Rank- and role-adjusted scouting from the OTPStats corpus: which factors actually correlate with winning, model-scored comp building from the pool, and best-available fills per role.
Sign in and request suite access on the Tools page — an admin approves each account.
Our five vs theirs: win probability from the wins-weighted composite (K calibrated on 997 harvested customs) plus a small comp-shape term from Ascendant+ ranked data; each team's bread & butter vs what's exploitable, judged against what each player's own role mix posts (a sentinel isn't "low kill output", a duelist isn't "never plants"); map records; per-player scouting cards. Pool: loading…. Names outside the pool price at the pool floor.
Your saved recruiting pool — signed-in site users only, persisted to your coach account. The chips are each player's trailing stats z-scored against the Ascendant+ population (green above, red below, dim = not enough data) so complementary profiles jump out when you mix and match.
Pick up to five players (your pool first, then signed-in users, then scouted candidates). The edge is the model's win probability against an average team of equal rank — driven by trailing pre-match profiles, never same-match stats.
Every profiled factor for this five, z-scored against the Ascendant+ population and sorted by how strongly that factor correlates with winning (predictive lens, r shown in each chip). Chase green on the left side of this list first — those are the stats that matter most.
Ranked by the edge the model predicts if they take the next seat next to the players selected above. Pulls from your pool by default — widen the scope when you want to scout beyond it. Close candidates are genuinely close — settle those in tryouts, not in a spreadsheet.
| PLAYER | SEAT EDGE | GAMES | TIER | SRC |
|---|---|---|---|---|
| Loading… | ||||
Two lenses. Predictive: the player's trailing pre-match profile vs their next match's result — the honest scouting signal. Descriptive: same-match values — winning inflates these (you get more kills because you're winning), so read them as "what winning looks like", never as "what to buy". Every cell shows its sample size; a correlation only means something once |r| clears the noise band ±1.96/√n — cells still inside it are dimmed and tagged "noise" (not enough games yet, not evidence of no effect). KAST/trade/clutch cells are the newest data and grow daily.
| FACTOR | ALL | DUELIST | INITIATOR | CONTROLLER | SENTINEL |
|---|---|---|---|---|---|
| Loading… | |||||