openskill 0.1.0 → 1.0.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: a7fcde3a56dfbdd99df6a10cfdebf1fb95c0b2b0fe21513792a3e3103e0606f1
-  data.tar.gz: be466ac8799195ba2a0353ac2bc94be71857c6c27d824c37db59d1ab088c13ca
+  metadata.gz: b2559ed1bbd59204cee9143bc49fc51ce1c0048abc5a84b03ba4793d437a4d31
+  data.tar.gz: 692090c1a809b7645f8acb0bd10be874ec01a32ae2d8d4177d1eacfebfbc041b
 SHA512:
-  metadata.gz: 9d3c96ae581d28403b828e90788582b06a6d77c7c6a418dfbe1b98abf0e164abdfcf4b1b28151a486a2ed7ab0ab902f7803fca4eb7861ca27c5550ab73c2e4a7
-  data.tar.gz: 867ce922496ccac80c0ea67195d2bf339afb253c7d52d73dbfd2c830848e9d762591b182e725f9e19262d06d209415b7b49921dffdb19b8fee92e8a12dca3903
+  metadata.gz: 1533d76307745a38d728ba4a0a35d093f0270cca0e36b77bc21d276bfa250107f10668260748bebd886fc2177e33001479297839e3890ba158235d2216a8c8d8
+  data.tar.gz: '09866e5514cac8a47f4534c404871cd8e15a15b8d94e4d00fe9b6c31cc907852c31ed3f2cd4b1320cd38636df49a5968085de62a1c2580171adb27a9853fef63'

data/README.md

@@ -2,7 +2,8 @@
 
 A Ruby implementation of the OpenSkill rating system for multiplayer games. OpenSkill is a Bayesian skill rating system that can handle teams of varying sizes, asymmetric matches, and complex game scenarios.
 
-[![Tests](https://img.shields.io/badge/tests-passing-brightgreen.svg)]()
+[![Gem Version](https://badge.fury.io/rb/openskill.svg)](https://badge.fury.io/rb/openskill)
+![](https://github.com/erdostom/openskill-ruby/actions/workflows/test.yml/badge.svg)
 [![Ruby](https://img.shields.io/badge/ruby-%3E%3D%203.1.0-red.svg)]()
 [![License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 
@@ -15,6 +16,7 @@ A Ruby implementation of the OpenSkill rating system for multiplayer games. Open
 - 🔢 **Player Weights**: Account for partial participation or contribution
 - 📈 **Score Margins**: Factor in impressive wins
 - 🔄 **Tie Handling**: Properly handle drawn matches
+- 🎲 **5 Rating Models**: PlackettLuce, BradleyTerryFull, BradleyTerryPart, ThurstoneMostellerFull, ThurstoneMostellerPart
 - ⚡ **Fast**: Efficient Ruby implementation
 - 🧪 **Well Tested**: Comprehensive test suite matching reference implementation
 
@@ -38,14 +40,32 @@ Or install it yourself as:
 gem install openskill
 ```
 
+## Available Models
+
+OpenSkill Ruby includes 5 rating models from the Weng-Lin family:
+
+- **PlackettLuce** - Multidimensional model, recommended default (Algorithm 4)
+- **BradleyTerryFull** - Full pairing with logistic regression (Algorithm 1)
+- **BradleyTerryPart** - Partial pairing with sliding window, more efficient (Algorithm 2)
+- **ThurstoneMostellerFull** - Full pairing with Gaussian CDF (Algorithm 3)
+- **ThurstoneMostellerPart** - Partial pairing with Gaussian CDF, most efficient
+
+All models support the same API and features. Choose based on your accuracy vs performance needs.
+
 ## Quick Start
 
 ```ruby
 require 'openskill'
 
-# Create a model (Plackett-Luce by default)
+# Create a model (PlackettLuce recommended)
 model = OpenSkill::Models::PlackettLuce.new
 
+# Or use other models:
+# model = OpenSkill::Models::BradleyTerryFull.new
+# model = OpenSkill::Models::BradleyTerryPart.new(window_size: 4)
+# model = OpenSkill::Models::ThurstoneMostellerFull.new(epsilon: 0.1)
+# model = OpenSkill::Models::ThurstoneMostellerPart.new(epsilon: 0.1, window_size: 4)
+
 # Create player ratings
 alice = model.create_rating(name: "Alice")
 bob = model.create_rating(name: "Bob")
@@ -239,6 +259,7 @@ OpenSkill uses a Bayesian approach to model player skill as a normal distributio
 - **σ (sigma)**: The uncertainty about the skill level
 
 After each match:
+
 1. Compute team strengths from individual player ratings
 2. Calculate expected outcomes based on team strengths
 3. Update ratings based on actual vs expected performance
@@ -249,12 +270,14 @@ The **ordinal** value (`μ - 3σ`) provides a conservative estimate where the tr
 ## Why OpenSkill?
 
 ### vs Elo
+
 - ✅ Handles multiplayer (3+ players/teams)
 - ✅ Works with team games
 - ✅ Accounts for rating uncertainty
 - ✅ Faster convergence to true skill
 
 ### vs TrueSkill
+
 - ✅ Open source (MIT license)
 - ✅ Faster computation
 - ✅ Similar accuracy
@@ -264,14 +287,14 @@ The **ordinal** value (`μ - 3σ`) provides a conservative estimate where the tr
 
 This Ruby implementation uses idiomatic Ruby naming conventions:
 
-| Python API | Ruby API |
-|------------|----------|
-| `model.rating()` | `model.create_rating` |
-| `model.create_rating([25, 8.3])` | `model.load_rating([25, 8.3])` |
-| `model.rate(teams)` | `model.calculate_ratings(teams)` |
-| `model.predict_win(teams)` | `model.predict_win_probability(teams)` |
-| `model.predict_draw(teams)` | `model.predict_draw_probability(teams)` |
-| `model.predict_rank(teams)` | `model.predict_rank_probability(teams)` |
+| Python API                       | Ruby API                                |
+| -------------------------------- | --------------------------------------- |
+| `model.rating()`                 | `model.create_rating`                   |
+| `model.create_rating([25, 8.3])` | `model.load_rating([25, 8.3])`          |
+| `model.rate(teams)`              | `model.calculate_ratings(teams)`        |
+| `model.predict_win(teams)`       | `model.predict_win_probability(teams)`  |
+| `model.predict_draw(teams)`      | `model.predict_draw_probability(teams)` |
+| `model.predict_rank(teams)`      | `model.predict_rank_probability(teams)` |
 
 ## Examples
 
@@ -312,23 +335,23 @@ updated = model.calculate_ratings(
 ```ruby
 class Player
   attr_accessor :name, :mu, :sigma
-  
+
   def initialize(name, model)
     @name = name
     rating = model.create_rating
     @mu = rating.mu
     @sigma = rating.sigma
   end
-  
+
   def to_rating(model)
     model.load_rating([@mu, @sigma], name: @name)
   end
-  
+
   def update_from_rating!(rating)
     @mu = rating.mu
     @sigma = rating.sigma
   end
-  
+
   def ordinal(z: 3.0)
     @mu - z * @sigma
   end