Find the Odd Disk

[lucb1e]

I was wondering if it got harder or if it's just random:

    function generateColors(difficulty, blacklist) {
        ...
        let sample = Math.floor(Math.min(Math.max(0, 1-Math.pow(1-difficulty, 1.5)), .99)*5);
        let distance = (5 - sample)/5;
        ...
    }
    function setupRound(blacklist) {
        ...
        const data = generateColors(currentRound/totalRounds, blacklist);
        ...
    }

Plotting that first magic: https://lucb1e.com/randomprojects/js/testformula.htm#%24%28%...

    round#  difference
     0-- 2  5
     3-- 5  4
     6-- 9  3
    10--13  2
    14--20  1

The "blacklist" parameter prevents that you get the same challenge twice. Note also that it submits every answer to the server (fine imo, but I think it would be even nicer if this was mentioned on the page)

[tennysont]

I appreciate https://lucb1e.com/randomprojects/js/testformula.html, it's a cool idea!

[readingnews]

Thanks for posting, I thought the same thing... my (useless data point of one) results showed 100% accuracy except the last four, which I thought "wow, I am just guessing now, can literally not see a difference".

[noduerme]

Got 18/20. I chalk that up to spending years as a graphic designer. I'd like to see a similar study about which text was perfectly kerned, or by how many pixels an element was off-center or misaligned. I can spot that on billboards a block away, and my life is therefore a constantly grating experience.

Marginally related. I paint oils as a hobby, and my studio gets northern light, usually overcast and cloudy, during the day. Differentiating tiny color variations under those conditions is very easy, and in general your objective "pitch perfect" impression of color is also pretty accurate. However, I've painted in the same room at night under a "warm" LED bulb, and been absolutely shocked at how wrong and blue everything turned out when seen in the light of day. Not just that, but the hues I intended to be close to one another are much farther apart than they appeared under LED lighting.

So if lighting conditions can shift not just your perception of a color, but also its relationship to the ones around it, then I think how much more does your screen gamma and range alter that? A fair test would be printed on the exact same Heidelberg in 4 colors.

[nine_k]

Regarding the LED lights: unless you use a lamp with CRI < 90, you see obvious, glaring color distortions, and some colors just "disappear", cannot be seen, because of the lack of a particular spectrum bands. Sadly, most inexpensive LED lamps have CRI around 80, and that light feeld definitely artificial, even if pleasant to the eye. A lamp with CRI 90 is okay, most things look natural, even though you can notice it's not sunlight. A lamp with CRI 95 is very fine, it's practically sunlight, and most tricky colors are visible well. I've never encountered a lamp with CRI, say, 97, but they exist and cost a lot.

(Source: doing object photography.)

[slowwriter]

I got 19/20. Turned off True Tone and cranked the screen brightness. Half the time I didn’t know if it was just my eyes playing tricks on me, but it was interesting to notice how the colors seemed completely indiscernible for a few seconds and then suddenly one stood out.

[01HNNWZ0MV43FF]

Swear on me mum I saw a game about kerning and alignment years ago on HN or proggit and of course it's impossible to find on search

[0xTJ]

I also got 18/20, I was confident on most of them, blinking I found "reset" my vision and made it easier. This was on my relatively comfortable (not too bright) monitor.

[rcxdude]

Yeah, metamers are a trip and a bad LED bulb will really screw with the appearance of colors. If anything, screens are more consistent, but more limited.

[zootboy]

I wish this had a "I can't tell" option. A few of the really hard ones I got right, but I'd say it was more of a lucky guess than a genuine ability to discriminate the difference.

[Freak_NL]

Quite. With the odds being 1 to 3 you'll pick the right one when you have to click one at random to proceed, the results get skewed.

[layer8]

This is from the creator of the ScienceClic YouTube channel [0]:

“As part of the next video, which will be out in a few weeks, l'd like to invite you to take part in an experiment about color perception. If you don't experience color blindness, l'd greatly appreciate it if you could take this test. Feel free to try it as many times as you like, think about it as a game!”

