Republicans hoping to overturn the election of Gov. Christine Gregoire have a steep mountain to climb because they rely heavily on an uncertain mathematical formula to prove Republican Dino Rossi collected more votes than Gregoire did.

Chelan County Superior Court Judge John Bridges set high standards — appropriately — for proving felons who voted illegally should be subtracted from the count for both candidates. Republicans must present clear and convincing evidence Rossi won, and that won't be easy. The trial begins May 23. The GOP will base part of its case on a math formula known as proportional, or statistical, analysis that assigns votes to felons voting illegally on the same mathematical basis as their home precinct voted.

For example, if 10 felons barred from voting live in a certain precinct in King County and that precinct voted 60 percent for Gregoire, six votes would be subtracted from her total, four from Rossi's.

Republicans have few other ways to prove how perhaps more than 1,000 felons voting illegally cast their ballots, aside from the impractical approach of bringing them into court and asking them.

So Republicans will bring in experts who will use math to describe likely voting patterns.

Likely is an important word. The math makes sense in an academic debate. But listen carefully. The experts speak in terms such as "likely" or "most likely," or "estimate" or "infer."

The math can be solid and the conclusion all wet. Bridges must be careful before allowing a dry statistical formula to overturn an election. Invalidating an election surrounded by uncertainty should be done only if a new approach provides greater certainty. Proportional analysis raises as many questions as it answers.

How do we know felons vote like the rest of voters in a given precinct?

Felons tend to be male; men vote differently than women. National studies show felons tend to be minorities and from low-income backgrounds who vote Democratic. Are minorities as big a share of the Washington state felon population as they are nationwide?

Another unknown: Gregoire, as former state attorney general, was the state's top law-enforcement officer. How did that affect voting preferences of felons and other illegal voters?

At this point, the burden of proof rests heavily on proving the statistical analysis is valid. It should.

In a contested election such as this, the goal has to be to bring more clarity, not less.