Quick Summary
- Rating = A number representing a wrestler's skill level
- Rating Deviation (RD) = How certain we are about the rating
- Starting rating = 1500 (average)
- Win = Gain points • Lose = Lose points
- Beat someone higher rated = Gain MORE points, RD decreases
What is Glicko-2?
The Glicko-2 rating system was developed by Professor Mark Glickman at Boston University as an improvement to the classic Elo system. It's used by many competitive platforms including chess servers and now MatMetrics for wrestling.
The key innovation is rating deviation (RD): a measure of how confident we are in a wrestler's rating. A wrestler with many recent matches has a low RD (we're confident), while a wrestler who hasn't competed recently has a high RD (less certain).
Why Glicko-2 is Better
- ✓Tracks uncertainty: Ratings are more reliable with RD
- ✓Handles inactivity: RD increases over time without matches
- ✓Better predictions: Confidence intervals improve matchups
- ✓Faster convergence: New wrestlers find their true rating quicker
Glicko-2 vs Elo
The original Elo system was invented by Arpad Elo in the 1960s for chess. While it works well, Glicko-2 addresses several limitations:
| Feature | Rating | Glicko-2 |
|---|---|---|
| Rating | Single number | Rating + RD + Volatility |
| Uncertainty | Not tracked | RD shows confidence |
| Inactivity | No change | RD increases |
| New wrestlers | Volatile ratings | High RD indicates uncertainty |
MatMetrics Transition
MatMetrics originally used Elo ratings but has transitioned to Glicko-2 as the primary rating system. You'll see both ratings during the transition, with Glicko-2 being used for all new calculations and rankings.
Rating Confidence Levels
The Rating Deviation (RD) tells you how confident we are in a wrestler's rating. Lower RD means more reliable ratings.
Many recent matches. Rating is very reliable and stable. The true skill is likely within ±120 of the displayed rating.
Some recent activity. Rating is reasonably reliable but may shift with more matches. The true skill is likely within ±300 of the displayed rating.
Limited recent matches or new wrestler. Rating may change significantly with more competition. Treat this rating as a rough estimate.
Reading Your Rating
Here's what different rating ranges generally mean:
| Rating | Level | Description |
|---|---|---|
| 2000+ | Elite | State champion caliber, nationally competitive |
| 1800-1999 | Excellent | State placer, top of weight class |
| 1600-1799 | Strong | Competitive varsity, regional contender |
| 1400-1599 | Average | Solid competitor, wins some/loses some |
| 1200-1399 | Developing | Building skills, limited experience |
| <1200 | Beginner | New to competition, learning fundamentals |
Important: These are guidelines, not guarantees. Always check the confidence level. A 1600-rated wrestler with High confidence is more reliable than a 1700-rated wrestler with Low confidence.
Using Ratings for Predictions
Rating differences translate to win probabilities. This is exactly how MatchIQ works.
Rating Difference → Win Probability
Glicko-2 improves predictions by considering RD. When both wrestlers have High confidence ratings, predictions are more reliable. When one or both have Low confidence, the prediction accounts for greater uncertainty.
Limitations
While Glicko-2 is a significant improvement over Elo, no rating system is perfect:
Doesn't Account for Injuries
A wrestler returning from injury might have a rating that doesn't reflect their current condition. RD helps somewhat by increasing during inactivity.
Style Matchups Matter
A Scrambler might consistently beat Technical wrestlers but lose to Grinders, regardless of ratings. Ratings don't capture this—but archetypes do!
Historical, Not Predictive of Growth
Ratings reflect past performance. They don't know that a freshman will improve dramatically over the season.