One rating · every discipline · zero conflict of interest
ISRSO maintains a universal, mathematically fair skill rating standard. We are a pure standard organization — like FIDE for chess or ISO for engineering. We certify platforms that host events; we do not host events ourselves. This separation is structural, not preferential.
Universal Standard
Tennis, badminton, chess, Go, pickleball: one ladder, one star rule, one rating math for every discipline. Per-sport calibration of the score lines comes in phase 3.
Glicko-2 + Elo
10 placement matches on Glicko-2, then Elo, the core FIDE has used for decades. The rating counts only win or loss: beat a stronger player and you gain more, lose to a weaker one and you drop more. The set score only affects stars.
Tiers Never Drop by Losing
Tiers are won in a promotion window and kept. Divisions can drop, with protection after each step up. Only a cheating sanction, or a season change that re-lands you by rating, can lower a tier.
Standard-Only Mandate
ISRSO publishes the algorithm + the certification spec. Tournament hosting, prize pools, broadcasting — all delegated to certified platforms. Zero financial interest in any specific event.
Where ISRSO ends · where platforms begin
A hard line — drawn on purpose. The standard organization stays out of commerce so its rulings are above suspicion. Platforms carry the operational load and earn the revenue.
- 📐 Publish the rating algorithm + tier specification
- 👑 Mint & maintain the King Roster
- 🛡 Audit + certify platforms (Tier A / B / C)
- ⚖️ Arbitrate cross-platform disputes
- 📊 Maintain rating baselines and score lines
- 🔬 Publish anti-fraud detection specifications
- 🎯 Host tournaments + leagues + prize pools
- 📺 Broadcasting + media coverage + sponsorships
- 👥 Athlete onboarding + coach matching
- 🔁 Day-to-day matchmaking + results submission
- 🛠 First-line dispute resolution
- 💼 Commerce, ticketing, merchandising
5 tiers × 5 divisions · then King
Each tier floor is a rating score line set by rarity. Each tier is roughly 10× rarer than the one below. King has no divisions; its stars simply add up.
A struck medal for every tier
The tier is told by the crest and its metal. Your progress inside the tier is told by the medal itself: notches, gem sockets, ribbon and halo.
Anatomy · Gold II, 2 of 5 stars, promotion window open
- 1Outer notchesStars of the current division (Bronze / Silver 4, Gold / Diamond 5, Master 6). A star lights up the moment you earn it.
- 25 gem socketsOne per division. Reached divisions hold a lit gem; the rest stay empty.
- 3CrestTier identity. Changes only when you move up a tier.
- 4RibbonReads “Gold II”; flips over when the division changes.
- 5FinishShows rating confidence (RD). Rough during placement, clearer with each match, slightly fogged after a long break.
- 6HaloOnly during a promotion window: 5 arcs (9 for King), one per counted match.
King has no divisions: the notches are replaced by a “★N” inscription. Others see a small medal next to your avatar on match cards, entry lists, draws, rankings and your profile; a gold ring around it means a promotion window is open.
10 placement matches, then Elo
The rating keeps it fair; stars and the medal make progress visible. Both phases share one scale (anchor 1500, 400-point log scale), so nothing is converted at the hand-over.
Glicko-2
- Start: rating 1500 · RD 250 · win or loss only
- Standard Glicko-2 update, no per-match point cap
- No stars yet
Placed
- Rating → division by score line
- Stars = position inside that division
- Above 2450 → King ★0
Elo
ΔR = K × w × (S − E)- K = 32 for the first 30 placed matches, then 24
- Winner never loses rating, loser never gains
Two values from one match: S for rating, S_set for stars
| Best of 5 | Best of 3 | Stars only: winner S_set / loser S_set |
|---|---|---|
| 3:0 | 2:0 | 1.0 / 0 |
| 3:1 | — | 0.9 / 0.1 |
| 3:2 | 2:1 | 0.75 / 0.25 |
| Single game / win-loss only | 1 / 0 | |
Rating and the promotion-window performance rating use only win = 1 / loss = 0. Feeding the set score into rating would squeeze the gaps between players to about 0.8 of the truth. The format comes from each sport's scoring rules; other formats (e.g. best of 7) interpolate linearly by the winner's share of games.
