Timothy Yeh <tenton@mac.com> wrote:
| 
| For the JP event, there’s a twitter account that supplies all the numbers

| (although this event, there’s a gap on one night); not sure if that
| would improve the accuracy of your predictions. :P

Recently the predictions are getting worse.. (;_;)
I'm too high for the score match and too low for the normal.
I think people are getting better and more people was waiting
to do the 4x/EX at the end, so the scores really shoot up for the
normal events.

I'll need to update the prediction algorithms.

As for using the hourly data, that would just give me too much noise.
Since the deltas (scores per hour) fluctuate depending on the time
of day, using just a couple data points per day smoothes things out.

| It looks like the scoring is doing the normal pick up near the end.
| I’ve been able to maintain a 3k-4k lead ahead of the 2nd tier pack,
| but that’s gone now. : P I haven’t looked at the JP event close
| enough for trends this time, yet. That was for my weekend (tomorrow). 

Right now I look at the charts comparing the previous same type
events.  That gives the best indication of the final score.

Looking at the last four scorematch events, you can see that
the current Honoka was pretty slow until a couple days ago.
It was at or below Nozomi's event.  Now it's catching up to
Rin's event, which finished above 60k.

  http://www.usagi.org/doi-bin/llcutoff.pl?me=25,27,29,31

--
Hitoshi Doi
                          " this space for rent "
http://www.usagi.org/doi/                                    doi@usagi.org