Preliminary Milestone 3 winners are:

1st: Edward & Willem

2nd: crescendo

To become the (non-preliminary) Milestone 3 winners, they'll need to document their methods, address any questions in the forum, etc. as specified by the rules.

Congratulations to Edward and Willem on the "last minute dash" to grab the 1st milestone 3 spot.

Also congratulations to crescendo on rising so quickly to grab 2nd spot.

I must say though, that I am a bit surprised that those in top 10 seems to be overfitting that much to the 30% I realize that some might decline the milestone prize to avoid revealing their techniques, but surely that is not the case for all who dropped a lot in the milestone 3 leaderboard?

I have done my share of overfitting to training-data but I figured that was just because I am a noob :)

Thats a massive overfit by the top guys! Will Kaggle be publishing the full rank list and scores? Also note the numer of entries by EXL, they might have got locked into hill climbing in the 30% with that many submissions.

Also since.4 is unlikely, the winners might decide to publish any how.

"Will Kaggle be publishing the full rank list and scores?"

Not sure what you mean. We've published the full leaderboard (with private scores) for each Milestone. We'll do so at the end as well for the final scores.

"Will Kaggle be publishing the full rank list and scores?"

Not sure what you mean. We've published the full leaderboard (with private scores) for each Milestone. We'll do so at the end as well for the final scores.

I would also like to see the leaderboard scores (public or private) for each Milestone. As far as I know, Milestone Leaderboard includes only the ranking - http://www.heritagehealthprize.com/c/hhp/leaderboard/milestone3. I don't see any scores in this Milestone Leaderboard. Am I missing something?

Thanks.

Oops! I didn't realize we were only showing the ranks. I'll have to look into this, then.

Yes, we will reveal the Milestone scores at the end.

End as in end of competition or after a few days ?

Wow. What happened? I looked at this comp for a while before moving onto other things. The change in rankings for Round 3 Leaderboard seems substantial.

Is there anyway mere mortals like me can judge whether the difference is driven by people overfitting and flaking out when exposed to the full data set or whether it's because teams deliberately hid their best submissions to avoid having to publish?

End of the competition. 

When will the two Milestone 3 winners papers be available for us to read?
Thanks
Dave

Maybe sooner, but by the rules they have 21 days from when I notified them (which was Sept. 6).