elo-ratings 0.0.0

data/.document

@@ -0,0 +1,5 @@
+lib/**/*.rb
+bin/*
+- 
+features/**/*.feature
+LICENSE.txt

data/Gemfile

@@ -0,0 +1,14 @@
+source "http://rubygems.org"
+# Add dependencies required to use your gem here.
+# Example:
+#   gem "activesupport", ">= 2.3.5"
+
+# Add dependencies to develop your gem here.
+# Include everything needed to run rake, tests, features, etc.
+group :development do
+  gem "shoulda", ">= 0"
+  gem "bundler", "~> 1.0.0"
+  gem "jeweler", "~> 1.5.2"
+  gem "rcov", ">= 0"
+  gem "rice"
+end

data/Gemfile.lock

@@ -0,0 +1,22 @@
+GEM
+  remote: http://rubygems.org/
+  specs:
+    git (1.2.5)
+    jeweler (1.5.2)
+      bundler (~> 1.0.0)
+      git (>= 1.2.5)
+      rake
+    rake (0.8.7)
+    rcov (0.9.9)
+    rice (1.4.0)
+    shoulda (2.11.3)
+
+PLATFORMS
+  ruby
+
+DEPENDENCIES
+  bundler (~> 1.0.0)
+  jeweler (~> 1.5.2)
+  rcov
+  rice
+  shoulda

data/LICENSE.txt

@@ -0,0 +1,20 @@
+Copyright (c) 2010 phillc
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

data/README.rdoc

@@ -0,0 +1,23 @@
+= EloRatings
+
+This is a ruby interface to the c++ classes of Remi Coulom's Bayeselo, a tool for estimating Elo ratings.
+
+= Bayeselo
+
+For more information or information on Bayeselo, see http://remi.coulom.free.fr/Bayesian-Elo/
+You can reach the website of Remi Coulom, the author of Bayeselo, at http://remi.coulom.free.fr
+
+Bayeselo is protected under the terms of the GNU GPL. See http://www.gnu.org/copyleft/gpl.html
+It is bundled under the ext directory of this project
+
+= EloRatings
+
+The goal of this gem was to provide a simple interface to Bayeselo in ruby.
+There has currently been no attempt to reproduce all functionality provided by Bayeselo.
+
+
+== Copyright
+
+Copyright (c) 2010 phillc. See LICENSE.txt for
+further details.
+

data/Rakefile

@@ -0,0 +1,54 @@
+require 'rubygems'
+require 'bundler'
+begin
+  Bundler.setup(:default, :development)
+rescue Bundler::BundlerError => e
+  $stderr.puts e.message
+  $stderr.puts "Run `bundle install` to install missing gems"
+  exit e.status_code
+end
+require 'rake'
+
+require 'jeweler'
+Jeweler::Tasks.new do |gem|
+  # gem is a Gem::Specification... see http://docs.rubygems.org/read/chapter/20 for more options
+  gem.name = "elo-ratings"
+  gem.homepage = "http://github.com/phillc/elo-ratings"
+  gem.license = "MIT"
+  gem.summary = "Ruby wrapper for bayeselo"
+  gem.description = "Ruby wrapper for bayeselo"
+  gem.email = "spyyderz@gmail.com"
+  gem.authors = ["phillc"]
+  gem.extensions << "ext/bayeselo/extconf.rb"
+  # Include your dependencies below. Runtime dependencies are required when using your gem,
+  # and development dependencies are only needed for development (ie running rake tasks, tests, etc)
+  #  gem.add_runtime_dependency 'jabber4r', '> 0.1'
+  #  gem.add_development_dependency 'rspec', '> 1.2.3'
+end
+Jeweler::RubygemsDotOrgTasks.new
+
+require 'rake/testtask'
+Rake::TestTask.new(:test) do |test|
+  test.libs << 'lib' << 'test'
+  test.pattern = 'test/**/test_*.rb'
+  test.verbose = true
+end
+
+require 'rcov/rcovtask'
+Rcov::RcovTask.new do |test|
+  test.libs << 'test'
+  test.pattern = 'test/**/test_*.rb'
+  test.verbose = true
+end
+
+task :default => :test
+
+require 'rake/rdoctask'
+Rake::RDocTask.new do |rdoc|
+  version = File.exist?('VERSION') ? File.read('VERSION') : ""
+
+  rdoc.rdoc_dir = 'rdoc'
+  rdoc.title = "elo ratings #{version}"
+  rdoc.rdoc_files.include('README*')
+  rdoc.rdoc_files.include('lib/**/*.rb')
+end