What counts
| Official event | w = 1.0 |
|---|---|
| Major event | w = 1.25 |
| Casual / self-organised | not rated |
| Placed vs. still placing | placed player's Δ × 0.5 |
| Format | one-on-one only |
| Sanctioned (frozen) | Δ = 0 |
Stars follow how you actually played
Not a flat +1 / −1. Stars depend on d = S_set − E: the set-score value minus your pre-match win chance. The rating counts only win or loss; the set score only affects stars. Stars come only from matches.
| Result | Condition | ★ |
|---|---|---|
| Win | d ≥ 0.75 | +3 |
| Win | 0.5 ≤ d < 0.75 | +2 |
| Win | d < 0.5 | +1 |
| Loss · 虽败犹荣 | d ≥ −0.1 | 0 |
| Loss | −0.25 ≤ d < −0.1 | −1 |
| Loss | d < −0.25 | −2 |
| + Upset | win with pre-match chance < 30% | +1 |
| + Major event | win at a major event | +1 |
| Cap | per match, bonuses included: Bronze / Silver +3, others +4 (so 0★ can never promote in one match) | +3 / +4 |
Strength line: if your shown division trails your rating by 2+ divisions, a win gives +1★ extra and a loss costs nothing. If it leads by 2+, a win gives 1★ less (minimum +1). The cap still applies.
Up and downALWAYS ON
- Full stars → up one division; extra stars carry over
- Lose a star at 0★ → down one division, landing on full − 1
- Tiers never drop by losing · V at 0★ is the floor
- New-division protection · the first two losses after moving up never drop you
- Top division full → no more stars; a win opens a promotion window only if your rating is within 50 of the next tier floor (next scene)
The only legitimate ways to a lower tier: an anti-cheat sanction, or a season change that re-lands you by rating.
Crossing a tier takes a performance, not a streak
Like a chess norm: over a window of matches, your performance rating (from wins and losses) must reach the next tier's floor. Tournament draws can't rob you, because strong opponents count for more.
| Opens | At division I with full stars, a win that would overflow opens the window only if rating ≥ next tier floor − 50; otherwise stars stay full and no window opens |
|---|---|
| Length | ≥ 5 rated matches (King: ≥ 9, with ≥ 3 opponents rated ≥ 2250) · at most 30 days or 3 events |
| During | Stars frozen, rating still moves · live “3/5 · PR 1612” |
| Settles | When the event ends, never mid-event |
| Success | PR ≥ next floor (1600 / 1800 / 2050 / 2250 / 2450) → next tier V, 0★, crowned |
| Fail | Stay at division I with half the stars (rounded down) |
| Expired | Too few matches in time → void, not a fail; stars stay full, the next win reopens it. A season change also voids it (rating carries over). |
Worked exampleGOLD I → DIAMOND
Gold I, 5/5★, rating 2012. Diamond floor is 2050, so the gate is 2050 − 50 = 2000: 2012 ≥ 2000, the next win opens the window. (At rating 1990 the stars would stay full and nothing would open.) Target: 2050. The next tournament gives 6 matches.
| # | Opponent R | Result | S |
|---|---|---|---|
| 1 | 1990 | Win | 1 |
| 2 | 2120 | Loss | 0 |
| 3 | 2060 | Win | 1 |
| 4 | 2010 | Win | 1 |
| 5 | 2180 | Loss | 0 |
| 6 | 2040 | Win | 1 |
| avg 2067 | Σ 4 |
Σ E(PR, Rᵢ) = Σ Sᵢ = 4
PR ≈ 2191
6 matches ≥ 5 and 2191 ≥ 2050 → Diamond V · 0★, crowned when the event ends.
PR uses win/loss only, so the set scores never enter it. Had PR come out at 2030: stay Gold I with ⌊5 / 2⌋ = 2★. If every match is won (or lost), PR = strongest opponent + 400 (or weakest − 400).
Every match visibly changes something
Medal-minting craft: each tier is a metal, and the motion is striking, setting stars, setting gems and crowning. It plays only when something happens.
Stakes shown in advance
A gold strip on every ranked match card, from each player's side. In a window it reads “Window 3/5 · PR 1612”. Casual matches show nothing.
