Back to the boards · All fighter profiles

Paulo Costa

Light Heavyweight · REAL v1.0 (registered) #2; v1.1 (shadow) #2

Descriptive profile. The resume rank is what has been earned; these statistics are context, not a prediction. A test on Oct 20, 2026 found that career statistics add nothing reliable to age, layoff and experience as predictors of the next result (experiments/REGISTER.md, EXP-2026-10-20-14).

Weight band 2-2: the 10th to 90th percentile of his position across weight vectors drawn near the 70/30 weights (the position does not depend on the weights). Rating 1,726, deviation 127: measured from 12 UFC bouts; a one-bout newcomer sits near 290. Division set by a documented override (config/division_overrides.yaml).

Age 35 (born 1991-04-21) · Height 6 ft 1 in (185 cm) · Reach 72 in (183 cm) · Stance Orthodox · 204 days since last bout (2026-04-11) · Last five 3-2 · Streak W2

Why he is ranked here

Rating 1,726 is the division's 3rd highest of 51 board fighters; the ledger stands at +2.8 (+2.3 from quality wins, +0.5 from close losses to top-5 fighters; losses count for three years); 5 quality wins in 15 years, 3 over top-10 opponents; no form penalty (3-2 last five). Score +0.71 = +0.51 (rating term, 70% weight) +0.20 (ledger term, 30% weight) -0.00 (form penalty); score order put him at #3; the title-cycle rule (a losing title challenger is held at #4 or lower for 270 days, and the fighters behind move up) placed him #2.

Rules, weights and worked examples: methodology.

Last ten bouts

Opp. official rank = the UFC media-panel rank of the opponent on the date (C champion, #1 to #15, unranked; n/a before Feb 2013 or where no official board existed). Model p = the predictive model's pre-fight win chance for him (frozen performance-adjusted Elo; bouts before 2020 were in its tuning sample). Linked opponents have a profile.

Career statistics

Through 2026-10-31, UFC bouts only. Percentile = share of the 51 board fighters at Light Heavyweight at or below this value; high is a position in the division, not a verdict.

Board position since Oct 2023

Position on the registered v1.0 board at each weekly snapshot (C = title holder or reserved, drawn at the top; lower is better). Absent from 1 of 132 snapshots (gaps in the line).

Predictive rating by bout

Rating after each UFC bout from the frozen predictive model (performance-adjusted Elo, data/processed/elo_history.csv; a debut starts at 1,500). The resume rating's history is not stored, so this is the predictive line, which is a different number from the resume rating in the header.

Trajectory

Recent performances are in line with the rating (-22 points).

Performance relative to the rating, by age

Each dot is one decided UFC bout at his age then: the performance score minus the rating's expectation, on the score's own scale from -1 to +1 (filled = win, hollow = loss; above zero is better than the rating implied; a finish as a slight favourite is about +0.45). The dashed line is the population curve for all fighters; the solid line is this fighter's own fitted level and trend as of the data date, weighted toward recent bouts. One unit sustained over many bouts corresponds to about 695 rating points; a single bout's residual is noisy.

Held-out evidence (walk-forward 2016 to 2026, EXP-2026-10-21-01): the age curve held in every test year; the individual component is a weak signal (correlation 0.06 with the next bout's residual; among fighters aged 33 or over, the lowest predicted quartile won 38% of next bouts against 53% expected, the highest 48% against 52%) and adds nothing to bout prediction beyond the shadow model. No board uses any of it: the rankings answer who has earned the place, not who wins next (EXP-2026-10-20-09). Method: src/mmalab/decline.py; settings: config/decline.yaml (expectation from the resume rating; one unit of residual is about 695 rating points).

Context flags

Descriptive only. Each line is a population fact about fighters like him, not a forecast for him; the register entries hold the measurements.

• Age band: 33 to 35 (age 35). On the board's own history (3,748 bouts between two board fighters, 2014 to 2026) the higher-scored fighter won 64% of bouts when under 30, 59% at 30 to 32, 57% at 33 to 35 and 42 to 46% at 36 and over (EXP-2026-10-20-09). A population rate, not a forecast for him.
• Not raised: no KO/TKO loss in the last 90 days; last bout within 12 months; 12 UFC bouts, more than the provisional limit of 3.

REAL v1.0 (registered method) · build c166ee8 · data through 2026-10-31 · page built 2026-11-06 · changelog. Independent fan research project built on public UFCStats data. Not affiliated with, sponsored or endorsed by UFC, Zuffa, LLC, TKO Group Holdings, or any athlete. All trademarks belong to their owners and are used only to identify athletes and events.