data/VERSION

@@ -0,0 +1 @@
+0.0.0

data/elo-ratings.gemspec

@@ -0,0 +1,148 @@
+# Generated by jeweler
+# DO NOT EDIT THIS FILE DIRECTLY
+# Instead, edit Jeweler::Tasks in Rakefile, and run 'rake gemspec'
+# -*- encoding: utf-8 -*-
+
+Gem::Specification.new do |s|
+  s.name = %q{elo-ratings}
+  s.version = "0.0.0"
+
+  s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
+  s.authors = ["phillc"]
+  s.date = %q{2011-01-19}
+  s.description = %q{Ruby wrapper for bayeselo}
+  s.email = %q{spyyderz@gmail.com}
+  s.extensions = ["ext/bayeselo/extconf.rb", "ext/bayeselo/extconf.rb"]
+  s.extra_rdoc_files = [
+    "LICENSE.txt",
+    "README.rdoc"
+  ]
+  s.files = [
+    ".document",
+    "Gemfile",
+    "Gemfile.lock",
+    "LICENSE.txt",
+    "README.rdoc",
+    "Rakefile",
+    "VERSION",
+    "elo-ratings.gemspec",
+    "ext/bayeselo/.gitignore",
+    "ext/bayeselo/CBradleyTerry.cpp",
+    "ext/bayeselo/CBradleyTerry.h",
+    "ext/bayeselo/CCDistribution.cpp",
+    "ext/bayeselo/CCDistribution.h",
+    "ext/bayeselo/CCDistributionCUI.cpp",
+    "ext/bayeselo/CCDistributionCUI.h",
+    "ext/bayeselo/CCondensedResults.cpp",
+    "ext/bayeselo/CCondensedResults.h",
+    "ext/bayeselo/CDiscretization.h",
+    "ext/bayeselo/CDistribution.cpp",
+    "ext/bayeselo/CDistribution.h",
+    "ext/bayeselo/CDistributionCollection.cpp",
+    "ext/bayeselo/CDistributionCollection.h",
+    "ext/bayeselo/CEloRatingCUI.cpp",
+    "ext/bayeselo/CEloRatingCUI.h",
+    "ext/bayeselo/CIndirectCompare.h",
+    "ext/bayeselo/CJointBayesian.cpp",
+    "ext/bayeselo/CJointBayesian.h",
+    "ext/bayeselo/CLUDecomposition.cpp",
+    "ext/bayeselo/CLUDecomposition.h",
+    "ext/bayeselo/CMatrix.cpp",
+    "ext/bayeselo/CMatrix.h",
+    "ext/bayeselo/CMatrixIO.cpp",
+    "ext/bayeselo/CMatrixIO.h",
+    "ext/bayeselo/CPredictionCUI.cpp",
+    "ext/bayeselo/CPredictionCUI.h",
+    "ext/bayeselo/CResultSet.cpp",
+    "ext/bayeselo/CResultSet.h",
+    "ext/bayeselo/CResultSetCUI.cpp",
+    "ext/bayeselo/CResultSetCUI.h",
+    "ext/bayeselo/CTimeIO.cpp",
+    "ext/bayeselo/CTimeIO.h",
+    "ext/bayeselo/CVector.cpp",
+    "ext/bayeselo/CVector.h",
+    "ext/bayeselo/EloDataFromFile.cpp",
+    "ext/bayeselo/EloDataFromFile.h",
+    "ext/bayeselo/README",
+    "ext/bayeselo/ReadLineToString.cpp",
+    "ext/bayeselo/ReadLineToString.h",
+    "ext/bayeselo/chtime.cpp",
+    "ext/bayeselo/chtime.h",
+    "ext/bayeselo/chtimer.h",
+    "ext/bayeselo/clktimer.cpp",
+    "ext/bayeselo/clktimer.h",
