scoring for calibration movies

Discuss your experiences with and ideas about Stardust@home here.

Moderators: Stardust@home Team, DustMods

Post Reply
entropydave
Posts: 35
Joined: Wed Aug 02, 2006 11:28 pm

scoring for calibration movies

Post by entropydave »

Hmmm... I just noticed that it takes ages for the sensitivity to creep up for each correct ID of the calibration movies (in my case at least 4-5 correct ones for an increase of 0.01%) so I decided to get one wrong to check out something I noticed before - I lose 0.06% for one incorrect one.
Surely this isn't right - if you need 4 or 5 to creep up a 100th of 1 percent then one wrong should not cause one to drop so much.
Any explanations or am I missing the bleedin' obvious?!


thanks!
dave
"I have the most excellent proof that x^3=y^3+z^3, however my signature is too small to contain it..."
fjgiie
DustMod
Posts: 1253
Joined: Sat May 20, 2006 8:47 am
Location: Hampton, SC, US

Post by fjgiie »

From "Definitions" we have:
Sensitivity is the total number of tracks you have correctly identified in calibration movies divided by the total number of calibration movies you have searched in which there were actually tracks.
All calibration movies correct = 920 , 460 is one half
Calibration movies incorrect = 120 , 60 is one half
We will divide three groups of numbers - sensitivity score only

correct 460 + incorrect 60 = searched 520

460 / 520 X 100% = 88.46 %

(miss one) 460 / 521 X 100% = 88.29 % |OR| (get one correct) 461 / 521 X 100% = 88.48 %
minus .17% on a miss, and plus .02% on a correct one.

what else,

fjgiie
Post Reply