Re: Aggregate of fastest times each day
I really like the idea - as somebody who rather enjoys the challenge of the timed puzzles, it would add to that. If it was calculated using ratios, the numbers jpoos put up seem relatively close to what the ratios seem to be from looking at the top scores. For example, the ratio of best scores of the top 50 for 4x4/5x5/6x6/8x8 is 1.0/2.6/5.4/26.5 - though the top 8 is a bit harder to assess since it includes only subscribers.
The big challenge I see with ratios is the effect of a small change. If the 4x4 is given 20 times the weight of an 8x8 puzzle, then it is far more likely the day's results will be determined by who gets a good (or lucky) 4x4 score since a 1 second change on the 4x4 is equivalent to a 20 second change in the 8x8. A 1s change on a smaller puzzle should be scored better than on a big puzzle, but I would also say a 1% change on a smaller puzzle should have less effect than on a larger puzzle.
Thinking on the original comparison to the decathlon, I looked up how the decathlon is actually scored:
https://en.wikipedia.org/wiki/Decathlon#Points_systemQuote:
Points = INT(A(B — P)^C)
A, B, and C are parameters for the event and P is the athlete's score for the event. The parameters are presumably defined to roughly equate the events and the range of abilities seen in the field. Potentially, historical timed puzzle data could be used to calculate appropriate values of A, B, and C to equalize performance across the 4 events (the calcudoku quadrathlon?) There could also be a triathlon version for non-subscribers.
It might also be interesting to have one-day, one-week, and one-month events. Especially with the 8x8 only having one per day, this might smooth out the chance of one bad day (or, add a 2nd 8x8 for sucscribers.)