Give it a claim and a target. It tests to 50-digit precision. Does this hold? The anti-confirmation-bias tool.
pip install begump
from gump.verify import check
result = check("sqrt(2/pi)", 0.79891)
print(result['verdict']) # "CLOSE"
print(result['ppm']) # 7.2 (parts per million off)
print(result['claim_value']) # 0.7978845608...
print(result['target']) # 0.79891
print(result['digits_agree'])# 3
# Thresholds:
# PASS = within 1 ppm (< 0.0001%)
# CLOSE = within 1000 ppm (< 0.1%)
# KILLED = more than 1000 ppm off
pi, e, phi (golden ratio), sqrt(x), cbrt(x), ln(x), log10(x), log2(x), sin(x), cos(x), tan(x), asin(x), acos(x), atan(x), exp(x), abs(x), factorial(x), plus all arithmetic (+, -, *, /, **, ^).
Confirmation bias lives in the gap between "close enough" and "exactly right." If your formula matches to 3 digits, it might be coincidence. If it matches to 50, it is real. This tool shows you exactly where agreement ends.
PASS — Within 1 ppm. This holds to at least 6 significant figures. Real.
CLOSE — Within 1000 ppm. Suggestive but not proven. Could be coincidence at this precision.
KILLED — More than 1000 ppm off. This does not hold. Move on.