+    "ext/bayeselo/consolui.cpp",
+    "ext/bayeselo/consolui.h",
+    "ext/bayeselo/const.cpp",
+    "ext/bayeselo/const.h",
+    "ext/bayeselo/date.cpp",
+    "ext/bayeselo/date.h",
+    "ext/bayeselo/debug.h",
+    "ext/bayeselo/elomain.cpp",
+    "ext/bayeselo/extconf.rb",
+    "ext/bayeselo/list.h",
+    "ext/bayeselo/listi.h",
+    "ext/bayeselo/move.cpp",
+    "ext/bayeselo/move.h",
+    "ext/bayeselo/pgn.h",
+    "ext/bayeselo/pgnlex.cpp",
+    "ext/bayeselo/pgnlex.h",
+    "ext/bayeselo/pgnstr.cpp",
+    "ext/bayeselo/piece.h",
+    "ext/bayeselo/player.h",
+    "ext/bayeselo/position.h",
+    "ext/bayeselo/random.cpp",
+    "ext/bayeselo/random.h",
+    "ext/bayeselo/rb_bayeselo.cpp",
+    "ext/bayeselo/readstr.cpp",
+    "ext/bayeselo/readstr.h",
+    "ext/bayeselo/square.h",
+    "ext/bayeselo/str.cpp",
+    "ext/bayeselo/str.h",
+    "ext/bayeselo/version.cpp",
+    "ext/bayeselo/version.h",
+    "ext/bayeselo/version_number.h",
+    "lib/bayeselo.rb",
+    "lib/bayeselo/bayeselo.rb",
+    "lib/bayeselo/c_bayeselo.rb",
+    "test/bayeselo/test_bayeselo.rb",
+    "test/bayeselo/test_c_bayeselo.rb",
+    "test/helper.rb"
+  ]
+  s.homepage = %q{http://github.com/phillc/elo-ratings}
+  s.licenses = ["MIT"]
+  s.require_paths = ["lib"]
+  s.rubygems_version = %q{1.3.7}
+  s.summary = %q{Ruby wrapper for bayeselo}
+  s.test_files = [
+    "test/bayeselo/test_bayeselo.rb",
+    "test/bayeselo/test_c_bayeselo.rb",
+    "test/helper.rb"
+  ]
+
+  if s.respond_to? :specification_version then
+    current_version = Gem::Specification::CURRENT_SPECIFICATION_VERSION
+    s.specification_version = 3
+
+    if Gem::Version.new(Gem::VERSION) >= Gem::Version.new('1.2.0') then
+      s.add_development_dependency(%q<shoulda>, [">= 0"])
+      s.add_development_dependency(%q<bundler>, ["~> 1.0.0"])
+      s.add_development_dependency(%q<jeweler>, ["~> 1.5.2"])
+      s.add_development_dependency(%q<rcov>, [">= 0"])
+      s.add_development_dependency(%q<rice>, [">= 0"])
+    else
+      s.add_dependency(%q<shoulda>, [">= 0"])
+      s.add_dependency(%q<bundler>, ["~> 1.0.0"])
+      s.add_dependency(%q<jeweler>, ["~> 1.5.2"])
+      s.add_dependency(%q<rcov>, [">= 0"])
+      s.add_dependency(%q<rice>, [">= 0"])
+    end
+  else
+    s.add_dependency(%q<shoulda>, [">= 0"])
+    s.add_dependency(%q<bundler>, ["~> 1.0.0"])
+    s.add_dependency(%q<jeweler>, ["~> 1.5.2"])
+    s.add_dependency(%q<rcov>, [">= 0"])
+    s.add_dependency(%q<rice>, [">= 0"])
+  end
+end
+

