It’s based on publicly available data, analyzed by NAMFREL, an independent and accredited watchdog. Their analysis shows 16.8 million senatorial overvotes across roughly 56 million voters, implying that nearly 30% of all senatorial ballots were voided due to overvoting.

At face value, this number is alarming. But what’s more troubling is this:

Every overvote figure across precincts is perfectly divisible by 12.

And 12 just happens to be the exact number of seats in the senatorial race.

This statistical pattern is unnatural. In a real-world voting scenario involving millions of individuals, random voter error would not produce such mathematically consistent results. So why is every precinct’s overvote total a multiple of 12?

A Misinterpretation of the Overvote Mechanism

In any election, an overvote occurs when a voter selects more candidates than allowed. In the senatorial race, that means shading 13 or more candidates instead of 12. A well-designed vote-counting system should register 1 overvote per ballot, and disregard that specific race on that ballot.

But based on the numbers, it appears the MIRU Automated Counting Machine (ACM) applied incorrect logic. Instead of recording 1 overvote per overvoted ballot, it recorded 12 overvotes, one for each candidate slot.

This kind of error could only occur at the software level. And this is where accountability must be demanded of:

Pro V&V, the independent software reviewer,
The Technical Evaluation Committee, which approved the source code,
And COMELEC’s own technical team.
Were these patterns not visible during testing? Was the overvote logic not reviewed line by line? Or worse, was it misunderstood entirely?

The Problem with the Formula Used by MIRU

Let’s look at how this inflated overvote number may have been derived:

Likely Faulty Formula:

X = (V × C) – Z – U
Where:
X = Overvotes
V = Valid Ballots
C = Number of Senators (12)
Z = Total Votes for All Candidates
U = Total Undervotes

This formula treats overvotes as the difference between the maximum theoretical votes and the sum of recorded votes and undervotes. This is incorrect. It creates the illusion of more overvotes than actually occurred.

Case Example:

Region V, Camarines Norte, Labo, Mabilo I (Cluster ID: 16060060)

V = 418
C = 12
Z = 3,038
U = 1,810

MIRU Calculation:
X = (418 × 12) – 3,038 – 1,810 = 168 overvotes

But let’s apply the correct definition, where an overvote is a discarded race per ballot, not a calculated leftover:

Correct Formula:

X = V – ((Z + U) / C)

Applying the data:
X = 418 – ((3,038 + 1,810) / 12) = 14 actual overvotes

The inflated figure (168) is 12x higher than the real one (14). The math doesn’t lie.

What This Means Nationally

If this pattern holds across precincts, then the 16.8 million overvotes reported could actually represent just 1.4 million ballots, not 16.8 million. That brings the real overvote rate down from 30% to just 2.5%—a normal and expected figure.

Accountability and Action Needed

We must now ask:

Did Pro V&V fail to detect this during source code review?
Did the Technical Evaluation Committee and COMELEC overlook or misunderstand the overvote logic?
Why were watchdog groups not granted access to deeper technical validation of how overvotes were recorded?
 

This is not a political accusation, it is a call for technical accountability. Systems fail. Bugs happen. But transparency and correction must follow.

We Call On COMELEC and Its Partners to:

1.  Publicly disclose the overvote counting logic and whether it follows per-ballot or per-candidate methodology.
   
   2. Explain why all overvote numbers are divisible by 12.
   Allow independent audits of MIRU’s source code and overvote logic, including a line-by-line review.
   
   3. Reassess the 2025 senatorial vote tallies if this bug is confirmed to have affected legitimate votes.
    

This issue is not about party or preference. It is about math, fairness, and the sanctity of every Filipino vote. The nation deserves a clear, public explanation and a commitment that it will never happen again.

Jay Jimenez, CISSP
Founder, OpenVCM.org