2017-07-11 - Issue with Project Discovery Evaluation Set

Well you can narrow the time frame down to “zoom in” and look at a small more detailed selection of this graph but there’s just no discernable dips or irregularities. I think they may have a bug or problem with the actual program that’s just picking these patterns at random. It says there’s a transit every x hours or days and marks all those little yellow dots. But you basically have to guess to even line up one or two of them because there’s no indication of where you should start.

The transits ate supposed to be rare but literally every slide I’ve done has been claiming there’s transits every where and there hasn’t been a single no transit slide. What I suspect is many of these slides (not all) are incorrectly marked and claiming there’s dips in light when there’s actually nothing there.

You can fold it and line it up perfectly when there’s actual dips and you have intervals. But if you look at the broken slides people are linking including myself the transits aren’t dips in light theyre just randomly scattered throughout the slides. You have to click random spots and random intervals and fold and you can only drag 0.10 at a time it seems on the epoch so if you actually want to scroll through the whole fold it takes hours. At least for me, idk if it has to do with me using a trackpad instead of a mouse but there’s no way to fold and line these up and find a pattern if you’re just clicking at random.

Help…

1 Like

If you click on one of the tiny yellow rhombuses after the analysis, it will fold the whole thing back, and you will see that the average of the highlighted area is 5 pixels lower than the average of the rest.

Hang on… Are all those transits that I missed in the random noise really correct? Blimey. We need better tools because no matter what buttons I press it still looks like random noise. I’m not even sure starting out on easier samples are going to help. Even after I fail and spend some time examining the data where the transits have been highlighted for me, I still can’t see the transits.

That’s all great but even if the evaluation data is changed so that we’re not dropped into the deep water at the start: will it actually help in achieving the goal?
Do you actually think that with the tools the minigame gives I will - after some time - be able to spot a transit in data sample like this one?
click

While I believe there are tools that can prove there is indeed a dip in light curve there associated with a planet transit - how to a human eye is this part of the plot different from a part to the left or right from there?
Edit: I had to only include a hyperlink coz despite having played this game for the past uhm 4-5 years now the forum considers me a new user. OK.

Basically, folding doesn’t help. It’s a rare thing to see two obvious transits next to each other so you sort of eyeball it but the changes in luminosity are so subtle that when you fold the data is all looks the same, so you try eyeballing the lines on the tiny graph in the middle to get it to line up with where you think the dips are. But I’m pretty sure it wasn’t designed to be used that way. Also, being restricted on the screen size doesn’t help. If it really is that subtle then I need it across all my screens so I can really look at it.

A moving average might help. or some other way to manipulate the data. I dable with audio editing and in software packages you get all sorts of noise reduction tools and FFT and all that sort of thing. What about borrowing from that and having an array of tools that lets you really dig into the data. It feels like I’m looking at it from only one angle and it’s all a bit imprecise at the moment. It’s not just that it’s the deep end right after the tutorial. In a way having the hard samples first may serve a purpose otherwise people may just mark everything no transits except the obvious ones. You need to show the player the hardest ones first to get him in that mindset about how it could be really really hard to see a transit.

But in any case - this is way way harder than I expected and much harder than the last project discovery.

Well, I think with time we will learn to better recognize the patterns. This video demonstrates some patterns clearly. It seems that we need to look for a ‘w’ pattern in some cases. It is very easy to be fooled by those fast dips, but remember passing in front of the sun is not instantaneous. It takes time and thus we need to look for wider valleys. I think that if you keep doing this you will get used to the patterns eventually, even if now you feel like nothing makes sense.

Of course, the difficulty should get fixed. I mean that if your accuracy drops then the system should give you easier samples, so you can raise your rating again.

this is my ongoing collection of derp samples. I’ll keep adding as i find them. I’ve missed a few here and there.

currently lvl 23 @ 85.7% accuracy.

1 Like

That video seems to have much more obvious samples than what I’m presented with. But… If the tough stuff really is real - as in these samples can be so subtle, then I’m not sure I can do this with the current tool set.

The Data points on these are meaningless hogwash…I have analaysed quite a few samples and there is nothing to distinguish transits from general background noise. Dips and spikes could be transits but arent and the transits exist in spaces that are as uselessly uniform as the rest of the data. Hogwash!

Plus also we need to have contnued access to the tutorial. Noy just use it once and its gone. The tutorial could do with some improvement. It tells you what to do but not exactly what you are trying to achieve or how to best utilise each stage.

1 Like

the tutorial can be redone just click on the (?) in the bottom right corner. but the tutorial needs more explaining or i am just really bad at this xD

1 Like

Nope, I dont have a question mark on mine…I also have a ridiclous fail, half the image is green and half red about a millimetre off for a failed analysis. Nonsense! Apparently new users cant link images. More CCP genius :slight_smile:

Sorry new users cannot upload images , i have been playing since 2012 … had a sample all green but one at right end and says failed …all were inline , program is busted , lol

1 Like

I think they mean new users of the new Forum…although its new so we are all new users…would make sense that we could submit examples of how this isnt working.

1 Like

Maybe we just mark them in mass no transits and consensus would work ?

1 Like

http://imgur.com/Q978p2G

1 Like

this whole game needs to be shutdown and redone from the ground up. or at least reimburse people for all the points they lost today. i was 98.9accuracy on the cell project, im current at around 8% on the new one.

every single slide im getting is incorrectly marked with transits that arent there and at this point im just clicking randomly and skewing results.

all results need to be reset, and you already mentioned everyone going back to 50% tomorrow but we need to get credit for all the slides we tried to do today cause ive done probably 50-60 so far and ive failed every single one despite giving the correct answers.

1 Like

After scanning quite a number of these, I believe that most of these transits are far too subtle for the human eye to determine in the graphs.

Speaking as a scientist, I’m terribly sorry, CCP Leeloo, but this data is not suitable for human analysis. It will take super-computers to evaluate this type of data.

3 Likes