# Numbers that look like performance claims but are not ours to measure.
#
# Format: <number>  <reason>
# Checked by tools/check_claims.py. Keep this list short: if a number can be
# measured, measure it instead of listing it here, and if it can be cited say
# so with {{claim:<run-id>.<field>}} rather than hoping the value is unique.
#
# The check reports entries that back nothing any more. An entry that has gone
# quiet is either a claim that was fixed or a scanner that stopped looking, and
# both are worth noticing -- so prune it and say which, as below.

# --- figures published by other authors, attributed, never presented as ours ---
86   published Schwartz & Hearst F1 range quoted in DECISIONS.md, attributed
89   published Schwartz & Hearst F1 range quoted in DECISIONS.md, attributed

# --- REMOVED, because they are no longer detected as claims at all ---
#
# Five entries were dropped when the scanner stopped treating digits welded
# into an identifier as free-standing numbers. Each was a workaround for a
# false positive, and none of them is a claim any reader would recognise:
#
#   2008   the "2008;9:402" of a BMC Bioinformatics citation -- one token, not
#          a number, so it is no longer read as one.
#   2003   the year in "Schwartz & Hearst (2003)".
#   1250   the "1250" of MED1250, a corpus name.
#   31     the size of Ab3P's SingTermFreq.dat. Its line was flagged only
#          because the filename "Lf1chSf" contains the letters "f1"; keywords
#          are now matched as words, so the line carries no metric keyword.
#   48     the size of Ab3P's Lf1chSf, same line, same cause.
#
# They are recorded here rather than deleted silently: if any of them starts
# being flagged again, the scanner regressed and this is the note that says so.
