I'm not sure whether this discussion at this point is still specific to ladder ranking or is also inclusive of the calculated rating level for each player seen on the player's profile page. I'm assuming the latter.
And if I am understanding this right conceptually, my calculated playing level is going to be affected day to day by all the matches reported in the entire GTN - is that correct?
If so, I'm wondering whether users might find it more valuable to have their ratings calculation inputs be limited to matches and other players within the specific sub-network(s) that they are members of. I'd argue that this should be the default way to calculate the rating, but maybe there are use-cases that I am not understanding and it should instead be a user-level preference. This is based on the idea that nearly everyone who becomes a member of GTN is doing so for the purpose of connecting and competing with others in their local community.
I know that in my particular little network, everyone who has registered has registered specifically for our particular community site and would be surprised to learn that a match played in Paris was affecting their calculated rating. Most of them don't really even know that the wider GTN exists.
This thread was started for the new Elo System for ranking a ladder. But, a similar Elo System is also used to calculate a player's playing levels (singles and doubles).
Any match played on GTN goes toward your calculated levels as long as all opponents are registered members also.
I think having a calculated playing levels per network could be confusing. Lots of members are in multiple networks. Also, if you went to a profile page from outside a network, what calculated levels would you show?
For people's calculated playing levels, the playing level they sign up with seems to play a big role. For instance, people who sign up as 4.0 need to win something like 30 consecutive matches in order to get up to 4.5. This seems especially strange since many people could sign up equally well as either 4.0 or 4.5 at the start. I think it might be better, once people have played any matches at all, to base their ratings solely on the matches they played, rather than on the playing level they signed up with.
I also think it would be good to base playing levels on just the past five or ten matches that someone played, so that the calculated playing level reflects the current or recent level of the player, and allows for the possibility of players improving their levels. Currently it seems that the calculated playing level mostly reflects the level at which the player signed up (e.g., on my ladder, people who signed up as 4.0's still have lower ratings than people who signed up as 4.5's, even in cases where the 4.0 has a better record and even won their head-to-head match).
The way it works now, is that the lower the accuracy of your playing level, the more it can fluctuate. So, new players could see big swings in their calculated playing levels, which helps them get to a more accurate level quicker.
As a player is less active (plays less matches), their playing level accuracy decreases. So, if someone hasn't played in a long time, their level could move more dramatically when they start playing again. Again, helping them get to their accurate level faster.
The more accurate your playing level becomes, the smaller the changes become. Helping a player stabilize.
Hope this all makes sense. It is pretty complicated stuff.
Thanks for replying, Trevor. Actually this doesn't seem complicated at all. And I still don't see why the level that the player initially signed up at should be involved in the calculated playing level, once the player has started playing matches. If two players play the exact same opponents on the same days, with the exact same scores, then common sense suggests they have the same playing level -- but their calculated playing levels would be wildly different from one another if one player signed up on the ladder as a 4.0 and the other as a 4.5.
The other issue I was pointing out is that if a player plays lots of matches, and steadily improves so he beats better and better opponents, then this improvement will not be reflected in the calculated playing level, since the calculated accuracy will always be high due to the player playing often. What would make more sense is if matches played recently had more of an impact on a player's calculated playing level than matches played a while ago. In other words, the calculated playing level should be a weighted average of the apparent playing levels from various matches, where the weights are bigger for more recent matches and smaller for earlier matches. It sounds like this is done now for just the first few matches after a player takes a long break from playing; but if a player plays steadily then this isn't done.