data/ext/bayeselo/.gitignore

@@ -0,0 +1,4 @@
+*.o
+*.bundle
+Makefile
+*.log

data/ext/bayeselo/CBradleyTerry.cpp

@@ -0,0 +1,670 @@
+/////////////////////////////////////////////////////////////////////////////
+//
+// R�mi Coulom
+//
+// December, 2004
+//
+/////////////////////////////////////////////////////////////////////////////
+#include "CBradleyTerry.h"
+#include "CCondensedResults.h"
+#include "CResultSet.h"
+#include "CCDistribution.h"
+#include "CMatrix.h"
+#include "CLUDecomposition.h"
+#include "random.h"
+
+#include <cmath>
+
+#include <iostream>
+#include "CMatrixIO.h"
+
+/////////////////////////////////////////////////////////////////////////////
+// One iteration of the MM algorithm on gammas
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::UpdateGammas()
+{
+ //
+ // Loop over players
+ //
+ for (int Player = crs.GetPlayers(); --Player >= 0;)
+ {
+  double A = 0;
+  double B = 0;
+  for (int j = crs.GetOpponents(Player); --j >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(Player, j);
+
+   double OpponentGamma;
+   if (cr.Opponent > Player)
+    OpponentGamma = pNextGamma[cr.Opponent];
+   else
+    OpponentGamma = pGamma[cr.Opponent];
+
+   A += cr.w_ij + cr.d_ij + cr.l_ji + cr.d_ji;
+   B += (cr.d_ij + cr.w_ij) * ThetaW /
+        (ThetaW * pGamma[Player] + ThetaD * OpponentGamma) +
+        (cr.d_ij + cr.l_ij) * ThetaD * ThetaW /
+        (ThetaD * ThetaW * pGamma[Player] + OpponentGamma) +
+        (cr.d_ji + cr.w_ji) * ThetaD /
+        (ThetaW * OpponentGamma + ThetaD * pGamma[Player]) +
+        (cr.d_ji + cr.l_ji) / 
+        (ThetaD * ThetaW * OpponentGamma + pGamma[Player]);
+  }
+  pNextGamma[Player] = A / B;
+ }
+
+ //
+ // Swap buffers to prepare next iteration
+ //
+ {
+  double *pTemp = pGamma;
+  pGamma = pNextGamma;
+  pNextGamma = pTemp;
+ }
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// MM on ThetaW
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::UpdateThetaW()
+{
+ double Numerator = 0;
+ double Denominator = 0;
+
+ for (int Player = crs.GetPlayers(); --Player >= 0;)
+  for (int j = crs.GetOpponents(Player); --j >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(Player, j);
+   double OpponentGamma = pGamma[cr.Opponent];
+
+   Numerator += cr.w_ij + cr.d_ij;
+   Denominator += (cr.d_ij + cr.w_ij) * pGamma[Player] /
+                  (ThetaW * pGamma[Player] + ThetaD * OpponentGamma) +
+                  (cr.d_ij + cr.l_ij) * ThetaD * pGamma[Player] /
+                  (ThetaD * ThetaW * pGamma[Player] + OpponentGamma);
+  }
+
+ return Numerator / Denominator;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// MM on ThetaD
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::UpdateThetaD()
+{
+ double Numerator = 0;
+ double Denominator = 0;
+
+ for (int Player = crs.GetPlayers(); --Player >= 0;)
+  for (int j = crs.GetOpponents(Player); --j >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(Player, j);
+   double OpponentGamma = pGamma[cr.Opponent];
+
+   Numerator += cr.d_ij;
+   Denominator += (cr.d_ij + cr.w_ij) * OpponentGamma /
+                  (ThetaW * pGamma[Player] + ThetaD * OpponentGamma) +
+                  (cr.d_ij + cr.l_ij) * ThetaW * pGamma[Player] /
+                  (ThetaD * ThetaW * pGamma[Player] + OpponentGamma);
+  }
+
+ double C = Numerator / Denominator;
+
+ return C + std::sqrt(C * C + 1);
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Compute the maximum relative difference between two iterations
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::GetDifference(int n,
+                                    const double *pd1,
+                                    const double *pd2)
+{
+ double Result = 0;
