]> git.sesse.net Git - wloh/blob - templates/rating.en.xml
Remove password that leaked out into the git repository. (It has also been changed...
[wloh] / templates / rating.en.xml
1 <?xml version="1.0" encoding="UTF-8" ?>
2 <!DOCTYPE
3   html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
4   "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
5 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="no" xmlns:t="http://template.sesse.net/">
6   <head>
7     <title>WLoH rating</title>
8     <link rel="stylesheet" href="style" type="text/css" />
9   </head>
10   <body>
11     <p style="font-size: smaller;" t:id="navbar">
12       <t:languages>
13         <t:separator>::</t:separator>
14         <t:lang-with-link><a /></t:lang-with-link>
15         <t:lang-no-link />
16       </t:languages>
17     </p>
18
19     <h1>WLoH rating</h1>
20
21     <p><em>This is a hobby project from a third party, and not an official part of <a href="http://wordfeud.aasmul.net/">Wordfeud League of Honour</a>.</em></p>
22
23     <p>Even so, the rating is based on play data from WLoH (thanks to Lobotommy for access!), and is updated every hour. It is completely unofficial and has no bearing on WLoH, but is used to estimate win probabilities in <a href="index">the probability calculation</a>.</p>
24
25     <p>The model can change at any time I feel like it :-) Do not take the rating too seriously, even though it is based on relatively reasonable mathematical models. Remember that all statistics say more about the past than about the future.</p>
26
27     <h2>Model parameters</h2>
28
29     <p>For those so inclined. There is also a longer, more detailed <a href="ratings-explained">explanation</a> intended for non-mathematicians.</p>
30
31     <ul>
32       <li>MLE-based model with one scalar (strength) per player and to global scalars (both standard deviations; see below), solved using cyclic MM (minorization-maximization). Number of iterations to convergence: <t:iterations />.</li>
33       <li>Likelihood function, prior: Gaussian distribution with µ=500, σ=<t:rating-prior-stddev /> (est.)</li>
34       <li>Likelihood function, for each match: Gaussian distribution with µ=(score1 - score2), σ=<t:match-stddev /> (est.)</li>
35       <li>Weight: Current and previous three seasons are given full weight (equal to the prior). Older seasons are weighted exponentially less, with half-life of three seasons. Matches stored as 0-0, 150-0, 0-150 eller 150-150 are ignored.</li>
36     </ul>
37
38     <h2>List of divisions</h2>
39
40     <table>
41       <thead>
42         <tr>
43           <th>Div.</th>
44           <th>Average</th>
45           <th>Std.dev</th>
46         </tr>
47       </thead>
48       <tbody t:id="divisions">
49         <tr>
50           <th t:id="rank" />
51           <td class="num" t:id="average" />
52           <td class="num" t:id="stddev" />
53           <t:subdivisions>
54             <td class="num"><a t:id="divlink" /></td>
55           </t:subdivisions>
56         </tr>
57       </tbody>
58     </table>
59
60     <h2>Ranking list</h2>
61   
62     <table>
63       <thead>
64         <tr>
65           <th></th>
66           <th>Nickname</th>
67           <th>Rating</th>
68           <th>Std.dev</th>
69           <th>Last seen</th>
70         </tr>
71       </thead>
72       <tbody t:id="players">
73         <tr>
74           <th t:id="rank" />
75           <td><a t:id="user" /></td>
76           <td class="num" t:id="rating" />
77           <td class="num" t:id="stddev" />
78           <td><a t:id="divlink" /></td>
79         </tr>
80       </tbody>
81     </table>
82
83     <p class="lastsync">The database was last synchronized <t:last-sync />.</p>
84   </body>
85 </html>