Would it be of any use to predict the most probable next game where it will be your turn? 

Background: Sometimes I find myself without a current turn, but wanting to play. In these situations I sometimes browse through my games planning my next moves. At some point I realized you could actually make a prediction in which game it will most likely be your turn next by looking at the average turn time for each player in that game.

On the other hand, this information is by nature of pure statistical value only and has no possibility of actually predicting your next turn. 

Is anybody else interested in this statistic?

@Tom: On a side note, does the algorithm that calculates the average turn time for each player adjust for outliers (http://en.wikipedia.org/wiki/Outlier), or does it simply calculate they plain average?

It's interesting but I think ineffective.  I have games that I'll leave for several days if not feeling motivated to take the turn (ie. Simulgear in a complex map against a formidable opponent), I have games I'll take seconds apart when playing real-time games, and I may check in 2 or 3 times a day, but my "average" turn time is spread across all games.

Some people have an average turn time of under 10 minutes due to majority of their games being real-time, but they may join a game with 3-day boot-time and miss their turn because they play real-time games on the week-ends and sporadically log in throughout the week.

Perhaps "average turn time" as a statistic should ignore real-time games.

Maybe median instead of mean.

Andernut wrote:

I have games I'll take seconds apart when playing real-time games, and I may check in 2 or 3 times a day, but my "average" turn time is spread across all games.

If outliers were removed from the average turn time that should compensate for the effect of those complex games.

Andernut wrote:

Perhaps "average turn time" as a statistic should ignore real-time games.

It may be overkill, but it shouldn't be hard to calculate different average turn times for the different possible turn times.

2 average turn time would make sense

http://www.wargear.net/forum/showthread/2115/Another_average_time_calculation

Andernut wrote:

It's interesting but I think ineffective. 

The average turn time would only be one variable that influences 'Expected Next Turn'. The other two will be player number and turn timer.

Player Number:
Let's say every player had an average turn time of 5 hours, then the expected next turn would predict 15h for a 4 player game and 2 days for a 10 player game - that's a significant difference. Regardless of the absolute value of these predictions, the relative predictions (as in which game will be first) could be fairly accurate.

Turn Timer:
If the algorithm used separate values for the different turn timers with removed outliers I suppose even the predictions for the absolute values could be quite meaningful.

Advanced:
If this were a professional research project, you'd also take into account two more statistics:
(a) The player's turn-taking probability distribution: If you have enough data about a player, you can predict at what time(s) of day he is most likely to take his turn.
(b) Whether the player is currently online or not. 
I could imagine this variant produced surprisingly accurate predictions.

The advanced version obviously would be more of a bachelor thesis project because it requires quite a bit of tweaking. But the lighter version should be quite easily implemented and could produce fairly accurate results.