+ for (int i = n; --i >= 0;)
+ {
+  double Diff = std::fabs(pd1[i] - pd2[i]) / (pd1[i] + pd2[i]);
+  if (Diff > Result)
+   Result = Diff;
+ }
+ return Result;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Constructor
+/////////////////////////////////////////////////////////////////////////////
+CBradleyTerry::CBradleyTerry(const CCondensedResults &crsInit):
+ crs(crsInit),
+ velo(crs.GetPlayers()),
+ eloAdvantage(32.8),
+ eloDraw(97.3),
+ v1(crs.GetPlayers()),
+ v2(crs.GetPlayers()),
+ pGamma(&v1[0]),
+ pNextGamma(&v2[0])
+{
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Convert Elos to gammas and Thetas
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::ConvertEloToGamma() const
+{
+ ThetaW = std::pow(10.0, eloAdvantage/400.0);
+ ThetaD = std::pow(10.0, eloDraw/400.0);
+ for (int i = crs.GetPlayers(); --i >= 0;)
+  pGamma[i] = std::pow(10.0, velo[i]/400.0);
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// MM Algorithm
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::MinorizationMaximization(int fThetaW,
+                                             int fThetaD,
+                                             double Epsilon)
+{
+ //
+ // Set initial values
+ //
+ ThetaW = fThetaW ? 1.0 : std::pow(10.0, eloAdvantage/400.0);
+ ThetaD = fThetaD ? 1.0 : std::pow(10.0, eloDraw/400.0);
+ for (int i = crs.GetPlayers(); --i >= 0;)
+  pGamma[i] = 1.0;
+
+ //
+ // Main MM loop
+ //
+ for (int i = 0; i < 10000; i++)
+ {
+  UpdateGammas();
+  double Diff = GetDifference(crs.GetPlayers(), pGamma, pNextGamma);
+
+  if (fThetaW)
+  {
+   double NewThetaW = UpdateThetaW();
+   double ThetaW_Diff = std::fabs(ThetaW - NewThetaW);
+   if (ThetaW_Diff > Diff)
+    Diff = ThetaW_Diff;
+   ThetaW = NewThetaW;
+  }
+
+  if (fThetaD)
+  {
+   double NewThetaD = UpdateThetaD();
+   double ThetaD_Diff = std::fabs(ThetaD - NewThetaD);
+   if (ThetaD_Diff > Diff)
+    Diff = ThetaD_Diff;
+   ThetaD = NewThetaD;
+  }
+
+  if (Diff < Epsilon)
+   break;
+
+  //
+  // Print iteration information
+  //
+#if 1
+  if ((i + 1) % 100 == 0)
+  {
+   std::cout << "Iteration " << i + 1 << ": ";
+   std::cout << Diff << ' ';
+   std::cout << '\n';
+   std::cout.flush();
+  }
+#endif
+ }
+ 
+ //
+ // Convert back to Elos
+ //
+ {
+  double Total = 0;
+  for (int i = crs.GetPlayers(); --i >= 0;)
+   Total += (velo[i] = std::log10(pGamma[i]) * 400);
+  double Offset = - Total / crs.GetPlayers();
+  for (int i = crs.GetPlayers(); --i >= 0;)
+   velo[i] += Offset;
+ }
+
+ if (fThetaW)
+  eloAdvantage = std::log10(ThetaW) * 400;
+
+ if (fThetaD)
+  eloDraw = std::log10(ThetaD) * 400;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Get the likelihood of the results of a player
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::LogLikelihood(int Player) const
+{
+ double Result = 0;
+ for (int i = crs.GetOpponents(Player); --i >= 0;)
+ {
+  const CCondensedResult &cr = crs.GetCondensedResult(Player, i);
+  double Delta = velo[Player] - velo[cr.Opponent];
+  if (cr.w_ij > 0)
+   Result += cr.w_ij * std::log(WinProbability(Delta));
+  if (cr.d_ij > 0)
+   Result += cr.d_ij * std::log(DrawProbability(Delta));
+  if (cr.l_ij > 0)
+   Result += cr.l_ij * std::log(LossProbability(Delta));
+  if (cr.w_ji > 0)
+   Result += cr.w_ji * std::log(WinProbability(-Delta));
+  if (cr.d_ji > 0)
+   Result += cr.d_ji * std::log(DrawProbability(-Delta));
+  if (cr.l_ji > 0)
+   Result += cr.l_ji * std::log(LossProbability(-Delta));
+ }
+ return Result;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Get the likelihood of a result set
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::LogLikelihood(const CResultSet &rs) const
+{
+ double Result = 0;
+
