I haven't checked into how it is calculated... if it is calculated by the rank someone held when the game started or when the game finished.

On ToS I couldn't help noticing that if you had multiple 1v1 games going it was beneficial to your ranking to "lose" all your losing games, then to win your winning games by stretching out the turns you take in winning games and taking losing game's turns quickly (potential for minor amounts of abuse).

Slightly different point - I was in a game that started in October.  Since being in that game, and since being eliminated from that game, my ranking has gone up a good bit.  It's been dragging on and when I "officially" lose that game I know my ranking will drop. 

Should the amount of ranking points to be lost be calculated when:

A) Game is started.

B) You are eliminated.  (Calculated when you are eliminated, not deducted when you are eliminated)

C) Game is completed. 

Just some food for thought.

-Andernut.

I think you should start the game with all the cards on the table. Use the scores from the beginning, and beat the players that you started playing with. 

That way, you can join games knowing what your going to end them getting. If you take a year to win it, so be it, but you still the same points you would have when you started. 

Eyes wide open.

agreed

It is calculated from the point where the game completed.  To determine this I looked at my last 3 "You lost/won the game" emails, looked at the ranking numbers, and noted that I was eliminated from the last one before the other two finished, yet the number that was displayed in my calculation for the last one was the number I was at after the previous one.

If I could change the ranking system, I'd switch to something based on the http://en.wikipedia.org/wiki/Glicko_rating_system so that active players would have ratings that converge at some level of skill.  My rating fluctuates in what looks like a 300-400 point range, and I assume that others do as well.

There was a previous forum post many a moon ago that I thought I brought up asking whether you could game the system by doing exactly as you proposed.  I can't for the life of me find the post now though.  Ozy has some incredible skill at resurrecting dead posts for future discussions, so maybe he'll find it.

Found it:  http://www.wargear.net/forum/showthread/1064p1/Tournament_tiebreaker

The question was in regards to tournament record, but the same could be applied to normal rankings.

Although it looks like Tom's workaround for tournaments was that rankings are recalculated on a per-round basis as the tourney progresses, meaning you can't game the tournament by playing fast or slow.

This doesn't fix the normal rankings issue being discussed however.

BorisTheFrugal wrote:

 Ozy has some incredible skill at resurrecting dead posts for future discussions, so maybe he'll find it.

I always thought that was more of a Yertle thing.  I just do a google search with "site:wargear.net/forum"

There was a previous forum post many a moon ago that I thought I brought up asking whether you could game the system by doing exactly as you proposed.

I do remember a discussion(s?) about ranking methods though.  Hugh had figured out some neat way to predict where someones ranking would eventually settle based upon games played so far, and I thought that would be an awesome addition to the many stats we already have here.

Ozyman wrote:

BorisTheFrugal wrote:

 Ozy has some incredible skill at resurrecting dead posts for future discussions, so maybe he'll find it.

I always thought that was more of a Yertle thing.  I just do a google search with "site:wargear.net/forum"

ooo, so simple, yet i never thought of it. would that work with subforum threads? that could be highly useful.

I think this is the thread you are talking about:

http://www.wargear.net/forum/showthread/1905/Gaming_the_Score

Ozyman wrote:

BorisTheFrugal wrote:

 Ozy has some incredible skill at resurrecting dead posts for future discussions, so maybe he'll find it.

I always thought that was more of a Yertle thing.  I just do a google search with "site:wargear.net/forum"

 

There was a previous forum post many a moon ago that I thought I brought up asking whether you could game the system by doing exactly as you proposed.

I do remember a discussion(s?) about ranking methods though.  Hugh had figured out some neat way to predict where someones ranking would eventually settle based upon games played so far, and I thought that would be an awesome addition to the many stats we already have here.

I had to find it, so http://www.wargear.net/forum/showthread/474/Ratings_Equilibrium is the discussion.