Settlement, about 2.5 s, skippable
- +N★notches light one by one
- ▲gem drops in, ribbon flips
- 0★“虽败犹荣” seal, notches still
- −1★one notch quietly dims
- ▼gem fades, protection rule shown
Its own moment
A gold “爆冷” stamp, an edge flash, then the stars. A shareable poster is generated and the result is posted to the same event and club.
Halo on, arc by arc
Entering a window lights the medal's halo and your avatar frame. Each counted match lights one arc, its depth showing how well you played. A failed window lets the halo fade: “PR 1580 / 1600 · half your stars kept”.
New tier, struck
On success the old medal dissolves into light and the new tier's medal is struck, with the strongest haptic. The event and the club are notified.
Crowned in the hall
A landscape event screen cast to the venue TV. During the event it rotates players in a window (name, medal, x/5, PR). After the results it plays each crowning in turn.
With “reduce motion” on, the final state is shown directly. Haptics follow system settings. No sound effects. Text on metal meets contrast requirements.
3 seasons / year · rating carries over
Three seasons a year gives every discipline time for meaningful matches. At each boundary your rating carries over untouched; only the stars are placed again by rating. Nothing is pulled back or reset.
Season boundary: rating carries over, stars re-land
R' = R
Your division and stars are placed again from R, as after placement; your shown division is the lower of the old one and the landing spot. Confidence (RD) is clamped to 150–350.
Stars are the only thing that moves: rating is never pulled toward the middle. Past-season tiers are archived permanently, so your S3 Diamond I stays a verifiable record.
A name you can wear · forever
Every King promotion mints a permanent number with a season stamp, an issue index and a discipline tag. It holds across disciplines and seasons and is never reissued, like a retired jersey number.
Numbering charter
#00007 can only ever refer to one person.#00347 in tennis and it is also yours in piano if you ever reach King there. One identity, many domains.#00333), palindromes (#01210) and single digits (#00001 – #00009) carry collector-grade prestige.Time off never costs a tier
A break does not change your tier, division or stars. It only lowers how sure the system is about your rating (RD), and the medal shows that as a light fog.
Placing
Rough, unfinished medal. It gets clearer with every match.
Playing
Polished medal. RD shrinks after every match.
On a break
Light fog on the medal. Tier, division and stars unchanged.
Back
Fog clears as RD converges. No re-placement needed.
Stars only come from matches
No paid protection
Stars and divisions cannot be bought or insured.
No tier for time served
Seniority, streaks or hours played do not move the rating.
No log-in stars
No daily rewards, no hidden score. What you see on the ladder is the rule.
Trust is the standard
A rating system worth following is one you can defend. ISRSO publishes its integrity rules publicly and in full — detection layers, appeal pathways, sanction ladder. How the records themselves can be checked is the next chapter. No hidden clauses in the rules; no shortcut to a verdict.
Anti-fraud · 4 detection layers
Statistical Anomaly
Rolling 30-day R-curve outlier detection (Z-score >3σ), win-rate spike alerts, sub-second move-time analysis per discipline. Flags surface within 24h.
Device + IP Fingerprint
Cross-account collusion detection. Smurf / alt account linkage via hardware signature. Shared-device tournament play flagged for human review.
Match Recording Audit
1% random sampling of ranked matches + 100% of King promotion-window matches reviewed by 3rd-party referees within 72h.
Unified ISRSO UUID
One identity per person — cross-platform, cross-discipline. Smurfs caught at registration via ID verification; biometric option for Tier A+ platforms.
Dispute resolution · the 4-step path
File at platform
Within 72h of match end. Platform freezes the disputed rating change pending review.
Platform investigation
Resolution within 7 days. Cleared, reversed, or escalated to ISRSO.
ISRSO Arbitration
Cross-platform / unresolved cases. Committee reviews match recording, anti-cheat logs, witness statements.
Verdict + 1 appeal
Final verdict within 30 days, published in the ISRSO Public Record. One appeal allowed, reviewed by independent panel.
Sanctions · 5-step graduated ladder
Warning
Profile flag · no rating impact. Visible internally for 90 days.
Match invalidation
Disputed match expunged · rating, stars and windows replayed without it.
