VCT
TYL0
G22
Final13h ago
TL1
PRX2
Final13h ago
NS0
NRG0
in 7h
KC0
XLG0
in 10h

Rating report

How accurate the rating is, how settled it is, and the records it has set.

How accurate is the rating?

Before every rated match, the rating implies a win probability. We keep those predictions and score them against what actually happened — across 4,864 decided matches. A well-calibrated rating means our 65% picks win about 65% of the time. Nothing here retunes the model; it only reports on it.

0%0%25%25%50%50%75%75%100%100%Predicted win probability
Each dot is a bucket of past matches. The dashed line is perfect calibration; dot size is how many matches landed in that bucket.
Predicted versus observed win rate by bucket
PredictedObservedMatches
7%17%142
16%31%244
25%36%344
35%42%505
46%47%821
54%57%1153
64%63%658
75%69%502
85%80%335
93%87%160

Brier score

0.228

Mean squared error of the prediction. Lower is better; 0.25 is what always guessing 50/50 would score.

Calibration error

0.048

Average gap between what we predicted and what happened, weighted by how many matches sit in each bucket. 0 is perfect.

Log loss

0.650

Penalises confident-and-wrong predictions much harder than cautious ones.

Measured over the matches the rating engine actually rates. Predicted 52.2% on average and observed 54.4% — those two should sit close together for an unbiased model.

How settled are the numbers?

A Glicko rating is an estimate with an error bar, not a fact. That error bar is the rating deviation — the ± 68 printed next to every rating on the site. It narrows as a team plays and widens again while a team sits idle, so a wide band means “we are not sure yet”, never “this team got worse”.

Rated teams
4113
Settled
761
Provisional
3352
Median RD
±154.8
  • SettledRD 0–600
  • Firming upRD 60–120166
  • WideRD 120+595

A further 3352 teams are still provisional — the engine flags a rating that way while its uncertainty is above 200 or it has played fewer than 5 rated matches. Those numbers exist, but they have not converged, so they are kept out of the main ladder and out of every record on this page.

Counted from the rating snapshot of 2026-09-25. What RD and volatility mean →

Records

Rating records

Methodology →

Highest ratings ever

  1. 1
    Enterprise Esports
    10 Aug 20262571±122
  2. 2
    Sharper Esports
    10 Aug 20262476±106
  3. 3
    Nongshim RedForce
    15 Mar 20262453±142
  4. 4
    Gentle Mates GC
    31 Jul 20262372±103
  5. 5
    S8UL Esports
    4 Jun 20262343±116

Highest this season

  1. 1
    Enterprise Esports
    10 Aug 20262571±122
  2. 2
    Sharper Esports
    10 Aug 20262476±106
  3. 3
    Nongshim RedForce
    15 Mar 20262453±142
  4. 4
    Gentle Mates GC
    31 Jul 20262372±103
  5. 5
    S8UL Esports
    4 Jun 20262343±116

Biggest single-match swings

  1. La Masia

    lost to FF · 6 Jun 2026

    2207 → 1964-243
  2. Equipe Caos

    beat HL · 18 Feb 2026

    1305 → 1524+219
  3. Havoc & Yonder GC

    beat NOVA.GC · 16 Sept 2026

    1448 → 1666+218
  4. FunPlus Phoenix

    beat AG · 3 Apr 2026

    1415 → 1620+205
  5. Karmine Corp GC

    beat G2G · 4 Feb 2026

    1464 → 1667+203

Longest runs at #1

  1. Enterprise Esports

    29 Jul 2026 – now

    58daysLive
  2. Team Falcons Vega

    13 Jul 2026 – 28 Jul 2026

    16days
  3. Team Falcons Vega

    7 Jun 2026 – 10 Jun 2026

    4days
  4. Enterprise Esports

    11 Apr 2026 – 14 Apr 2026

    4days
  5. S8UL Esports

    26 Jun 2026 – 28 Jun 2026

    3days

Records cover the public Glicko engine's own rating history, which begins 2 Jan 2026. Ratings the engine still flags provisional are excluded, so a wilder number from a team that had not converged will not appear here. “This season” is measured from 1 Jan 2026.

The full rating methodology — the Glicko-2 algorithm, decay, data sources, and known limitations — lives on the methodology page.