When you connect to this website, you send your IP address and sometimes some cookies. You may also give us personal identifying information, such as your name and contact information. All this data is used to securely provide you with the services that you request. We encourage you to review our privacy policy to make sure that you understand how your data is managed, and to contact us if you have any questions. View Privacy Policy

Difference between revisions of "Rating system overview"

From NASPAWiki

You are viewing a condensed mobile version of this NASPA webpage.
Switch to full version.

(Added logistic formula)
Line 14: Line 14:
  
 
3. Your win-loss record (ties count as a half a win.)
 
3. Your win-loss record (ties count as a half a win.)
 +
 +
<div style="float:right; width:20em; border: 1px dotted black; padding: 0.25em; margin:0.25em;font-size:80%">
 +
How exactly are Expected Wins calculated?
 +
They are calculated fractionally for each opponent, then added up.
 +
If the between your rating and your opponent&rsquo;s rating is ''d'',
 +
you are expected to win 1&divide;(1+''e''<sup>0.0031879&middot;''d''</sup>) games against them.
 +
</div>
  
 
The rating program first calculates a number to estimate how many games you will probably win.  This is known as your &ldquo;Expected Wins.&rdquo;  The rating program uses your intial rating and the ratings of all your opponents to perform this calculation.  It then subtracts your Expected Wins from your actual number of wins to get the difference.  It then multiplies the difference by your K value (see below) to calculate your rating change.  If you won fewer games than what was expected, then the difference will be negative meaning you lose rating points.  If you won more games than expected, then you will gain rating points.   
 
The rating program first calculates a number to estimate how many games you will probably win.  This is known as your &ldquo;Expected Wins.&rdquo;  The rating program uses your intial rating and the ratings of all your opponents to perform this calculation.  It then subtracts your Expected Wins from your actual number of wins to get the difference.  It then multiplies the difference by your K value (see below) to calculate your rating change.  If you won fewer games than what was expected, then the difference will be negative meaning you lose rating points.  If you won more games than expected, then you will gain rating points.   

Revision as of 13:32, 24 April 2018

This rating system overview describes how the rating system works at a very high (and not too technical) level.

Each tournament player has a numerical rating. This rating is most commonly used to assign players to their appropriate divisions in a tournament. The rating is also used, in part, to determine which players qualify for certain tournaments, for example the World SCRABBLE Championship.

The obvious question is "How is the rating calculated? (Don’t forget, this is supposed to be a non-technical article, so don't throw in any of that fancy math stuff!)"

Ok, here goes:

You have just completed a tournament so we know several things:

1. Your rating going into the tournament (determining your initial rating is complex, so I will leave that to a different page),

2. The ratings of your opponents, and

3. Your win-loss record (ties count as a half a win.)

How exactly are Expected Wins calculated? They are calculated fractionally for each opponent, then added up. If the between your rating and your opponent’s rating is d, you are expected to win 1÷(1+e0.0031879·d) games against them.

The rating program first calculates a number to estimate how many games you will probably win. This is known as your “Expected Wins.” The rating program uses your intial rating and the ratings of all your opponents to perform this calculation. It then subtracts your Expected Wins from your actual number of wins to get the difference. It then multiplies the difference by your K value (see below) to calculate your rating change. If you won fewer games than what was expected, then the difference will be negative meaning you lose rating points. If you won more games than expected, then you will gain rating points.

The K value, or sometimes called the Multiplier, is usually 20. However players with less than 50 tournament games will have a higher K value, the idea being this helps accelerate them to their “proper” rating. Also, once a player's rating goes above 1800 or 2000, their K value drops to 16 or 10, respectively.

The rating system also implements Acceleration and Feedback points. If someone does exceedingly well - that is gains a lot of rating points - they are awarded Acceleration points and their opponents are awarded Feedback points. The underlying assumption in this case is that the player was “underrated” going into the tournament and they should have been rated higher. Hence the Acceleration points. Furthermore, because they were underrated, their opponents probably lost more (or didn't gain as many) points than they would have, hence the Feedback points.

This leads to the common question “Is my rating an indication of how good a Scrabble player I am?”

In a pure mathematical sense, your rating is essentially your overall winning percentage weighted by your opponents’ weighted winning percentages. It does not directly measure your word knowledge, board vision, time and rack management skills or other attributes required to be a good Scrabble player.

On the other hand, the object is to win the game. The better players will tend to win more games, and the ratings reflect that. Remember that Scrabble has a certain amount of inherent randomness, so in any given game, the lower rated player has a chance. And upsets do happen.

I hope this article gave you some insight into how the rating system works. I plan on publishing additional, more technical articles in the future. If there is something specific you would like to see documented, or have questions / comments, please feel free to drop me a note.

Jim Hughes

Chair, Ratings and Recognition Committee