Billcast Runs on Muse Spark 1.3

Paste a California bill. See if it becomes law.

Muse Spark reads it, a model trained on 19 years of bills scores it, and Muse Spark explains and improves it.

  • ranked right on a session it never saw
  • 19 yearsof votes and vetoes
  • 3Muse Spark jobs
Or try one:
Muse Spark fills in the subject, committee and fiscal flags, then explains the odds.

Chance of being signedstatistical model

Band: likely range. Line: average bill in ().

In plain Englishwritten by Muse Spark

Paste a bill or pick a sample above. Muse Spark will explain what drives the odds, where bills like this stall, and what would help.

Rewrite it to passdrafted by Muse Spark, scored by the model

Muse Spark amends the bill for its weakest steps, like a staffer before a hearing. The model scores each draft, so the gain is measured, not claimed.

What moves this estimate

Change from each input, compared with a typical bill.

Similar past bills

Past bills whose titles share words with the subject line.
Model inputsMuse Spark fills these in. Adjust anything.

Context

The author and the committee

Loading rosters…

How accurate is it?

Tested on , a session it had not seen: AUC (0.5 is a coin flip). Bills grouped by predicted chance, against the share actually signed:
What the model can and cannot see

How Billcast works

The odds come from a tested statistical model that runs in your browser. Muse Spark does the reading and writing around it, and never sets the number. Trained on .

  1. Read

    Muse Spark turns a digest into model inputs and quotes the phrase behind each fiscal flag.

    Muse Spark 1.3
  2. Score

    Five logistic regressions, one per step, trained on every bill since 2007 and checked against a session they never saw.

    Statistical model
  3. Explain

    Muse Spark reads the step odds and factor weights and says why, in plain English.

    Muse Spark 1.3
  4. Rewrite

    Muse Spark drafts amendments aimed at the weak steps. The model re-scores them, so you see the real change.

    Muse Spark + model