I found it by searching through http://www.wargear.net/forum/listthreads//2/?viewtype=active&filter=Hugh for likely looking threads, found one where Hugh linked to it, and jumped there.

weathertop wrote:

Ozyman wrote:

BorisTheFrugal wrote:

 Ozy has some incredible skill at resurrecting dead posts for future discussions, so maybe he'll find it.

I always thought that was more of a Yertle thing.  I just do a google search with "site:wargear.net/forum"

ooo, so simple, yet i never thought of it. would that work with subforum threads? that could be highly useful.

 

I think it will search all the of forums (subforums?).  For example:

https://www.google.com/search?q=site%3Awargear.net%2Fforum+scores

In fact the search button on wargear is supposed to do this for you, but it wasn't working right.  There was some discussion here:

https://www.google.com/search?q=site%3Awargear.net%2Fforum+scores

Tom said he needed to do something to fix it, but I don't know if that thing ever got done.

Ozy's second link doesn't point to the right place (I think he made a cut/paste error, as it matches his first link.  The one he's trying to point to is my post about searching for baseball having some funky results:

http://www.wargear.net/forum/showthread/3125/Forum_search_function

Thanks Boris!

i was thinking more like, "i know it was in this particular forum, so i want to limit to only that one" like: 

"site:http://www.wargear.net/forum/listthreads/Board_Designers_Chat + whatever"

I think there's arguments for both sides - in theory your ranking starts from 1000 and stabilises at a score reflecting your ability. Hence if you take your ranking score at the end of a game it is more likely to be closer to your true ranking.

Also taking starting ranking scores offers a way to game the system - if a new board is released, it is to the advantage of a good player to start as many games as possible whilst his/her ranking is 1000 as they will then get the maximum benefit from points won in those games.

Or another way of looking at it - it's swings and roundabouts, some times it will benefit you other times it won't.

Actually in both theory and practice your ranking cannot stabilize.  It has to swing over a surprisingly broad range.

That is why the glicko system that I linked to above keeps track of both a score and an estimate of how uncertain that score is.  If it thinks that it is certain about your score, then you stand to gain/lose less per game.  If uncertain, you gain/lose more.  The more you play, the more certain it is.  But as time goes on it becomes less certain.

Just one more hidden parameter, but the chess sites that I've seen use it come up with extremely consistent scores.  And if I see that a player is 100 points stronger than me, it is clear which is different.  (By contrast my score here has swung 300 points in the last couple of months, and I think that is mostly due to chance.)

Regarding the original question: In principle, you have more information about a player at the end of a game, so that's when it should be calculated. Gaming the scores using the losing quick / winning slow technique offers a smaller gain than someone getting their score calculated from the 1000 spot for many games.

Regarding the side conversation: Glicko is excellent. Conan pointed me to Trueskill, which looks really interesting. I have many criticisms/analsyses of the current calculation, though I've never collected them into a single post. When Conan and I were discussing this, I noticed an interesting feature that even Elo (which Glicko expands upon) has that this system doesn't. This feature explains why equilibrium is so hard to reach for good players.  Most locals are happy with the current calculation, so such conversations end up being largely irrelevant!! (The math is fun at least...)

Hugh wrote:

 Most locals are happy with the current calculation,..

Not me.  Unfortunately, Global Rankings seem pretty firmly entrenched.   I would prefer to see something that is less volatile, yet indicative of current ability.

M57 wrote:

Hugh wrote:

 Most locals are happy with the current calculation,..

Not me.  Unfortunately, Global Rankings seem pretty firmly entrenched.   I would prefer to see something that is less volatile, yet indicative of current ability.

Not me neither. It has been discussed many times without any success, but a 40 point cap (instead of a crazy 100) and a moving average would be great improvments.

The problem with a 40 point cap is that people who have played only a few games will have wildly incorrect ratings.  A constant cap always has this problem.  A large cap means that active players will never have their rating settle down.  A small cap means that people who have just played a few games have wildly incorrect ratings.

That is why the glicko system has a cap that varies based on how active you've been recently.