I Published the Grading Rules Before I Graded Anybody

Every political scorecard you have ever seen has a number on it and no formula behind it.

One group gives an official a 92. Another group gives the same official a 41. Both of them are certain. Neither one will show you the math.

That is not accountability. That is an opinion wearing a lab coat.

When I built RepWatchr, I made myself one rule before a single official got a grade. Publish the method first. Put it on its own page, in public, where anybody can read it and come after me with it.

It is up. It is called Performance Grade Method, version 1.0.

Here is what is in it.

The grade is built from five things, and every one of them carries a fixed, published weight.

Read those five again, because what is missing matters more than what is there.

Party is not in there. Ideology is not in there. Fundraising totals are not in there. Follower counts, press volume, and how much people like the man are not in there. The method page says it in plain words. Those things do not change the grade.

A yea and a nay are not scored as morally correct or incorrect. The voting weight asks a neutral question. Did the official take a documented position when they were eligible to take one. Showing up to the vote is the thing being measured. Which way they went gets reported, not graded.

That rule right there is the one I care about most, and I will tell you exactly why.

I know what it is like to be sentenced by rumor. I know what a cropped screenshot and a confident stranger can do to a man long before anybody in a robe has ruled on anything.

If I built a scoring system where an accusation could quietly shave points off somebody I disagree with, I would be running the same play that got run on me. I am not doing that. Not to a Republican, not to a Democrat, not to anybody.

So the method sorts evidence into tiers and puts a number on each tier.

Official roll calls, journals, filings, audits, final ethics orders, and court records carry full weight at 1.0. Official transcripts, meeting video, agency datasets, and direct official statements carry 0.9. Named independent reporting tied to primary records carries 0.7.

Campaign, advocacy, and social material carries 0.4, and the page is blunt about why. It is proof of the speaker's statement. It is not proof that the accusation is true.

Unverified material carries a zero. It is a research lead. It cannot move a score.

Then there is the part almost nobody else does.

If the record is thin, the official does not get a bad grade. They get no grade.

A full number and letter require at least eighty percent of the planned weight to be scoreable, seventy percent weighted coverage, sixty five percent confidence, four categories, and a scoreable ethics category. Below that there is a provisional number with no letter attached. Below that it is NR, not rated, published with a visible checklist of the evidence that is still missing.

Missing evidence never becomes a zero. It never becomes a 50. It never becomes a guess.

There is a response process, and it is on the same page anybody can read.

Verify and classify the evidence. Notify the official of a score moving integrity event or a major score change. Give ten business days for a response. Require a second reviewer on high impact records. Publish the source, the calculation, and the response. Route appeals to a reviewer who did not make the first decision.

Corrections create a new version instead of quietly erasing the old calculation. That one matters more than it sounds. If you can edit history with no trail, you do not have a record. You have a draft.

Community sentiment does live on a profile, but it is fenced off from the grade completely and labeled for what it is. A self selected signal. Not an election, not a scientific poll, and never an input to job performance.

Version one uses whole letters only. A, B, C, D, F. No pluses and no minuses. The method page calls that plus minus theater, and it is right. The evidence does not justify that kind of precision yet, so I am not going to fake it.

Here is where the whole thing sits today. The counters on the RepWatchr home page read 16,809 public profiles, 8,209 federal and state records loaded, 57 authority roles, and 59,289 source URLs.

Not 59,289 opinions. Links. Things you can open, read, and check against me.

I want to pull one lesson out of this for anybody building anything, because it is not really about scorecards.

If your business makes a judgment about somebody, publish the standard before you publish the judgment.

Publish it when it is still inconvenient. Publish it before you know who it is going to help and who it is going to hurt. Publish it before you have a favorite.

A standard written after the fact is not a standard. It is a defense.

The day I write a rule that protects a man I cannot stand is the day the rule is worth something. Until then it is just a preference with a number stapled to it.

They can argue with my grades all day. That is fine. That is the point.

What they cannot say is that I made the rules up after I saw the scores.