Devblog: Exoplanets: The Next Phase Of Project Discovery

Ok, I am done with PD.
It was a noble idea with bad data.
Care to explain how this is a transit when even folded the points are all over the place?
There are no dips or even spikes.
Just a big mess.
So if you see something there that is really a transit, tell me how to find it, because it might help everyone improve their analysis and help research.

Until who ever provide the validation of the data can explain HOW they find transit in this case there is no point continuing PD because it is only frustrating.
So either we are not provided with all the data we need (or a way to use them in a meaningful way), or the validation process is garbage.

Personally I go for the later.
So byebye PD, it was not fun while it lasted.

At least The Agency is here now and that will provide a fun experience guaranteed.

i have printscrens of things that need to be repaired in proect discovery, can not upload ā€¦

It is impossible to fix what one cannot see.
Youā€™re gonna have to find someway to upload them to ccp.

Letā€™s see ā€œDouble XP sample of today 0/10ā€. A new day, as marked by downtime.

[quote=ā€œEdlorna_Tinebe, post:177, topic:8026, full:trueā€]
But thatā€™s exactly the job that PD is paying you to do, at 50,000 ISK per sample.[/quote]

Or USD0.000625 per sample.

Hello! I have some ideas that occurred to me while i was trying to find those planets.

  1. If possible, folded view should be a toggle option, whithout having to select a possible transit first. Dont know about others, but for me, i kinda detect possible transits easier that way, rather than having to stretch the timeline out greatly to try to detect the faint transit values that are so difficult to spot and hide between the rest of the zigzags. plus i dont have an ultra widescreen 50inch monitor.

  2. when i scroll through the timeline of the observed star, and it has positive or negative luminosity values that differ greatly from the usual limits, when i scroll the timeline, the current values jump up and down when the spikes appear in the window. The zoom window should consider the highest and lowest possible values on the entire timeline window at a timeline (i hope my description made sense)

I think it would be interesting to add an api for this little game that would allow coders who play eve, or who donā€™t, to hook into the api to pull in these files as images.

Using the data points alone would also be interesting, but I personally believe it would defeat the purpose of this experiment because the whole idea is to find things that they eye notices which an algorithm might not.

So the idea would be, the coders would be able to use the api to request the next graph, as an image, and then they would write code which would try to analyze the image for the transits that we are using our eyes to look for anyway.

I know this is possible because there is code and programs, for example, that are used to count the number of flies on a cow.

Of course, it would be broken if anyone did come up with good code for doing this, so it would probably be important to put a limitation on how many api requests could be made per minute, etc. To the extent that the program would be requesting images from the api much more slowly than a person can click through, but the idea would be that it would also be extremely more accurate, assuming good code implementation.

This is really my only two cents about this thing and to be honest, I am not sure if I even would be able to make code that could do thisā€¦ I just know that it is possible.

Overall, this event is great and I appreciate the fact that CCP seems to be interested in adding content like this to the game.

Cheers!

-Big

My wishlist for Project Discovery:

  • flagging to report abnormal datasets (test set / experimental set)
  • vertical zoom, sometimes the autoscaling fails due to noise peaks
  • remember detrending setting and time frame between runs
  • allow overlay to show numerical addition of segments rather than layered view
  • way to increase (SHIFT) and decrease (CTRL) horizontal shift with cursor
  • setting to resize/remove useless planetary view
  • setting to remove background stars/animation
  • setting to change color scheme
2 Likes

Another strange sample [#200218762] image

Iā€™ve run into a couple WTFs today:

The first almost looks like the data is inverted:

The second is the first time Iā€™ve noticed this marking. Apparently the ā€œcorrectā€ answer is that there are two different planets, but amazingly the both have the same period and are exactly 180-degrees out of phase:

This is way too complicated to figure out!

Maybe hire a game designer who has experience creating fun, intuitive and functional mini games.

I totaly aggree with my prespeakers, the samples at high accuracy become much to dificult, or at least the evaluation makes no sense.
The problem is, I can quite easily keep an accuracy of 70%, without thinking about any of the samples. I just mark anything obvious and pass everything else. Thinking more is time consuming and due to the unsolvable samples just not worth the try.
Like this, the whole scientific approach is obsolete, as high ratings arenā€™t really rewardet appropritly.
EDIT: Maybe there should be some extra bonus for 95% and above, or something like that.

would be nice to have project discovery on EVE Portal app

1 Like

if it was a game, it would be more fun, i guess. Its just wrapped as mini-game, not actually being one, though.

Iā€™m Level 360 at 99% accuracy.

Canā€™t Stop. Wont Stop.

1 Like

Ahoi and hello to all! :slight_smile:

Iā€™m quiet new to EVE and this Project is great! I make 2 videos of searching Exoplanets with up to 99% accuracy rating and high grind factor:

-the first one, I just keep most of transists I found and cut-off the fails - up to Level 20 and 92% Accuracy

-the second video shows how fast possible it could go, to grind ISK (and to have a picture to the computer-generated musicā€¦) - Level 45 to 50 and holding 99% accuracy.

-the 3rd is just the tutorial with all probes positive.

Our youtube-channel (just this Project Discovery Videos there atm):

https://www.youtube.com/channel/UCiUfFlC7WzrRV2NAPnHBDWg

Greetings!

Ohm

(bad english, I guess, sorry therefore!)

Modest suggestion.

Would it be possible to have the name of the star in the samples?
We already get a 9-digit number in the lower right corner when we submit the sample, so changing the number with HD24562+13 wouldnā€™t be too much of a diference.
Some of us nerds would love to search and learn more of this paranoid star that gave us this psycho spectra :slight_smile:

Overall great project.

Wonder how long this is suppose to go on forā€¦Iā€™ve currently set my target level to 1 per day. I really canā€™t stand much more than that

I think they left it out intentionally, to NOT to have people search via google and skew the results.

Well they can do like they do now and not display it until after the slide has been analyzed.

Also, Can we get a leaderboard?