How accurate is the ENSI.Score rating

How the number is made

Within the two hours before a match's announced start, the probability that team 1 wins is computed from the two ENSI.Score ratings shown on the site at that moment and written down. It is computed again whenever the teams, a rating or the start time change, until five minutes before the announced start or until the match goes live. The last computation is the one the match page prints and the one measured here; the earlier ones are kept but not measured, and none is ever edited. If a match starts earlier than announced, the last computation may fall after the real start: it still uses only ratings from before the match, because a rating changes only when a series ends.

p1 = 1 / (1 + 10(r2 − r1 − 0) / 400)

r1 and r2 are the two ENSI.Score ratings at the moment the probability is locked. The published model is elo-v1: an offset of 0 and a scale of 400.

A match counts towards the figures above when all of this is true of it, inside the 365 days the window covers:

• The match was played and ended inside the window, and the score names a winner.
• Draws and matches decided by a forfeit are left out: there was nothing on the server to predict.
• The prediction still describes the match. If the start or a rating changes more than five minutes before the announced start, it is marked and a new computation replaces it; after that it stays as written. If the sides are swapped, it is marked at any time, since its number belongs to the other team 1, and a match left without a valid computation is not measured. A marked prediction stays readable as the record of what was claimed.
• Nothing is measured on a match that was already over. The accuracy of a model is what it said beforehand, and only that.

The calibration bars are wide on purpose. Telling a group that says 70 % apart from one that wins 75 % of the time needs some 16 500 matches — about a year of them — and until there are that many the honest thing to print is the interval rather than a verdict.

php artisan e:elo-accuracy --report=469

select series_id, p1, y::int as y from elo_accuracy_report_rows where report_id = ?

php artisan e:elo-accuracy --live --window-days=365 --window-end=2026-10-20

select sp.series_id, sp.p1,
       (case when s.score_1 > s.score_2 then 1 else 0 end) as y
from series_predictions sp
join series s on s.id = sp.series_id
where sp.model = ? and sp.status = 'active'
  and s.date_end is not null
  and s.date_end >= (?::date - make_interval(days => ?::int - 1))
  and s.date_end  < (?::date + interval '1 day')
  and s.score_1 is not null and s.score_2 is not null and s.score_1 <> s.score_2
  and coalesce(s.forfeit_team_1, false) = false
  and coalesce(s.forfeit_team_2, false) = false
  and (?::int is null or s.game_id = ?)