Tier rollback
−1 small division. Repeated infractions within a season.
Suspension
30 / 90 / 365-day cooldown across all certified platforms.
Permanent Ban
Cross-discipline removal · public Disqualified Registry · King number retired.
Sanction closed loop — in production since 2026-07. A verdict automatically reverses the rating, rebuilds the tier state and notifies the federation. Under v2 a voided match or upheld appeal replays the athlete's matches in order, rating, stars and promotion windows included, with the same result every time.
Don’t take our word for it
Trust should not rest on what we say about ourselves. It rests on three things: open rules, records that cannot be rewritten, and results that can be recomputed. We are not asking anyone to vouch for us first. Instead, the system is built so that if the platform touches a record it shouldn’t, it leaves a trace that can be caught — and the records about you are yours to check.
Rules in the open
Every rule is published and numbered by version. You know how points are counted before you play.
Records that can’t be rewritten
What has been written stays written. Mistakes are fixed by adding a new line, never by erasing an old one.
Results that can be recomputed
The same record and the same rules always give the same numbers, so the math can be redone and compared.
The trust ledger is built and is being switched on. Until then, this chapter describes how it will work — not a service that is already running.
Four steps, from match to proof
Record
Each rated match, correction and platform action is written down, in order.
Seal
The ledger is sealed at regular intervals. Anyone can keep a copy of the seal.
Recompute
Same ledger, same rule version, same result — checked with an independent tool.
Confirm
Players and organisers can sign the score on their own phone.
Eight ideas behind it
A ledger that only moves forward
Every rated match, every void and correction, and every back-office action by the platform goes into one public ledger, in order. Entries can only be added — never edited, never deleted. Think of a bank statement, or a notary’s register.
Sealed regularly, copies for everyone
Whenever the ledger grows, the platform stamps it with a seal every 10 minutes, plus one more each day. Seals are public and anyone may keep one. If history were quietly rewritten later, the new seal would no longer match the old one, and anyone who kept an earlier seal could spot it. A seal carries only a fingerprint of the records — no names, no scores.
Rules first, every calculation labelled
All rating rules are public and numbered by version, and every calculation states which version it used. Changing a rule means publishing a new version and recording it in the ledger; it applies from then on, while past matches keep the version they were counted under. If the running settings ever drift from the published version, the system refuses to record.
Same ledger, same rules, same answer
We provide an independent recompute tool: an auditor with the audit data can rerun every rating from the start under the public rules and compare it, line by line, with the platform’s results. The ratings the platform seals are themselves recomputed from the ledger; if anyone changed back-office data directly, the two would disagree, and the system would refuse to seal and lock until a person checks. There is no back-office button to set a score directly.
Mistakes are corrected, never erased
Voiding a match, changing a score, ruling on an appeal — each one adds a new entry, and the original stays forever. Every back-office action by the platform must name who did it and why. Who changed what, when and why can always be traced.
Your confirmation, your key
Players and organisers can sign the score using the security chip in their own phone, unlocked with a fingerprint or face. The key is created inside the phone and cannot be exported, so the platform never holds it. Adding or removing a signing device is recorded in the ledger; your phone checks this and warns you if an unfamiliar device appears. Signing never changes a rating — it is an extra piece of evidence.
Check your own record
In the app you can check that your own rating and your signing devices are written into the sealed ledger. Only when the check passes does your profile show ‘rating verifiable’; if an unfamiliar device is added to your account, the home screen warns you clearly.
Test and real kept apart
Events from test organisations and matches marked as tests are never rated and never enter the ledger. The test list is publicly searchable, so you can see whether any of your matches was marked as a test, and when.
Our limits — we don’t say ‘absolute’
An honest system starts by saying what it cannot do. On the left: what you will be able to check yourself once the ledger is on. On the right: what, for now, still rests on trusting the platform.
- The seal really comes from the platform and is not a forgery.
- If you kept an earlier seal: the ledger has only grown since, with no history rewritten.
- Your own rating and signing devices are written into the sealed ledger.
- Your own matches and scores — match details are visible only to the people involved.
- Whether any of your matches was marked as a test.
- Auditors with the audit data: every rating, recomputed from the start.
- Whether a match really took place, and who organised it — the platform’s record is what counts.
