TopBJJ Index Youth

TopBJJ Index

Search TopBJJ Index
Language selector
Menu

TopBJJ Index methodology

How the Rating works

A transparent, reproducible rating built from indexed youth competition results.

Only events in the NL–BE–DE–FR–LU region count toward this rating and position.

What the rating measures

The rating estimates recent, opponent-adjusted performance inside the indexed youth competition pool.

  • The rating uses eligible results from the last twelve months.

  • Eligible verified matches follow the same model rules whether they were imported automatically or entered by an administrator.

  • Published source corrections update the visible result and the next rating and ranking publication.

  • Wins and losses are weighted by the expected result and the strength of the evidence.

  • A public position requires at least five matches, 3.5 effective evidence, three opponents and two events.

  • Overall combines Gi and No-Gi according to the evidence actually available.

  • Overall is a performance index, not advice to match athletes of different ages or weights.

  • Uncertainty describes how stable the estimate is, but does not determine eligibility by itself.

How evidence is counted

Eligible results are weighted for independence and reliability before they support a published rating.

  1. Result Each eligible match starts with its recorded win or loss.
  2. Opponent strength Results against stronger opponents carry different context than results against less-tested opponents.
  3. Evidence strength Independent events and opponents support a rating more than repeated evidence.
  4. Status and stability Evidence volume and diversity determine rating status. Estimated uncertainty and competition-network connectivity describe how stable the rating is.

What does not add a bonus

The rating is based on results, opposition and evidence, not on presentation details.

  • Medals do not add a rating bonus.

  • Submissions do not add a rating bonus.

  • Country and academy do not add a rating bonus.

  • Belt and weight do not add a rating bonus.

  • There is no automatic inactivity penalty.

Formats

Each format is rated from its own eligible evidence so comparisons keep their competition context.

Gi

Gi

Gi rating uses eligible Gi matches.

No-Gi

No-Gi

No-Gi rating uses eligible No-Gi matches.

Overall

Overall

Overall combines available Gi and No-Gi evidence.

With little evidence in a second format, Overall stays close to the actual evidence share. As both formats reach 3.0 effective evidence, the mix grows smoothly toward the 25%–75% bounds. One available format contributes 100%.

When an athlete receives a published index position

The overall ranking is the standard. The list can be refined once by age or by weight, but never by both at the same time.

  • The athlete is active, public and has enough independent rating evidence.

  • Boys and girls are included in the same ranking.

  • An age ranking includes every weight within one age band.

  • A weight ranking includes every age within one broad TopBJJ competition-weight group.

  • A competition-weight group reflects recent indexed division context, not measured body weight; groups below 20 athletes receive no separate position.

Ranking limitations

A broad ranking creates a more meaningful position in a small sport, but it is not advice to match children with large age or weight differences.

  • Overall places performances from the broad indexed youth pool on one scale. This is a performance index based on opponents and results, not a prediction of a responsible match between different ages or weights.
  • The visible rating is rounded; positions use the more precise model estimate. Only fully equal model outcomes share a position.
  • Only active athletes with enough evidence count toward the population size.
  • Missing or corrected source data can change ratings and positions.
  • The index only covers the indexed competition pool and is not an official world ranking.
  • Broad competition-weight groups with fewer than 20 athletes do not receive a separate published position.

Technical methodology

The same completed run supplies the public ratings, evidence figures and cohort positions.

The published rating is a weighted rolling 12-month Bradley-Terry model.

Configuration

Rolling window in months
12
Ranking eligibility policy
C
Minimum rated matches required
5
Minimum independent-evidence threshold
3.5
Minimum unique opponents
3
Minimum distinct events
2
Full-evidence level per format
3
Same-event evidence discount exponent
0.25
Repeated-opponent evidence discount exponent
0.5
Published scope minimum size
20

Formulas

Context v1 expected win probability: 1 / (1 + exp(−(athlete skill − opponent skill + context offset))).

Evidence weight: w = event_match_count^(-same_event_exponent) × pair_meeting_count^(-repeated_opponent_exponent).

Completed run

rating@2026-08-31-72befccf6acc57c8

Reproducibility

Public figures come from a stored, completed model run with its version and calculation date.

  • The public rating follows the opponent-adjusted skill estimate. Uncertainty stays separate from the rating; evidence status shows how much data supports the estimate.
  • A merged division is shown as event context and never overwrites the personal age category.
  • Match impacts are current marginal estimates that may change with later results; they are not additive points.
  • The published period contains twelve complete calendar months and advances once per month.
  • Older matches stay visible as history but have impact 0 after leaving the rolling period.
View rankings