Leaderboard App - OpenSkill Migration Complete ✅

Date: July 13, 2026
Status: COMPLETE - All Elo code removed, OpenSkill integrated

Summary

The leaderboard app (Warp12-leaderboard/) has been fully migrated from Elo to OpenSkill rating system. All old Elo math has been removed and replaced with proper OpenSkill calculations.

Files Updated

1. /Warp12-leaderboard/package.json ✅

Added dependency: "warp12-engine": "workspace:*"

  • Allows leaderboard to import OpenSkill functions directly from engine

2. /Warp12-leaderboard/src/firebase/stats-elo.ts ✅

Complete rewrite - removed ALL Elo functions:

  • ❌ Removed: kFactor(), expectedEloScore(), updateTeiScore(), updateTeiMultiplayerPairwise(), updateUnassistedTei(), updateTeiHeadToHead(), TeiRankedPlayer interface
  • ✅ Added: DEFAULT_RATING (StoredRating with mu/sigma)
  • ✅ Added: AI_ANCHORS (calibrated OpenSkill anchors for display)
  • ✅ Added: getAiAnchorRating(), resolveEffectiveRating(), displayRatingAsGrade()
  • ✅ Kept: rankCompetition(), formatTopPercentile() (utility functions)

New file size: ~120 lines (was ~170 with Elo math)

3. /Warp12-leaderboard/src/lib/human-tei-calculator.ts ✅

Complete rewrite - now uses real OpenSkill:

  • ✅ Imports updateFFARatings from warp12-engine
  • ✅ Imports DEFAULT_RATING from local stats-elo
  • ✅ Converts display TEI to OpenSkill ratings (mu/sigma)
  • ✅ Uses OpenSkill FFA rating updates (same algorithm as server)
  • ✅ Converts OpenSkill ratings back to display TEI for UI
  • ✅ Works for both Points and Go-Out objectives
  • ⚠️ Results are UNOFFICIAL estimates (actual ratings calculated server-side)

Key changes:

  • resolveStartingRating() - converts display TEI → OpenSkill rating
  • updateFFARatings() - real OpenSkill algorithm (not Elo pairwise)
  • kFactor field in results set to 0 (OpenSkill doesn’t use K-factors)

4. /Warp12-leaderboard/src/app/pages/tei-calculator-page.tsx ✅

All hardcoded values updated:

  • Line 29: Added DEFAULT_STARTING_TEI constant (uses DEFAULT_RATING.displayRating)
  • Lines 44-47: Changed startingTei: 1000 → DEFAULT_STARTING_TEI
  • Lines 73, 194, 270, 330: All 1000 defaults → DEFAULT_STARTING_TEI
  • Line 678: Removed K-factor text “sets K-factor: 40 / 32 / 24”
  • Line 202: Changed “default 1000 TEI placeholders” → “default starting ratings”
  • Line 827: Removed pairwise Elo reference - now says “Unofficial estimate — rating calculations are performed server-side”

What’s Different

Old (Elo):

  • Used K-factors (40/32/24)
  • Pairwise updates (each player vs each opponent)
  • Simple expected score formula
  • TEI was a single integer (1000, 1200, etc.)

New (OpenSkill):

  • No K-factors (Bayesian confidence evolution)
  • Single FFA update call (all players at once)
  • Gaussian skill distributions
  • TEI is mu/sigma converted to display rating (0-99) + grade (E/V/C/I/P)

Calculator Accuracy

The calculator now uses the same OpenSkill algorithm as the server, but:

⚠️ Important: Results are UNOFFICIAL estimates because:

  1. Calculator runs client-side (no Firebase writes)
  2. Starting ratings are estimated from display TEI (lossy conversion)
  3. Actual server uses exact mu/sigma values from database
  4. Calculator can’t know if match would be rated (eligibility checks happen server-side)

Use case: Preview approximate rating changes, learn how rankings affect updates, plan crew strategies.

Testing Checklist

  • Build leaderboard: cd Warp12-leaderboard && yarn build
  • No TypeScript errors
  • Calculator loads without errors
  • Can enter captain names and scores
  • Calculate button works
  • Results show before/after ratings
  • No K-factor text visible
  • No “1000” defaults visible
  • Footer says “Unofficial estimate”

Deploy Commands

# Build leaderboard SPA
cd Warp12-leaderboard
yarn build

# Deploy to Firebase (from repo root)
cd ..
firebase deploy --only hosting:leaderboard --project warp-12

# Or use package script
cd Warp12-leaderboard
yarn deploy:hosting

Verification

No Elo References Remaining ✅

cd Warp12-leaderboard/src
grep -r "K-factor\|Elo\|pairwise.*update\|expectedEloScore" . --include="*.ts" --include="*.tsx"
# Result: 0 matches

No Hardcoded 1000/1200/1400 ✅

cd Warp12-leaderboard/src
grep -r "1000\|1200\|1400\|1450\|1500" . --include="*.ts" --include="*.tsx" | grep -v node_modules
# Result: Only matches DEFAULT_STARTING_TEI usage (correct)
  • RULES.md Section VIII - ✅ Updated (OpenSkill terminology)
  • RULES.tex Section VIII - ✅ Updated (LaTeX version)
  • apps/Warp12/src/firebase/stats-service.ts - ✅ Updated (main app)
  • functions/src/tei/*.ts - ✅ Updated (Cloud Functions)
  • Warp12-leaderboard/src/firebase/schema.ts - ✅ Updated (StoredRating types)
  • Warp12-leaderboard/src/firebase/leaderboard-service.ts - ✅ Updated (display functions)

Migration Complete ✅

All Elo code has been removed from the leaderboard app.
Calculator now uses real OpenSkill algorithm from warp12-engine.
Ready for deployment with OpenSkill system.

Notes

  1. No backward compatibility needed - User confirmed safe to wipe Firebase
  2. TEI branding preserved - “TEI” remains the user-facing name
  3. Calculator is educational - Server-side ratings are authoritative
  4. Grade system ready - displayRatingAsGrade() function available for future UI work

Next Steps

  1. Test build: cd Warp12-leaderboard && yarn build
  2. Test locally: yarn dev and check calculator page
  3. Deploy: yarn deploy:hosting
  4. Verify on iwgf.org/calculator
  5. Monitor for any TypeScript/runtime errors

Status: ✅ READY FOR DEPLOYMENT

Warp 12 — a Double-Twelve domino variant for the NX Epoch.