- Nothing left out: the ledger proves that what was written hasn’t changed, not that everything was written. You can notice a missing match of your own; others cannot.
- History from before the ledger is switched on: entered once by the platform as the starting point.
- A signature shows that a phone registered to your account signed. If someone else used your account, or a parent signed for a child, it is not you in person.
- The app itself, and the two calculation programs that cross-check each other, are all built by us.
- Ordinary users can check only their own share; recomputing everyone needs the audit data.
Next: make the right-hand column shorter. External copies and third-party timestamps are being introduced: seals will also be kept by someone outside the platform, and a third party will stamp them with the time. Joint witnessing by organisations comes after that. Our own rule: until external copies and timestamps are in place, we will not claim the seals are in effect.
What makes records trustworthy is being public, unchangeable and recomputable — not mining or tokens. We use the public-ledger approach long proven by web security certificates and financial auditing: cheaper, faster and greener.
The math under the medallion
The rating is the source of truth; divisions, stars and windows are derived from it. Everything is recorded per match, so any result can be replayed and audited.
Match confirmed
A certified platform reports both players, the result, the set score and the event type (official / major / casual).
Rating updated
Glicko-2 during placement, Elo afterwards. S is win or loss only; E comes from both ratings before the match.
Stars & division
d = S_set − E sets the stars; bonuses and the strength line apply; divisions move with protection rules.
Window settled
When the event ends, open windows are judged by performance rating (win/loss only); crownings go to the app and the venue screen.
(R, RD) = Glicko2( R, RD, R_opp, RD_opp, S ) start R = 1500, RD = 250 · S = win / loss · no per-match cap
# Ranked · Elo
E = 1 / ( 1 + 10^((R_opp − R) / 400) )
ΔR = K × w × ( S − E ) K = 32 (first 30) → 24 · S = 1 / 0 · w = 1.0 official / 1.25 major
# winner ΔR ≥ 0, loser ΔR ≤ 0 · placed vs placing: placed ΔR × 0.5
# Stars
d = S_set − E → +3 (≥ 0.75) / +2 (≥ 0.5) / +1 · loss 0 (≥ −0.1) / −1 / −2 (< −0.25) · upset +1 · major +1 · cap +3 (Bronze / Silver) / +4
# Promotion window · performance rating
Σ E( PR, R_opp_i ) = Σ S_i S_i = win / loss · opens only if R ≥ next floor − 50 · success if PR ≥ next tier floor
# Season rollover · rating carries over
R' = R → re-place division + stars by rating
# Sanctions
FROZEN → ΔR = 0 · VOID / upheld appeal → replay the athlete's matches in order (rating, stars, windows)
SUSPEND / BAN → propagated to the federation via signed outbound webhook
Three committees · one charter · pinned versions
A standard worth following is one you can read, version, and challenge. ISRSO operates under three independent committees, a public founding charter, and semantically-versioned releases. The standard evolves — but never silently.
Governance bodies
Algorithm Committee
7 members · chess / sports / eSports / academia, 3-year rotating terms.
Maintains rating calibration, the K table, the rarity score lines and the star rules; per-sport calibration in phase 3.
Arbitration Committee
5 members · legal + technical backgrounds, independent of any certified platform.
Final word on cross-platform disputes, sanction appeals, and integrity escalations. Decisions published in the Public Record.
Certification Board
Independent audit professionals · annual rotation.
Audits applicant platforms; grants and revokes Tier A / B / C status. Publishes quarterly conformance reports.
Version timeline
Where ISRSO is administered
Platforms pass the ISRSO conformance audit — algorithm fidelity, anti-fraud, badge rendering, public leaderboard — to issue ratings. ISRSO does not host events; platforms do. Below: the four ecosystem platforms certified at the time of this writing.
GrowingX
The training management OS for academies. Eight-module SaaS that runs schools and writes ratings into the ISRSO ledger.
G-Match
Full-stack tournament operations — formats, seeding, scoring, anti-fraud, escrowed prize pools. Every cleared match writes to ISRSO.
XCoach
AI-powered training paired with ISRSO ratings for skill progression feedback.
PickBoll
Pickleball match results feed live into ISRSO with full anti-fraud verification.