A Simple Exoplanets Project Discovery Guide


(Rexxar Santaro) #1


To make a right decision while scanning a satellite diagram it’s important to understand the basics of how the diagram is formed and what kind of information it holds in. One of my hobbies is astronomy and I have some tips for you about this project and how to play with it (better experience with telescopes than with microscopes, if to compare this project to previous). It’s very interesting to have access to the satellite data, even if it is simplified and modelled.

The Project Discovery Concept
This project works under one of distributed computing project criteria. The main idea is to write a program code which should launch on many different client computers, just to distribute massive computing operations over a huge number of machines. The computing power works additively, which makes possible to replace an extreme expensive supercomputer with many normal computers. Time ago I participated as a volunteer in such a project – I just installed an application which had a scheduled work mode to perform DNA operation analyses. As the fact that the program knows the correct or relatively correct result before submitting yours analyze just shows us how it works. This project works under an entertainment game style mode. In return, of our successful deeds, the project program gives us a “cookie” (nice EVE goods), which is especially tasty for all Alpha clones.

The Diagram
Every photometric wave diagram curve, as a model concept, is formed additively based on 3 different photometric sources of influence:

  1. Star. The main and most powerful source of light. Creating the basic “star-wave” luminosity diagram, which we can see on Solar Activity Classification UI menu. The binary star systems can have a transit wave effect caused by the second star, after covering first star.

  2. Exoplanet(s) eclipse. Is defined by “transit-wave” – a wave with zero intensity amplitude by default (has no effect on star-wave by default), which changes into negative “sink” amplitude during transition time (as a negative impulse). The transit-wave’s intensity sinks, as moon eclipse happens, decreases the star-wave luminosity no matter what kind of curve the star has.

  3. White noise. In signal processing, white noise is a random signal having equal intensity at different frequencies, giving it a constant power spectral density. In our case this is displayed by a noise-wave of very low luminosity amplitude with much higher frequency than star-wave and transition-wave have. The sources of noise-wave are: light from closely positioned stars in celestial angle, but which are situated far far away; glowing/reflection from nebulas; high frequency thermonuclear deviations, which causes different temperature on different zones over star surface; even from reflected beams of our Sun from meteorites situated in our solar system. Read more.

A star is a luminous sphere of plasma held together by its own gravity. The nearest star to Earth is the Sun.


A star’s life begins with the gravitational collapse of a gaseous nebula of material composed primarily of hydrogen, along with helium and trace amounts of heavier elements. For at least a portion of its life, a star shines due to thermonuclear fusion of hydrogen into helium in its core, releasing energy that traverses the star’s interior and then radiates into outer space. Almost all naturally occurring elements heavier than helium are created by stellar nucleosynthesis during the star’s lifetime, and for some stars by supernova nucleosynthesis when it explodes. As the star expands it throws a part of its mass, enriched with those heavier elements, into the interstellar environment, to be recycled later as new stars.[3] Meanwhile, the core becomes a stellar remnant: a white dwarf, a neutron star, or if it is sufficiently massive a black hole. Binary and multi-star systems consist of two or more stars that are gravitationally bound and generally move around each other in stable orbits.

Basic idea. In our galaxy are over 100B stars. Most of those stars haven’t planets. Most of the solar systems without planets have a young “blue” star or binary-star model. Mostly, the ancient stars, like our Sun, can have planets and only on veeeery few of those planets can exist life.

Tip 1. Don’t try to find a planet transition on every diagram sample. Most of those diagrams shows us solar systems without planets or the data isn’t enough to determine a planet.

In the Project Discovery UI are presented the most common star activities under the Solar Activity Classification menu: eclipsing binary (binary stars; life is not possible), rotating stars (like our Sun; life is possible), eruptive variable (dying stars: giants, bright giants, supergiants; life is impossible) and pulsating stars (white dwarfs, neutron stars; life is impossible).

Detrend tools – is designed to reduce the star wave – amplitude. It works like a traditional filter which extinguishes low frequency waves and amplifies high frequency waves.

Tip 2. Detrend should be used always when the star-wave significantly: decreases, increases, none periodically oscillating, periodically oscillating. The most common detrend value should be 10h or 24h and rarely 5h. The Detrend value of 1h is not usable (rarely usable to analyze high frequency pulsar stars).

A rare example of binary eclipse star
Eclipsing binary stars usually have a sharp and short luminosity sink, which is well displayed on eclipsing binary menu and are easy to track. It has a transition sink like a planet has, but it isn’t. The eclipsing binary star has higher sink amplitude and narrow spike. In the example below is displayed a complex case:

The sample shows us that one star is a bit bigger then another one, which is marked by greater transition amplitude sink (1.5- 2 times). It will be correct to mark them with different Epoch values and identical periods. Mostly, binary stars haven’t planets.

