In analyzing the lists of alleged felon voters provided by the political parties, The Seattle Times applied a statistical method called proportional deduction, subtracting illegal votes precinct by precinct, in proportion to how the candidates fared in each. The Republicans have asked the court to use this method.

The Times first looked at all 1,740 felons on both lists: 946 from the Republicans and 794 from the Democrats. Proportionally deducting these felons would take 837 votes from Gregoire and 820 from Rossi (most of the rest are subtracted from Libertarian Ruth Bennett and write-in candidates). That leaves a net loss of 17 votes for Gregoire, cutting into her 129 lead but still leaving her ahead by 112 votes.

However, the Times' determined the lists were not entirely accurate. So reporters set up a two-step process for analyzing the felon-voter lists with adjustment for errors.

First, they checked the Republican list, which was filed with the court in early March. The Times studied 289 randomly selected felons from the GOP list. The three-month investigation found about 11 percent of the felons were wrongly included for various reasons.

Taking that 11 percent error rate into account and proportionally deducting votes using only the GOP list, Gregoire loses 486 votes and Rossi loses 306. That is a net loss of 180 votes for Gregoire, erasing her 129-vote lead and temporarily giving Rossi a 51-vote lead.

Next, the reporters looked at the Democrats' list, which cited alleged felon voters that were overwhelmingly from areas that favored Rossi in the election.

The Times reporters were not able to determine an error rate for the Democrats as they did for the GOP because the list was not submitted until the end of April, then updated May 6. The party also would not provide court case numbers for the alleged felony convictions, which the Republicans included.

Reporters did do some initial checking and found 33 felons were wrongly included because they had a received an official certificate restoring their voting rights. (The Times found 32 felons that had been discharged on the Republicans' entire list.)

Without a definitive error rate, the newspaper instead used proportional deduction to show how many felon voters the Democrats would need to overcome the hypothetical 51-vote lead from the GOP list. Because the list is so skewed toward Rossi precincts, The Times determined the Democrats could win with just 210 of their 794 felon voters, assuming any errors were distributed evenly throughout the list.

That would equate to an error rate of 74 percent. Any error rate lower than that, and Gregoire would prevail.