These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

EVE General Discussion

 
  • Topic is locked indefinitely.
 

Analysis Failed - Feedback to Project Discovery

Author
Caseidae
Seniorenresidenz
Funatix
#1 - 2017-07-18 21:21:56 UTC  |  Edited by: Caseidae
o/,
after some hard days here is some of my feedback. Maybe its just me being too stupid in that case your flame will let me know ...


1) Your tutorial is too short for the difficult ones. I would need to see much more examples to train my brainish neural network.

2) If an test analysis fails I want to have the chance to get back to it, zoom in and try to find out why I have missed it !! Currently its just like "beeep - loser" & "next". Do I need a photographic memory or screenshots to compare what I have seen to what the real solution was? Makes me so furious!

3) I had it serveral times that I have "found" too many drops - Analysis Failed. Are you afraid of false positives? I also have sometimes that I have found the transit but with a multiple period, especially for high frequency transits. Is that soo wrong? You still have the transit, do you?

4) Detrend is nice. But isnt it just a low pass filter? How about also having a high pass filter to detrend the high frequency flicker !? If we detrend down a few hours - do we really see transits of higher period still? Does it make sense to look for transits in the range of hours?
I would really love to see the sample filtering out high frequencies. I really doubt we checking transits in the range of a few hours.

5) Your test samples are either ridiculous easy or unfair. You give my either a Grand Canyon drop or its a high frequency transit in a flicker time series. At the later I always think how I shall I ever find that?

I have the seen the Twitch Broadcast "I am now looking for 6 month to it ..." he said. Tbh Do you expect us the same learning curve? Then its way underpaid. Tell me please, what is the amount of time one should spent a flicker curve to find a difficult drop?

6) What does "xx% marked transits" mean? xx% of my fellow players have marked transits - how is that a criteria? My transits match to xx% of what other have found?

I think the current player strategies are:
- rely on the "transits are rare" - always no transits unless its obvious und try not to fall for the tests
- mark every drop as single or as multiple transits - that gives me the impression if I see the consensus results and also try not to fall for test.

7) My final thing to say: If we have a 27d window and transit orbital period larger than that - you miss the periods to confirm it. But is then not every drop a potential transit? How can you deny this unless you can give a good description how a transit drop looks like. And if so, why not giving this description?
2 drops but not giving a period within the sample can be two transits. How can you confirm or deny it? I am not accepting the "its totally rare scenario" - unless you accept it not being considered by the players because of its rareness.


My personal impression:
I would love to do it. If I would see a chance to fulfil the expectations but currently I am not. I am still guessing. I am still totally surprised by the correct results of a difficult test sample.

Regards.

CCP: dont check this char for discovery status, played with Alt.
Malphas Vynneve
Aliastra
Gallente Federation
#2 - 2017-07-18 22:48:27 UTC
I understand your frustrations and had similar complaints during my first few analysis after the tutorial.

I'm no expert at it yet and i'm sure there are more people at a much higher rank, but i'm currently at Analyst 13. Went from like Analyst 4 to 13 in a combined matter of two or three hours. I would never be able to explain how i "see" indications of transits or a lack thereof without sitting down with you and talking about it with PD up in front of us, but my best advice to give you is to just blaze through a couple by hitting no transit and pay attention to the ones that say fail, then you'll start to get an idea of what to look for and, more importantly, what not to look for.

With that having been said, i think it's safe to say that i may be completely and utterly wrong and that Analyst 13 is rookie numbers XD
Phil Stargazer
TimeZone Warriors
Brack Regen
#3 - 2017-07-19 19:03:40 UTC  |  Edited by: Phil Stargazer
I agree 1000% on #2. When I get a failed analyzes, I'd like to try and do it myself.
But the lack of yellow points on your own fold , the absense of zooming in on the fold ánd not being able to retry it is a defenite lack of help we could get.

BtW, the magnifying glass is NOT the zoom I'd like on the fold. I want a scalable one.
Caseidae
Seniorenresidenz
Funatix
#4 - 2017-07-19 21:19:54 UTC
Thanks for your feedback.

For 2) I noticed: I can go back into and look into it. Sorry, if I missed that. I was sure I couldnt.

But this allows me to get more specific. I just failed a test and I cut the "to be find" sections out:

Results of 200042841

The pictures show the drops that I had to find. While two of it may have a recognizable drop the others are not differ from the usual flicker.

The bottom two pictures with the vertical line through it are what I have marked and was indicated wrong. I am pretty sure its a remarkable drop although I appreciate it doesnt mean its a transit :)


So please tell me how to solve this? What is the difference to the other flickering data, the difference to my findings?
I am so sorry guys but I think I am lost for this.
Saqara
Equestrian Postal Service
#5 - 2017-07-21 03:14:38 UTC
Yeah, this discovery method is a royal pain. I get analysis failed more times then not, mostly due to the very teeny tiny drops that look like ALL the other drops in the squiggly line mess. I look at the failed results and ask, "How the F*** was I supposed to SEE that?!" Least with the cells one it was kinda obvious on most of them. Watching my accuracy % plummet is kinda depressing so I just gave up on it, tbh. Cry