On Leaderboards we now support three very nice headers:
Details
Pre-game predicted outcome
Post-game predicted outcome
On the latter two we list the Trueskill Mu, which is the expected performance of a player in a given game (as distinct from their rating with is that less 3 times the TruesKill Standard Deviation)
TrueSkill is not very obvious in how it works, to most people and a common experience is to wonder why ratings changed as they did. We added those Pre-game and Post-game predictions to help folk understand that. I think we can improve on this in a few ways:
Support another option which annotates session details with some more info, namely for eac player their Mu (Predicted performance) and the difference in their Mu and the player beneath them in the ranking, as a measure of the expected tension. The idea here is to help use see how the calibre of the people you beat impacts your change in rating! For that we might want to see the deltas between all other players maybe as a ToolTip list, where 0 is in a list of deltas and represents the player, and the ones to left the deltas of those players that beat them and to the right those that they beat. Could be a ToolTip, could be added as a column? Main deal is we have some way of easily seeing it.
Currently on Pre and Post game predictions we just list Mu, perhaps in the Tooltipl we should list the rating and the Mu/Delta pair? Or perhaps again extra option to request showing rating and/or confidence as columns in that header (or parenthesised data)
Consider adding Mu and/or SD options for display as columns in the leaderboard table itself!
If we're feeling really fancy we should have a quick way to sort an existing leaderboard by Mu rather than Rating, or toggle between them. To see what the predicted rankings would be quickly based on existing ratings,
I can't help but feel this will make https://github.com/bernd-wechner/CoGs/issues/13 all the more important, as we begin to see the impact that low play frequency has on inflated ratings (due to early successes against well rated players)
I think the aim in general to help us see the kind of odds we overcame to rank as we did or that others beat to rank above us, and how that interplays with the Mu in particular and to keep the MU/SD pair more visible in these diagnostic or learning drill downs. They currently appear as tooltips in the leaderboard rankings.
On Leaderboards we now support three very nice headers:
On the latter two we list the Trueskill Mu, which is the expected performance of a player in a given game (as distinct from their rating with is that less 3 times the TruesKill Standard Deviation)
TrueSkill is not very obvious in how it works, to most people and a common experience is to wonder why ratings changed as they did. We added those Pre-game and Post-game predictions to help folk understand that. I think we can improve on this in a few ways:
Support another option which annotates session details with some more info, namely for eac player their Mu (Predicted performance) and the difference in their Mu and the player beneath them in the ranking, as a measure of the expected tension. The idea here is to help use see how the calibre of the people you beat impacts your change in rating! For that we might want to see the deltas between all other players maybe as a ToolTip list, where 0 is in a list of deltas and represents the player, and the ones to left the deltas of those players that beat them and to the right those that they beat. Could be a ToolTip, could be added as a column? Main deal is we have some way of easily seeing it.
Currently on Pre and Post game predictions we just list Mu, perhaps in the Tooltipl we should list the rating and the Mu/Delta pair? Or perhaps again extra option to request showing rating and/or confidence as columns in that header (or parenthesised data)
Consider adding Mu and/or SD options for display as columns in the leaderboard table itself!
If we're feeling really fancy we should have a quick way to sort an existing leaderboard by Mu rather than Rating, or toggle between them. To see what the predicted rankings would be quickly based on existing ratings,
I can't help but feel this will make https://github.com/bernd-wechner/CoGs/issues/13 all the more important, as we begin to see the impact that low play frequency has on inflated ratings (due to early successes against well rated players)
I think the aim in general to help us see the kind of odds we overcame to rank as we did or that others beat to rank above us, and how that interplays with the Mu in particular and to keep the MU/SD pair more visible in these diagnostic or learning drill downs. They currently appear as tooltips in the leaderboard rankings.