+ for (int i = rs.GetGames(); --i >= 0;)
+ {
+  double Delta = velo[rs.GetWhite(i)] - velo[rs.GetBlack(i)];
+  switch(rs.GetResult(i))
+  {
+   case 0: Result += std::log(LossProbability(Delta)); break;
+   case 1: Result += std::log(DrawProbability(Delta)); break;
+   case 2: Result += std::log(WinProbability(Delta)); break;
+  }
+ }
+
+ return Result;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Get the likelihood of the full set of results
+/////////////////////////////////////////////////////////////////////////////
+double CBradleyTerry::LogLikelihood() const
+{
+ double Result = 0;
+ for (int Player = crs.GetPlayers(); --Player >= 0;)
+  for (int i = crs.GetOpponents(Player); --i >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(Player, i);
+   double Delta = velo[Player] - velo[cr.Opponent];
+   if (cr.w_ij > 0)
+    Result += cr.w_ij * std::log(WinProbability(Delta));
+   if (cr.d_ij > 0)
+    Result += cr.d_ij * std::log(DrawProbability(Delta));
+   if (cr.l_ij > 0)
+    Result += cr.l_ij * std::log(LossProbability(Delta));
+  }
+ return Result;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Get the likelihood distribution of one player
+// The ratings of opponents are supposed to be exact
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::GetPlayerDist(int Player, CCDistribution &cdist) const
+{
+ std::vector<double> veloBackup(velo);
+
+ for (int i = cdist.GetSize(); --i >= 0;)
+ {
+  velo[Player] = cdist.ValueFromIndex(i);
+  if (crs.GetPlayers() > 1)
+  {
+   double Delta = (velo[Player] - veloBackup[Player]) / (crs.GetPlayers() - 1);
+   for (int j = crs.GetPlayers(); --j >= 0;)
+    if (j != Player)
+     velo[j] = veloBackup[j] - Delta;
+  }
+  cdist.SetProbability(i, LogLikelihood(Player));
+ }
+ cdist.LogNormalize(); 
+
+ velo = veloBackup;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Likelihood distribution of the eloAdvantage parameter
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::GetAdvantageDist(CCDistribution &cdist) const
+{
+ double eloPrevious = eloAdvantage;
+ for (int i = cdist.GetSize(); --i >= 0;)
+ {
+  eloAdvantage = cdist.ValueFromIndex(i);
+  cdist.SetProbability(i, LogLikelihood());
+ }
+ cdist.LogNormalize();
+ eloAdvantage = eloPrevious;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Likelihood distribution of the eloDraw parameter
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::GetDrawEloDist(CCDistribution &cdist) const
+{
+ double eloPrevious = eloDraw;
+ for (int i = cdist.GetSize(); --i >= 0;)
+ {
+  eloDraw = cdist.ValueFromIndex(i);
+  cdist.SetProbability(i, LogLikelihood());
+ }
+ cdist.LogNormalize();
+ eloDraw = eloPrevious;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Compute variance with the diagonal of the Hessian
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::GetVariance(double *pdVariance) const
+{
+ ConvertEloToGamma();
+
+ const double x = std::log(10.0) / 400;
+ const double xx = x * x;
+
+ for (int Player = crs.GetPlayers(); --Player >= 0;)
+ {
+  double Diag = 0;
+  double PlayerGamma = pGamma[Player];
+
+  for (int j = crs.GetOpponents(Player); --j >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(Player, j);
+   double OpponentGamma = pGamma[cr.Opponent];
+ 
+   double h = 0;
+
+   {
+    double d = ThetaW * PlayerGamma + ThetaD * OpponentGamma;
+    h += (cr.w_ij + cr.d_ij) / (d * d);
+   }
+   {
+    double d = ThetaD * ThetaW * PlayerGamma + OpponentGamma;
+    h += (cr.l_ij + cr.d_ij) / (d * d);
+   }
+   {
+    double d = ThetaW * OpponentGamma + ThetaD * PlayerGamma;
+    h += (cr.w_ji + cr.d_ji) / (d * d);
+   }
+   {
+    double d = ThetaD * ThetaW * OpponentGamma + PlayerGamma;
+    h += (cr.l_ji + cr.d_ji) / (d * d);
