Getting your Trinity Audio player ready...

Assuming two definitions of over votes:
 Votes Cast (VC) > Registered voters on the roll (R)
 Votes Cast (VC) > Number of valid ballots issued to voters (B) (not including spoilt ballots)
 The number of valid ballots issued to voters (B) cannot be greater than the registered voters on the roll (R). Then, B ≤ R
 Assuming an 80% voter turnout at the polling station, then the no of valid ballots issued to voters (B) should equal 80 percent registered voters on the roll. That is, B = 0.8R
 It thus implies that the votes cast should be less than or equal to 0.8R (VC ≤ 0.8R) with VC = 0.8R being the maximum.
 If the sum the total votes cast (VC) which includes the rejected ballots and votes for each candidate exceeds the number of ballots issued (B = 0.8R), then, there is overvoting
 The true magnitude of over votes equal VC – 0.8R.
“There can be only one true magnitude of overvotes at a given polling station. By definition, this magnitude must be equal to the difference between the recorded total number of votes, VC, cast and the actual number of votes cast by eligible voters (equal to 0.8R in the above example), based on the principle oneperson onevote”. That is if VC ≥ 0.8R, then, it implies there are indeed overvotes.
It is time we pay more attention to ballot accounting and verification.
Link to full paper: http://www.uoguelph.ca/~jamegash/OverVoting.pdf