Public player and replay data
Saved statistics with HTML tables, Dataset metadata and matching JSON. No account is required.
Find a dataset
Open a player profile or replay and follow its public dataset links. Identity uses the numeric CFN ID; display names can change.
HTML: /data/players/CFN_ID/overview or /data/replays/REPLAY_ID/overview. JSON: /api/public/v1/players/CFN_ID/overview or /api/public/v1/replays/REPLAY_ID/overview. Replace overview with a dataset name below.
Download the v1 JSON Schema · Dataset sitemap index · General data limitations
JSON contract and pagination
The versioned URL and schemaVersion: 1 identify this format. Breaking changes require a new version. Every response contains kind, public id, dataset, name, description, source, timeWindow, updatedAt, limitations, subjects, coverage, links, pagination and tables.
Each table has a stable key, title, totalRows, ordered columns and rows. Columns define each field's key, meaning in the label, and unit. Rows contain only those keys, with string, number, boolean or null values. Null means unavailable; zero is an observed value. Identifiers are strings where needed to preserve precision. Dates are ISO 8601 UTC; daily dates use YYYY-MM-DD. Round codes are comma-separated in original round order.
Add ?page=2, or follow pagination.next.json / pagination.next.html. Each table has up to 100 rows per page and is paginated independently. The page count covers the longest table. Small tables may be empty on later pages. Follow every page to collect the full dataset. Invalid parameters return 400 in JSON; missing records and out-of-range pages return 404. HTML invalid requests return 404.
Matches and ratings sort newest first, with stable internal tie-breaking; internal row IDs are never exported. Pagination is a live view, not an immutable snapshot: concurrent syncs can move rows between pages. For archival collection, record timestamps and deduplicate observations. Rating archive and Battle Log tables intentionally overlap. The interactive chart deduplicates timestamp/rating observations, favoring Battle Log evidence.
links connects landing pages, JSON, related datasets, source profile/replay and this reference. HTTP Link headers expose HTML, the schema, and adjacent JSON pages. subjects links public player identities through stable identifiers shared with JSON-LD.
Freshness and units
Reads use stored data and never launch syncs or captures. updatedAt is the last relevant saved update, not request time. Profile synchronizedAt, benchmark time, match date and report capture time are distinct in coverage. JSON may be cached for 60 seconds. Upstream outages can leave data stale; a recent page request is not evidence of recent source data.
MR and LP are independent rating points. Win rates are percentages (0–100); gauge shares are fractions (0–1). Local result win rates count draws as half a win; official rates use wins/games. Habit rates are events/game except corner rates, whose exact time unit is not verified. Play time uses seconds. Play points are game counters, not currency or reviews; the additional profile point counters retain unverified Buckler semantics. League codes and input/action codes retain their game-specific meanings.
Round result codes: 0 lost, 1 standard victory, 2 chip damage, 3 time over, 4 draw, 5 Overdrive, 6 Super Art, 7 Critical Art, 8 perfect. Unknown codes remain unknown. Replay telemetry rounds and candidate/move indices are zero-based; player sides are 1 and 2. LP/MP/HP and LK/MK/HK mean light/medium/heavy punch and kick; di, throw and parry are observed button chords.
Player datasets
overview — Player identity and results
Public CFN identity, current character ratings, official mode totals and results in the stored match sample for this Street Fighter 6 player.
Source: Stored Capcom Buckler profile, current leagues and Battle Log records.
Window: Latest saved profile; local results cover at most the newest 10,000 stored matches across all modes.
- A CFN name is a game handle, not a verified real name or affiliation with SF6 Stats.
- Local win rate gives a draw half a win; official character win rates count wins only. Both use percent (0–100).
- MR and LP are separate game ratings, not reviews. Missing values are null, not zero.
characters — Characters and rankings
Per-character Master Rating, League Points, official rank codes and ladder position, character wins and games, and locally observed character usage.
Source: Stored Buckler character leagues, character win rates and local Battle Log counts.
Window: Latest saved official profile aggregates; local character counts use the newest 10,000 stored matches.
- Ladder position is the reported MR ranking, not a rank inferred from rating thresholds.
- Peaks use current leagues, merged saved chart history and the newest 500 displayed matches; they are not guaranteed lifetime maxima.
- Official totals and local samples have different coverage; neither is a complete downloadable Capcom match archive.
matchups — Character matchups
Player-versus-character results, split by played character and opponent control type for local records, alongside official matchup totals and available community win rates.
Source: Stored Battle Log results, Buckler official matchup aggregates and latest stored community matchup month.
Window: Local: newest 10,000 stored matches. Official: latest saved aggregate. Community: the stated published month.
- Local rates include draws as half wins; official rates use wins / games.
- Community rates describe the broader population for the stated month, not this player's results.
activity — Play time and opponents
Play time in seconds, character play points, daily win/loss results and frequent or favorable opponents in the saved match sample.
Source: Stored Buckler play time and play points; daily results and rivals computed from local Battle Log records.
