Why Smartmatic's PCOS/SAES Passed the SBAC Accuracy Test
Posted from http://pmana.multiply.com

The report on the 26-item testing of Smartmatic's PCOS/SAES indicated that the SAES passed the accuracy test at 100% accuracy.  The accuracy test only requires that the PCOS machine be able to correctly read marks on the ballot with an accuracy of 99.9995% or a rate of 0.999995. This means only 5 errors in 1,000,000 are allowed, or one error in 200,000. The SBAC keeps insisting that 99.9995% accuracy means one error in 20,000.  This is wrong. The percentage 99.9995 or rate of 0.999995 means 5 errors in one million or one error in 200,000, and it does not need a PhD in Mathematics to explain this Grade 5 exercise on percentage computations.  During the testing, the SBAC allowed the TWG to use 1000 pre-marked ballots, each with around 20 marks in the ovals, for a total of 20,000 marks. The markings were all properly and evenly applied to completely fill each marked oval with felt-tip ink. Even the worst scanner will be able to read such marked ballots at an accuracy of 100%.

I claim that in the default configuration of the SAES, using substance 24 or thicker white paper ballots, under actual election conditions where people may use pencil, ball pen, rolling ball jotter, and felt-tip pen, and using all allowable marking styles (dot, check mark, cross mark, and complete shade), the SAES will be lucky to achieve an accuracy of 50%.

I have two reasons for this claim:

(1) Experiential.  During the demo of the SAES at UNTV Saturday afternoon last week, with TV personality Jay Sonza watching, the SAES refused to read any marking that is a single dot in the middle of the oval.  We had to redo all these marks and completely shade the entire oval before the SAES accepted our ballot.  On election day, May 10, 2010, the SAES will not return your ballot to you, so that you can re-shade all your dots.  It will just simply NOT count your dots, because under default configuration, it does not see your dots.

(2) Probabilistic. Let us remove ballot paper as a variable (assuming that Comelec will use the paper quality and whiteness that is sufficient for the job), and consider only marking instrument and marking style as contributing variables. Assume that voters use the following marking instruments: 2B pencil, HB pencil, 0.5 ball pen, 0.9 ball pen, fine felt-tip pen, and medium felt-tip pen, or a total of six different kinds of marking instruments.  Assume also that voters will use all marking styles (dot, check, cross mark, and complete shade), or a total of four marking styles. So there are a total of six times four or 24 combinations of marking instrument and marking style, possibly producing at least 24 shades of gray markings in the ovals.  But the SAES has the most primitive 4-bit scanner, capable to recognizing only 16 shades of gray, and we are asking it to distinguish more than 24 shades of gray marking that voters can use for marking their ballots.  I know that this is an oversimplification, but I think that you get my point that the SAES hardware can not cut it.

If I were to design a proper test instrument to determine election-day accuracy of the PCOS machine, I have to prepare 40,000 pre-marked 30-inch long ballots, each ballot with 25 marks (1 mark for president, vice-president, governor, vice-governor, 12 for senators, etc), for a total of 1,000,000 marks. These marks will contain the four different varieties of markings. Then I have another file containing an interpretation by a team of 10 intelligent human beings, for each of these 40,000 ballots. For each marked ballot, the ten have to agree unanimously which are marks and which are not. This human interpretation will provide a check to the PCOS machine. The PCOS will read each ballot and write a tiff file (the ballot image), and its interpretation.  These tiff file ballot images and the PCOS interpretation will be transfered to another computer, which will do the comparison of the PCOS interpretation with the human interpretation.  For each mark or lack of mark in which the PCOS and humans disagree, this disagreement will be marked as a PCOS read error, because we take the human interpretation as the correct one, being the collegial effort of a team of 10 intelligent humans.  But this was not the way the SBAC conducted its accuracy test.  There was no controversy as to whether a marking was a vote or not, since all marks were made by complete shading using felt-tip pen, a condition which will not apply on election day.

Now you decide for yourself whether the testing was done with integrity and fairness.

visit http://pmana.multiply.com