A rare example of rotating star
The star has a sharp star-wave with very low noise-wave. Around star collides a relatively big planet (maybe a gas planet) with 2d period.

Detrend tool let us to filter star-wave curve and to improve transition-curve respectively. And folded diagram looks like:

As a result

A rare example of pulsating star with high frequency
The example looks very confusing, because the sinusoid isn’t uniform and pulses over time:

Here is a folded diagram of the sample above:

The result is – 22.6% the star is a pulsar and, obviously, without a planet:

Another difficult example
I marked the sample as “No Transit” because it’s not possible to identify any transiting sink on folded diagrams.

Here is the view of folded diagrams of example above:

The example above hasn’t obvious transition sinks, like binary star sinks or planet transiting sinks, or sinusoid folded diagram like pulsars usually have. The star is, possible, a young monotonous glowing star without any planet colliding around.

Tip 3. The stars with sharp standard sinusoid or needle-like sinusoid folded diagram, mostly, haven’t any planets or the data is not sufficient for more precise classification. You can mark such samples as “No Transit”.

Discoveries of exoplanets
This is a new direction of astronomy science evolution. Welcome to modern astronomy where everything is displayed in white dot’s over a dark background and multispectral diagram curves. There are 3 main methods (among many others which are less popular) to discover an exoplanet: imaged directly (by telescopes), transit method (easy, require to determine changes of luminosity of a star over time; like moon eclipse) and radial-velocity method (difficult, requires to study Doppler shifts in the spectrum of the planet’s parent star). This project uses the Transit Photometry Method. The photometric method can determine the planet’s radius. If a planet crosses (transits) in front of its parent star’s disk, then the observed visual brightness of the star drops by a small amount; depending on the relative sizes of the star and the planet.


An exoplanet (extrasolar planet) is a planet located outside the Solar System. The first planet was announced in 1992, with two planets found orbiting a pulsar. The first confirmation of an exoplanet orbiting a main-sequence star was made in 1995, when a giant planet was found in a four-day orbit around the nearby star 51 Pegasi. Some exoplanets have been imaged directly by telescopes, but the vast majority have been detected through indirect methods such as the transit method and the radial-velocity method.

The Transit Photometry Method has three major disadvantages:

  • First, planetary transits are only observable when the planet’s orbit happens to be perfectly aligned from the astronomers’ vantage point;
  • The second disadvantage of this method is a high rate of false detections. For this reason, a star with a single transit detection requires additional confirmation, typically from the radial-velocity method or orbital brightness modulation method;
  • Red giant branch stars (big stars) have another issue for detecting planets around them: while planets around these stars are much more likely to transit due to the larger size, these transit signals are hard to separate from the main star’s brightness light curve as red giants have frequent pulsations in brightness with a period of few hours to days.

Basic idea. Most samples in the project have observation time near 30 days. They can have many transition-wave sinks, which marking the planet period. In some cases, we can notice a single transiting sink over star-wave just because the planet period can be between a month and few years. The planet transitions have small amplitude and larger spike compared to binary stars eclipse. The larger is the spike amplitude means the bigger is the planet (gas planet) and it is farthest from star. Transition-wave spikes mainly are hard to determine because they can be covered by star-wave spikes and intensive noise-wave. The only one tool to filter the star-wave and noise-wave influence is to use folding over specified periods.

Single transition example of, possible, a big gas planet:

Periodic transition example of a small planet (like Venus):

Folding. To fold the diagram to amplify the periodic transition-wave sinks just click on some potentially periodic sink and drag the mouse horizontally just to spread period markers over the diagram. Try to place each marker over all visible periodic sinks and fold it. The folded diagram can be scrolled left and right to match and amplify a possible transition signal.

Some obvious periods:

A complex exoplanet transition example
Here is the initial data. My first attempt to mark the sinks, which looks correct, had 2.345.

The folded diagram below displays 3 folded sinks for period 2.345. This is wrong period because we should have only one sink. So, the period should be divided on number of sinks, in this case, 2.345/3 = 0.782

The result

Some examples contain transitions of two planets, which are harder to identify with current UI tools.

Tip 4. A complex diagram view, like above, should be “folded” at different period markers to determine the possible transition(s) and it period.

Current analyzing tricks are not absolute correct, but they let me to get results with accuracy rating between 70 and 75%. With more experience and performed samples its possible to reach accuracy 80-85%. Higher accuracy is hard to achieve because the available data not always is accurate and some random mistakes can happen, like in example below.