Window: Official cumulative snapshot; local daily results and up to 15 rivals per group from the newest 10,000 stored matches. Recent MR changes cover ranked matches among the newest 500 displayed matches.
- Play points are game progression counters, not ratings or currency.
- Daily dates use UTC. Rival groups are a sample-based ranking, not an exhaustive opponent list.
- Recent MR changes reuse dashboard calculations: consecutive ratings where available, Elo estimates otherwise. A partial sum is not a verified total change.
habits — Habits and comparisons
Reported habit rates and gauge shares, elite benchmark comparison values, and available 7-day and 30-day changes between saved last-100-match snapshots.
Source: Stored Buckler battle_stats, saved elite benchmark and same-season HabitSnapshot comparisons.
Window: Reported habits reflect Buckler's last-100-match trends. 7d/30d compare snapshots near those dates, not matches isolated to those periods.
- Gauge shares are fractions (0–1); per-game rates are counts/game. Corner values retain Buckler's aggregate rate with no verified time unit.
- Benchmark players and sample size may be limited or stale. Player habits are profile-wide, even when compared to a character ladder.
- Snapshot deltas are changes in overlapping aggregate windows, not interval event counts or causal improvement. Missing baselines remain collecting.
matches — Stored match history
Paginated stored Street Fighter 6 matches with participants, selected and played characters, control types, outcomes, round result codes and pre-match MR/LP.
Source: Persisted Capcom Battle Log records, projected from the selected player's perspective.
Window: All locally stored matches, newest first; 100 matches per page.
- Only matches collected by this site are available; upstream battle logs have limited history.
- MR changes are estimates when consecutive observations are missing. This export publishes recorded pre-match MR/LP without inventing a post-match value.
ratings — MR and LP history
Recorded pre-match ranked ratings and archived MR/LP observations, with source, character, timestamp and available estimated MR change kept explicit.
Source: Ranked Battle Log observations and persisted local, sfbuff or imported rating points.
Window: All stored observations; each table independently paged newest first, 100 rows per page.
- The two tables overlap intentionally; the interactive chart merges timestamp/rating duplicates with live local battles preferred.
- Archived local deltas may be derived from consecutive ratings or estimated with Elo; imported deltas retain source semantics. Null character means unknown.
- MR and LP are independent. LP-only observations are retained, not recast as MR.
Replay datasets
overview — Replay participants and rounds
Participants, characters, controls, outcome, round result codes and available telemetry coverage for one Street Fighter 6 replay.
Source: Stored Capcom Battle Log and validated replay capture report when available.
Window: One replay; match date and report capture timestamp are separate.
- Unknown replay codes do not produce fabricated datasets. A report can exist without a stored Battle Log match.
- Telemetry survives video expiry; video availability does not establish telemetry completeness.
damage — Observed health changes
Observed health decreases with recipient side, round, sampled frame, estimated video time and candidate action identifiers from the replay capture.
Source: Validated, retained replay capture report (supported parser version 1 or 2).
Window: One captured replay; at most 240 retained health-change events, paginated 100 per page.
- Health decreases are not confirmed hit counts or causal move attribution. Player denotes the damage recipient.
- Timing assumes 60 render frames per second, not authoritative simulation frames or verified video synchronization.
- Omitted events and truncation are explicit; sums cover retained observations only.
inputs — Observed inputs and round samples
Observed button and chord transitions, per-round input counts and sample counts, health at sampled round boundaries, and estimated timing.
Source: Validated input transitions and round summaries from replay capture reports.
Window: One captured replay; at most 1,800 retained transitions and 100 round summaries, independently paginated.
- Observed inputs do not prove a move occurred, connected, succeeded, or was a distinct decision.
- Counts may include more transitions than the retained event list; omitted input events are reported separately.
- No measured drive gauge, burnout, positions, guard or wake-up state is available.
combos — Inferred combo candidates
Inferred combo sequences, candidate hit counts and damage, estimated move notation and action IDs, with estimated start and end video times.
Source: Validated combo candidates inferred from sampled replay actions and health changes.
Window: One captured replay; at most 80 retained candidates.
- Candidates, hit counts and notation are inferred, not verified combos or patch-matched move data.
- Timing assumes 60 render frames per second. Raw action identifiers are evidence, not authoritative move names.
Publication and search interpretation
Only public product statistics are exported. Authentication, capture jobs, worker state, file paths, tokens, browser activity, feedback and private coaching annotations are excluded. Raw replay samples are not retained in these public reports; unavailable telemetry is not reconstructed.
Dataset metadata describes data and links its JSON distribution. Domain-specific values remain in the tables and JSON. Imported CFN records do not establish site affiliation, so profile pages use WebPage and Person rather than claiming eligibility for Google's ProfilePage feature. MR, LP and match statistics are never review ratings or social interaction counts.
See Google's ProfilePage guidelines, Google's Dataset guidance and Schema.org Dataset. Valid markup and crawler access do not guarantee indexing, rich results or AI citations. No new license to Capcom data is asserted.