Skip to content
大相撲 SumoFans
夏場所 Natsu (Summer)
methodology · caveats · last build June 3, 2026 at 1:52 PM

注釈 Caveats & assumptions

What's an estimate, what's bounded, what this site deliberately doesn't try to be. The fine print, in one place.

01 / Assumptions & estimates

What's an approximation

Money

Where Number Source / reasoning
F5 / Salaries 150 ¥ / 1 USD Approximate market rate. Conversions rounded for readability. Real rate fluctuates; treat USD as ±10%.
F5 / Stipends Lower divisions have per-basho stipends (¥165k/¥110k/¥88k/¥77k for Makushita/Sandanme/Jonidan/Jonokuchi) JSA published rates. Annual figure is stipend × 6 bashos.

Time

Where Number Source / reasoning
P6 / Calendar grid Basho windows hard-coded to 2026 dates. Jungyō windows estimated JSA publishes basho dates for the calendar year. Jungyō tour days are estimated based on typical post-basho gaps.
P6 / Match time ~6 seconds avg per bout Public timing analyses of Makuuchi matches. Most end in seconds; a ~30s match is unusually long.
P6 / Year-of-rikishi hours Sleep 9h, training 5h, meals 2h, chores 2.5h, bath 1.5h, etc. Composite of public stable schedules + media coverage. Approximate - lower-ranked rikishi do far more chores; yokozuna do more events. Per-day numbers × 365 (approximate).

Attendance

Where Number Source / reasoning
F6 / Attendance Capacity = listed venue capacity. 15-day total = capacity × 15 × ~80-100% utilization Tokyo bashos typically sell out; regional (Osaka/Nagoya/Kyushu) average ~80-90% across 15 days. Estimates only.
T1 / Live banner - attendance so far Capacity × utilization × liveDay. Utilization: 97% Tokyo, 85% regional "~X have come through the gates so far" on the live banner is a multiplicative estimate, not actual gate counts. Tokyo (Hatsu/Natsu/Aki) typically sells out at Ryogoku; Osaka/Nagoya/Kyushu run lower on average across 15 days.

Bodies & roster

Where Number Source / reasoning
P5 / Bodies Heights and weights are self-reported to the JSA Official figures, but updated infrequently. A rikishi's actual current weight may drift before the next official update.
P1 / Athlete cards Country flags use shusshinCountry; defaults to 🇯🇵 if unknown Limited country list - additions made as new origins appear in the data.

Methodology choices

Where Number Source / reasoning
P2 / Career leaders Win % filtered to ≥100 matches Cuts out small-sample noise (early-career rikishi). Adjustable in scripts/build-data.ts.
P3 / Stables - "best last basho" Filtered to stables with ≥10 bouts in the last basho Avoids tiny stables with one absent wrestler appearing as "100% winner".
T1 / Yusho race leaderboard Top 3 rikishi sorted by best W-L; tie-broken by fewer absences → fewer losses → higher banzuke rank Rank-delta arrows use position within Makuuchi, not raw rankValue (which is encoded as tier × 100 + offset). Δ = previous_position − current_position.
02 / Caveats & known limitations

What this site is not

  • Static snapshot. The site is rebuilt from derived.json at deploy time. Live tournaments only update when someone re-runs npm run data and pushes. There's no client-side polling.
  • Makuuchi-only focus. Most analyses (banzuke board, athletes, kimarite-by-tier, etc.) are restricted to the top division. Lower divisions appear only in 04 (Six divisions) and 15 (Bodies).
  • Window-bounded. Kimarite frequencies, last-basho records, and outcome Sankey use a 6-basho rolling window (~1 year). Career stats are full lifetime. Deeper history (1958-present, 67 years) lives at /history.
  • Photo coverage. Wikipedia is tried first (CC BY-SA), with the JSA's official profile page as fallback. Active sekitori coverage is now ~100%; profile-page coverage across the full historical Makuuchi corpus (558 rikishi since 1958) is 55% - famous yokozuna and most post-2000 wrestlers resolve via Wikipedia, while low-rank one-tournament wrestlers (mostly pre-2000) with no Wikipedia article and no current JSA profile fall through to a 力 silhouette card. New entrants may take a basho or two to appear in either source.
  • Salary numbers exclude bonuses. Yusho (¥10M / ~$67K), sansho (¥2M / ~$13K), kensho (sponsor envelopes, ~¥18K / $120 net per match), and zenshō-yūshō (perfect record, additional ¥5M) are not included in the base-pay figures.
  • Time-of-year approximations. The 365-day calendar grid in section 16 is a typical year, not the exact current year's basho dates. Actual jungyō schedules vary.
  • Long-tail kimarite. The 82 official kimarite include rare ones (e.g. watashikomi, nichonage). New or extremely rare kimarite may show up without an English gloss until added to the lookup tables.
  • Currency. All ¥ values are JPY; all $ values are USD at ~150 ¥/$. Treat USD as approximate, not exchange-rate-precise.