+   }
+
+   h *= PlayerGamma * OpponentGamma * ThetaD * ThetaW;
+   Diag -= h;
+  }
+
+  pdVariance[Player] = -1.0 / (Diag * xx);
+ }
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Compute the covariance matrix
+// This function assumes that ratings are maximum-likelihood ratings
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::ComputeCovariance()
+{
+ //
+ // Compute the truncated opposite of the Hessian
+ //
+ CMatrix mTruncatedHessian(crs.GetPlayers() - 1, crs.GetPlayers() - 1);
+ {
+  ConvertEloToGamma();
+
+  const double x = std::log(10.0) / 400;
+  const double xx = -x * x;
+
+  mTruncatedHessian.Zero();
+
+  for (int Player = crs.GetPlayers() - 1; --Player >= 0;)
+  {
+   double Diag = 0;
+   double PlayerGamma = pGamma[Player];
+
+   for (int j = crs.GetOpponents(Player); --j >= 0;)
+   {
+    const CCondensedResult &cr = crs.GetCondensedResult(Player, j);
+    double OpponentGamma = pGamma[cr.Opponent];
+  
+    double h = 0;
+
+    {
+     double d = ThetaW * PlayerGamma + ThetaD * OpponentGamma;
+     h += (cr.w_ij + cr.d_ij) / (d * d);
+    }
+    {
+     double d = ThetaD * ThetaW * PlayerGamma + OpponentGamma;
+     h += (cr.l_ij + cr.d_ij) / (d * d);
+    }
+    {
+     double d = ThetaW * OpponentGamma + ThetaD * PlayerGamma;
+     h += (cr.w_ji + cr.d_ji) / (d * d);
+    }
+    {
+     double d = ThetaD * ThetaW * OpponentGamma + PlayerGamma;
+     h += (cr.l_ji + cr.d_ji) / (d * d);
+    }
+
+    h *= PlayerGamma * OpponentGamma * ThetaD * ThetaW;
+    Diag -= h;
+    if (cr.Opponent != crs.GetPlayers() - 1)
+     mTruncatedHessian.SetElement(Player, cr.Opponent, h * xx);
+   }
+
+   mTruncatedHessian.SetElement(Player, Player, Diag * xx);
+  }
+ }
+
+ //
+ // LU-Decompose it
+ //
+ CLUDecomposition lud(crs.GetPlayers() - 1);
+ std::vector<int> vIndex(crs.GetPlayers() - 1);
+ lud.Decompose(mTruncatedHessian, &vIndex[0]);
+
+ //
+ // Fill A
+ //
+ CMatrix mA(crs.GetPlayers(), crs.GetPlayers() - 1);
+ {
+  double x = -1.0 / crs.GetPlayers();
+  for (int i = crs.GetPlayers() * (crs.GetPlayers() - 1); --i >= 0;)
+   mA[i] = x;
+  for (int i = crs.GetPlayers() - 1; --i >= 0;)
+   mA.SetElement(i, i, 1.0 + x);
+ }
+
+ //
+ // Compute AC
+ //
+ CMatrix mAC(crs.GetPlayers(), crs.GetPlayers() - 1);
+ for (int i = crs.GetPlayers(); --i >= 0;)
+ {
+  int Index = i * (crs.GetPlayers() - 1);
+  lud.Solve(mTruncatedHessian, &vIndex[0], mA + Index, mAC + Index);
+ }
+
+ //
+ // Compute the covariance
+ //
+ mCovariance.SetProductByTranspose(mAC, mA);
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Compute the likelihood of superiority
+// This function assumes that the covariance matrix and ratings are computed
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::ComputeLikelihoodOfSuperiority()
+{
+ mLOS.SetSize(crs.GetPlayers(), crs.GetPlayers());
+
+ for (int i = mLOS.GetRows(); --i > 0;)
+  for (int j = i; --j >= 0;)
+  {
+   double Sigma2 = mCovariance.GetElement(i, i) +
+                   mCovariance.GetElement(j, j) -
+                   mCovariance.GetElement(i, j) -
+                   mCovariance.GetElement(j, i);
+   double Mu = velo[j] - velo[i];
+   double x = Mu / std::sqrt(2 * Sigma2);
+   mLOS.SetElement(i, j, erfc(x) / 2);
+   mLOS.SetElement(j, i, erfc(-x) / 2);
+  }
+
+ for (int i = mLOS.GetRows(); --i >= 0;)
+  mLOS.SetElement(i, i, 0.0);
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// ELOstat bounding of probabilities
+/////////////////////////////////////////////////////////////////////////////
+static double ELOstatBound(double p)
+{
+ static const double MaxP = 1 / (1 + std::pow(10.0, -600 / double(400)));
+ static const double MinP = 1 / (1 + std::pow(10.0, 600 / double(400)));