Tip 5. To minimize the phasing errors of period markers (like on images above) try to place epoch close to the middle of diagram. Folded diagrams scroll to get sharpest view.

«Эта неделя в EVE» №163 (22 июля-4 августа)
This Week in EVE #163 (July 29 - August 4)
EVE Wochenschau #163 (29. Juli- 4. August)
Project discovery is a new job in eve
(Rexxar Santaro) #2

P.S. Sorry about fluent english. o7

Strange Samples

IDK why Epoch 16.9 is not a transition with period over 30 days.

When the transition sink can be atop the star-wave:

It’s my 7th sample marked as “Consensus results: None, you are the first to classify this sample”

Another example when folding isn’t exact enough to make correct marks:

A nice example when the “deepest” sink isn’t a true option:

I learned the lesson above and made some “correct” analysis below:

A strange rotating star:

A hard smaple. I have no evidence, but the folded diagram was very distorted. I made right markers intuitively:

This is one unexpected. The first one unique:

IDK, but it looks like a CPU signal model - not a star luminosity:

This is just EPIC!


Who can see a transition here?

(Lady Ayeipsia) #3

Don’t stress. Your English, though not perfect, is more than skilled enough. You have written a very nice guide. Thank you! I look forward to putting what I have learned from this into use on project discovery.

(Tedric) #4

I like your guide.
It cemented what little i know with what i was seeing.
The ones buried in the noise are the hardest to see.
I’m wanting the first Project Discovery patch to come out which includes that magnifier glass tool.
I find the detrending tool only really useful in the 24h region, less than that it is quite useless.
Strangely, i’m finding NOT using the detrending tool is giving better results.

(Rexxar Santaro) #5

Current UI has very limited tools to analyze transitions on hard samples, like one below

Maybe mathematically project discovery processing data ok, but it displays data in very low quality.
Magnifier glass tool will not help us to perform more exact conclusions. We need more filters like: Gaussian filter, integral filter, spectral filters. Integral filter, which can be implemented mathematically, is very good to reduce white noise influence. We need better scrolling tools for folded diagrams.

Not always, there are samples which requires lower values of detrend to get sharper folded diagrams. If star-wave changes significantly, it’s better to detrend it, otherwise the folded diagrams will have distorted view.

The guide was updated with Tip 5.

Project Discovery - Two Questions
(Cat Evergreen) #6

[quote=“Rexxar_Santaro, post:5, topic:10314”]We need more filters like: Gaussian filter, integral filter, spectral filters. Integral filter, which can be implemented mathematically, is very good to reduce white noise influence. We need better scrolling tools for folded diagrams.

I support this proposal.

(Rexxar Santaro) #7

A UI magnifier and divider, which are implemented in the last patch, hardly can help to improve the analysis. It can help in situation if you truly know if the transition persists and where it can be situated (approximately). I managed to reach the stable accuracy rating of 80% (+/- 5%). Higher rating is hard to acquire due to failed samples, which are marked correct (samples above), and wrong marked diagrams. I consider the average “true” fail rating is between 15-20%.

(Rexxar Santaro) #8

Finally, I reached accuracy rating 99%.

To climb from 90% to 99% was as hard as to climb from 50% to 80%. Beginning from 95% the accuracy increases by 0.2% per success sample. I noticed a new bug – I can’t reach accuracy 100% no matter how many successful attempts I performed at rating 99%. I tested it last 3 days. Does anybody have any idea?

The guide was updated with new samples.

(Gwyn Blaidd) #9

I guess that even bigger problem than difficulty in reaching level 99% is complete lack of incentives to go much above 90%. At 90% accuracy the points reward is about 80 and ISK reward about 90 000. So it makes no sense to struggle to reach 95% or 99% for the sake of time that has to be spent on analysing some dubious samples. It is much better from the perspective of progressing further to accept a given level of fails but simply skip on hard samples.

(Rexxar Santaro) #10

The difficulty to reach accuracy 99% is based on your luck – how many samples with transition you get and how many of them you mark correct. Somehow, I performed around 30 samples before I got one with transition. If you are focused on profiting it makes sense to reach accuracy above 98-99%, but it makes no sense if you are focused on climbing your Analyst Rank level for SKINS and rewards. Even after failing dubious samples it’s possible to climb your rank up. Knowing the principles of these diagrams and having Analyst Rank 100 behind it’s much easy to make correct transition marks.

My current rank is 167. I can determine if the current sample have a transition with accuracy near 85%, previewing the diagram, during first few seconds. It takes like 2-3 Folds to identify the correct period in most cases. There are very complex samples, which requires much more time, much more folding combinations and with the available for dummies UI tools I cancel ‘em (No Transit) to save time.