[0] https://youtube.com/@scienceclicen

[nkrisc]

Would be interesting to get some basic analysis of my results. From a glance it appeared that the ones I missed (6) tended towards red. The low saturation ones and green ones I found to be easiest, but was there any actual significance of the distribution of my errors? Simply too small a set to say?

[hilbert42]

I got the same number wrong but I've passed every Ishihara test ever thrown at me. I did this test on a cheap mobile that's not calibrated, so it's anyone guess what its gamma and transfer curves are like.

One should only take such tests seriously if one's using a properly calibrated monitor and it's viewed under ideal viewing conditions.

[robertclaus]

Interesting! All of my misses were blue leaning a bit towards violet.

[tripdout]

17 out of 20. Was super easy until #10 and I had to stop and think more carefully (which was actually my first mistake), and then I got #14 and #15 wrong. The score was about what I expected, though - would've been surprised if it was <15 correct.

I wonder how much of this would come down to screen calibration / color accuracy? If everything's consistently off in 1 direction I guess not much, but I would imagine certain shades might appear effectively the same on some cheaper screens?

[blueflow]

How much of the result is vision accuracy and how much is dependant of the display?

[schobi]

There is certainly also a device limitation. I would expect that with less than full 24 bits of color, some fields might just look the same and the results do not depend on your vision any more.

Let's say the device has a "24 bit color display". What about eye protection color shifting? This limits the color space used could reduce the effective remaining bit depth. Or maybe they do temporal dithering to get more bit depth? Or maybe the 24 bits are already achieved with temporal dithering?

It does not need to be a calibrated display, but a cheap tablet in sunlight will be worse than a color grading monitor in a reference environment.

I hope they also register the devices used and analyze the statistics on that.

[hilbert42]

As I've mentioned, you can't take this seriously unless you've a properly calibrated monitor.

[nuancebydefault]

What stood out a lot in this exercise is that when looking at, versus near a disc, its luminance (or maybe the color as well) is perceived as changing. Almost the same i have when staring at not too bright stars, they seem to disappear when staring directly on them.

And related, I once had an 'eye migrane'. During that half an hour, the figures of a clock disappeared the moment i looked at them.

[Liftyee]

I experienced this too. IIRC the brightness-sensitive rod cells are more concentrated in your peripheral vision while your central vision has more colour-sensitive cone cells. This makes the centre of your vision less sensitive to dim objects, so you can see them only while looking indirectly (and they "disappear" when staring at them)

Another related effect is flickering of badly designed lighting only in my peripheral vision. When looking directly at the lights they look fine, but when the lights are in my peripheral vision they appear to flash distractingly. I think the peripheral vision is optimised to detect fast changes/movements. At least, that makes sense based on evolution.

[fallinghawks]

I'm curious how the eye migraine is related. I had one many, many years ago. It was a smallish (palm at arm's length) oval in the center of my vision that looked like snow on an analog TV, accompanied by a feeling of overwhelmed by all the colors of the products on the shelves (I was in a grocery). It stuck around for about half an hour for me as well.

I've also had eye floaters which cause things to distort and can be hard to see through. For about 6 months I had a large one in the center of my left eye vision, which was a bit scary when I discovered I might not see a car reflected in my wing mirror.

[RetroTechie]

Similar experience. What surprised me: sometimes the odd-one-out was really quick/easy to see, and other times it took much longer & I thought "at the end, they'll say all disks were the same color".

Got 14/20

[fallinghawks]

19/20. The first many were very obvious. I missed #14 and I think right around there I started to slow down because the differences were getting smaller. Looking away and then coming back helped, I think. It seemed like if you look at them too long, whichever circle you're looking at seems to shift color and it becomes really hard to discern the difference.

[archmaster]

Interesting. Got 19/20 too, and also missed #14!