+
+ if (p > MaxP)
+  p = MaxP;
+ if (p < MinP)
+  p = MinP;
+
+ return p;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// ELOstat algorithm
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::ELOstat(double Epsilon)
+{
+ double *pElo = &v1[0];
+ double *pNextElo = &v2[0];
+ for (int i = crs.GetPlayers(); --i >= 0;)
+  pElo[i] = 0;
+
+ for (int i = 0; i < 10000; i++)
+ {
+  //
+  // Start by copying ratings
+  //
+  for (int j = crs.GetPlayers(); --j >= 0;)
+   pNextElo[j] = pElo[j];
+
+  //
+  // Loop over players to compute their new ratings
+  //
+  double TotalElo = 0;
+  double GrandTotalGames = 0;
+  for (int j = crs.GetPlayers(); --j >= 0;)
+  {
+   double TotalGames = 0;
+   double TotalScore = 0;
+   double TotalOpponentElo = 0;
+
+   //
+   // Loop over opponents
+   //
+   for (int k = crs.GetOpponents(j); --k >= 0;)
+   {
+    const CCondensedResult &cr = crs.GetCondensedResult(j, k);
+    double Games = cr.Games();
+    TotalGames += Games;
+    TotalScore += cr.Score();
+    TotalOpponentElo += Games * pElo[cr.Opponent];
+   }
+
+   double p = ELOstatBound(TotalScore / TotalGames);
+   double Delta = 400 * std::log10(p / (1 - p));
+   pNextElo[j] = TotalOpponentElo / TotalGames + Delta;
+   TotalElo += TotalGames * pNextElo[j];
+   GrandTotalGames += TotalGames;
+  }
+
+  //
+  // Normalize and swap
+  //
+  for (int j = crs.GetPlayers(); --j >= 0;)
+   pNextElo[j] -= TotalElo / GrandTotalGames;
+  {
+   double *pd = pElo;
+   pElo = pNextElo;
+   pNextElo = pd;
+  }
+
+  //
+  // Compute difference
+  //
+  double MaxDiff = 0;
+  for (int j = crs.GetPlayers(); --j >= 0;)
+  {
+   double Diff = std::fabs(pNextElo[j] - pElo[j]);
+   if (Diff > MaxDiff)
+    MaxDiff = Diff;
+  }
+  if (MaxDiff < Epsilon)
+  {
+   std::cout << i + 1 << " iterations\n";
+   break;
+  }
+ }
+
+ //
+ // Store computed Elo ratings into vElo
+ //
+ for (int i = crs.GetPlayers(); --i >= 0;)
+  velo[i] = pElo[i];
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// ELOstat way of computing confidence intervals
+/////////////////////////////////////////////////////////////////////////////
+void CBradleyTerry::ELOstatIntervals(double *peloLower,
+                                     double *peloUpper) const
+{
+ //
+ // Loop over players to compute confidence intervals
+ //
+ for (int j = crs.GetPlayers(); --j >= 0;)
+ {
+  //
+  // Loop over opponents
+  //
+  double TotalGames = 0;
+  double TotalScore = 0;
+  for (int k = crs.GetOpponents(j); --k >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(j, k);
+   TotalGames += cr.Games();
+   TotalScore += cr.Score();
+  }
+  double MeanScore = TotalScore / TotalGames;
+
+  //
+  // Loop again for variance
+  //
+  double TotalVariance = 0;
+  for (int k = crs.GetOpponents(j); --k >= 0;)
+  {
+   const CCondensedResult &cr = crs.GetCondensedResult(j, k);
+   TotalVariance += (cr.w_ij + cr.l_ji)*(1 - MeanScore)*(1 - MeanScore) +
+                    (cr.d_ij + cr.d_ji)*(0.5 - MeanScore)*(0.5 - MeanScore) +
+                    (cr.w_ji + cr.l_ij)*(0 - MeanScore)*(0 - MeanScore);
+  }
+  if (TotalVariance < 0)
+   TotalVariance = 0;
+  double Variance = TotalVariance / (TotalGames * TotalGames);
+  double StandardDeviation = std::sqrt(Variance);
+  double pLower = ELOstatBound(MeanScore - 1.95996 * StandardDeviation);
+  double pUpper = ELOstatBound(MeanScore + 1.95996 * StandardDeviation);
+  double p = ELOstatBound(MeanScore);
+  double DeltaLower = 400 * std::log10(pLower / (1 - pLower));
+  double DeltaUpper = 400 * std::log10(pUpper / (1 - pUpper));
+  double Delta = 400 * std::log10(p / (1 - p));
+  peloLower[j] = Delta - DeltaLower;
+  peloUpper[j] = DeltaUpper - Delta;
+ }
+}