Examining the tally of votes for positions audited in the July mock polls, experts said that there were instances where the PCOS machine overcounted and undercounted on a per candidate basis. At times, the overcount would be cancelled out by the undercount, presenting a zero variance. Clearly, this is “dagdag-bawas” in digital form. But in Smartmatic’s flawed logic, the accuracy rate would be 100% based merely on the zero variance. The correct approach is actually to sum up the absolute value of the differences between the manual count and PCOS count to get the correct variance. Based on independent computations of the mock polls results, the total number of miscounted votes is 231 out of the 8,295 scanned marks. This translates to a failing accuracy rate of 97.21519%, and an alarming counting error rate of 2.78481%. The error rate is 557 times higher than the Comelec’s specified error rate of 1 out of 20,000 marks!
So what are the implications? A study by Dr. Felix Muga II, Ateneo math professor and senior fellow at the Center for People Empowerment in Governance (Cenpeg), showed that based on the given PCOS error rate, the 2010 election results in several areas would be questionable. He cited the results in the congressional election in the lone district of Biliran province, where the winning margin was only 361 votes. Using the 2.78481% error rate, he said 2,287 votes were miscounted by PCOS, a figure that is more than six times the winning margin.