[js2]

The X-Rite Color Challenge and Hue Test:

Are you among the 1 in 255 women and 1 in 12 men who have some form of color vision deficiency? If you work in a field where color is important, or you’re just curious about your color IQ, take our online challenge to find out. Based on the Farnsworth Munsell 100 Hue Test, this online challenge is a fun, quick way to better understand your color vision acuity.

https://www.xrite.com/hue-test

(My memory is that the full test used to be online.)

[red75prime]

What is "color IQ"? Color identification quotient?

[susam]

19/20 - <https://i.imgur.com/GplfQbO.png>

By the way, I keep Night Shift enabled all the time: <https://i.imgur.com/LGkSlJZ.png>. I don't know how much it matters in a game like this.

See also <https://susam.net/myrgb.html> for a colour guessing game I wrote last year.

[richrichardsson]

> I don't know how much it matters in a game like this.

Probably significantly since it explicitly tells you to disable any blue light filters on your screen.

I think the website needs to be more explicit that this is trying to gather data and that deviating from the test parameters will skew the results.

[havan_agrawal]

Shouldn't there also be a few "control" challenges sprinkled in where all three are the same color and there's no "right" answer? If the test is implemented well and/or there is no human bias (either from the previous question or from the positioning of the circles), then you'd expect to see a uniform distribution of answers on the control. If there is bias (e.g. some innate preference for the top circle (say)), that should get adjusted for in the final analysis.

[wrs]

Toward the end, I think the confounding effect of the afterimages of the discs was bigger than the difference between the discs. (That is, when looking at the three discs in turn, the effect of the afterimage of disc 1 on judging disc 2, etc.)

[aimor]

I noticed this too. I did much better covering up two of the disks and looking at them one at a time, or comparing pairs of disks.

[gblargg]

It could rearrange the discs each time to avoid this issue, perhaps just flipping the pattern vertically each time.

[Retr0id]

Interesting. I only got 15/20, and previously considered myself "above average" at colour distinction tests but based on other replies that's not an especially good score. I'll try again, going more carefully.

[Retr0id]

Ah, yes, 19/20 the second time (only the last one wrong).

The first time I kept my eyes fixed in the same place roughly in the middle which clearly wasn't a good idea. On the second attempt I glanced between each circle in turn, trying to discern the difference over two points in time rather than two points in space.

[thechao]

I have poor color discrimination, but excellent flicker detection (?). This last skill was discovered by the senior devs when I was doing GPU driver debug, and “we” were looking for an extremely transient high-refresh rate tile clear issue. The issue only occurred at 120Hz (or higher) refresh rate with solid clear color on a large screen, with nearly identical colors. About one 4x8 pixel tile every minute or so. That was a boring few days, let me tell you.

[ryao]

This seems to be a test of the color accuracy of a display. I got 17/20 and the 3 I got wrong were the last 3. I did this on my iPad Pro. For the last ones, they all looked the same to me.

[gblargg]

I wondered whether any display issues caused some to have no difference. I wonder whether it could somehow do a test to be sure your display is up to the task. It's annoying not knowing whether it's a visual limitation or hardware issue that causes wrong ones.

[Animux]

Was way easier after deactivating the Android night light.

[throwaway0665]

Towards the end the disk I was looking at would change color and become brighter than the others. I didn't notice until I focused on each one at a time and the "different" one became the one I was looking at.

[yellowapple]

13 out of 20 for me. Would've been 14 but I misclicked on one early on.

[tristor]

Got 18/20, which I credit mostly to my calibrated display and photographer's eye for color. That said, both of the ones I missed were blue and I am wearing glasses that have some blue light filtering, which likely negatively impacted my perception of the blue parts of the color spectrum. It's food for thought as to whether or not I can effectively do post-processing of my photos while wearing glasses like these.

Interesting idea though. I wonder what the distribution looks like?

[Night_Thastus]

17/20, not too bad. My failures were a blue, purple and brown example.

As others have said, a lot of this also comes down to what monitor and OS you're using, glossy/matte screen type, interior lighting and how everything is configured. This will be a lot easier with an OLED and all but impossible with an old TN panel. And whether HDR is enabled or sRGB mode, etc.