It makes no sense to grind SKINS for market. They are cheap on ‘filled’ market and completely don’t worth the spent time for ranking. I’m calling this SKIN-Scam – trying to sell a SKIN for it’s true value, let’s say 90M ISK, after paying millions in taxes and broker fees.

Project Discovery is the most profitable ‘job’ available to Alpha Clones. Average profit is 2.8M ISK/10 min comparing to mining on Venture – 0.8M ISK/10mins:

Pls, don’t call this discovery-ISK!

So, an Alpha can easily grind 16.6M ISK/hour at accuracy 98-99%. I’m not focused on this kind of venue. My goal is to reach level 250 to acquire the rewards, especially the PD-suit.

(Rexxar Santaro) #11

A 0.2 days of folded period is required:

(Dobermano4ka) #12

? :hushed:

(Chan'aar) #13

yeah those results SUCK :rage:

(Rexxar Santaro) #14

You can’t avoid those kind of errors. The current UI haven’t more accurate tools to point exact period in some cases.

Anyway, last 3 months there is only one type of diagram in DB, which confusing me mostly. That one which contains two luminosity sinks: one is very obvious while other is hardly visible (close to luminosity noise). Pointing them with two Epoch will give an error. So, the solution is, just to point the period for obvious luminosity diagram then divide the period on 2 (T/2) just to check for low luminosity sink. This method isn’t absolute correct.

(Aergri Evingod) #15

What confuses me most is the diagramm with a strong luminosity sink and a small luminosity sink those can be interpreted as a binary star and a planet also compared to analysing them to be two planets with the same orbital period, whichlooks odd like two planets at opposite position of the star orbiting it.

(Eazyjam Maverick) #16

heres another one for the records

(Eazyjam Maverick) #17


sometimes the UI can ruin it for you… not exactly sure why i failed this… but okay?

(Rexxar Santaro) #18

In your sample, the correct transition should be marked as one Epoch (both 10.235 and 8.323 are correct) with a period of 1.279/2 = 0.6395 d. In this virtually modelled sample this means next: around a small pulsar Star is orbiting a very small celestial object (not a planet can orbit only) with an orbital period of 0.6395d = 15h 30mins. The difference between small and big sinks is created, because the small sink is formed while the Star’s luminosity(radiation) is maximal. I should to note, that an orbital period around one day is an incredibly small period. For example, the Earth’s orbit period is one year (365 d) while the Mars’s orbit period is 687 days (1.9 years).

Why is only one Epoch?

  • Dual Stars (two stars are orbiting around each other), both of them, are big and can cover each other pretty well. They should create a thin sink spike diagram and their amplitudes shouldn’t differ more than 30-35%. On the sample above, the second Epoch has much smaller sink transitions by indicating that the second Star is much more smaller. This, in return, must throwing away the smaller Star to an elliptical orbit by making it a satellite for a short period of time, until it will be absorbed by the bigger Star. Elliptical orbit must have bigger orbital period than 1.279 d and asymmetrical sink diagram.

  • Dual Epoch of two planets isn’t possible. Two planets on different orbits can’t have the same period as I mentioned on Earth/Mars case above. If it’s an unreal rare case, when two planets are orbiting on the same orbit in antiphase (180’), then the second planet is too small compared to first one (due to diagram sinks) and it can’t orbit on the same distance from the Star as the second planet. Two different planets can orbit in antiphase if they will have the same mass M (one gas plane and another terrestrial). But gas planets must have higher radius R (size) to have the same mass M, which should be projected with more thick diagram sinks. They have identical width.

But it’s a modelled version and this can be a simple Dual Star diagram…

Yeah, this happens much more frequent than you thought.

(Hir Miriel) #19

I wonder if Project Discovery is not really about planets but about the people playing Project Discovery. Mostly a test of how long someone can go on against some pretty weird returns of information. I dislike the ones I know I will fail because I just can’t the line exactly where it needs to be, the line goes a bit too far to the left or right, and in the end I give up and cop the failure. And I always detrend and hit 24 hours, just wish I could have that as my default, rather than those two clicks and a slide each and every time. Then I fail sometimes by being too quick to click and end up accidentally clicking submit in my trance mode. Project Discovery is a wonderful idea, and no matter which model was picked there would always be some frustrations. Hopefully a few of them get sorted out. I have other gripes but I’m 85% for level 22 or so, which isn’t bad, but I’ve slipped into only doing the 10 double reward examinations, and even that isn’t because of the reward, more because I’m trying to support an idea which I think is good.

(Abaris Fagimasad) #20

Does anyone know if the samples which measure your accuracy are fabricated data or are real data which have been already analyzed and confirmed professionally?