I was looking through posts that need sourcing, and came across post #199073 . Now, its child post #672036 is higher resolution. But the source image is only the size of the parent post. They are both identical aside from the resolution. Tying to see which would be better to keep...

Bigger sizes are usually better, unless it's an upscale. Chances are somebody got the direct image from the artist.

Furrin_Gok said:
Bigger sizes are usually better, unless it's an upscale. Chances are somebody got the direct image from the artist.

eh, there's still too many variables to even say "usually" with only single aspect. There can also be immense difference between compression values and even with lossless files, there can be difference between scaling interpolation used with both versions. Traditional work there's also notion of digitizing the work again using new equipment and differend settings.

In this case however, it's artificial upscale.

Original version is pretty compressed making it harder to notice, as main giveaway of upscale is more soft look, but original is already extremely soft because of compression, as compression also softens everything.

However you can see stuff like macroblocking (8x8 squares on grid) which is how JPG compression works and it becomes more evident the higher contrast diagonal edges and more compression you introduce. In lower resolution it's 8x8, but on higher resolution it appears to be ~10x10 (bit hard to calculate visually) which would line up with 120% dimension scaling and the grid seems to be laid similarly.

Filesize is irrelevant, but factor to note is that higher resolution is 5 times larger, even though it's only 20% dimension increase. This is most likely thanks to much lower compression and it does appear lower res is 91 quality where higher res is 100 quality. Both of these look far more compressed than this, but especially 100 quality is always suspicious, really commonly that's done to decrease visual downgrade when saving JPG file again. 91 in this case is also bit suspicious as the compression looks like it was below 80, but as it's clearly downscaled work this could also be artists handling their files akwardly or alternatively the compression calculator is having a hiccup as it calculates compression from visuals rather than checking objective value.

Additionally like said, the higher resolution does not match any given sources and lower resolution hash matches given source. Known good source and exact match are usually preferred in unsure cases. On top of this for some reason the uploader of higher resolution did put up the lower resolution as parent post with no tags or sources which is really odd.

Usually in these cases, flagging either way is fine as staff needs to take a look at both versions regardless and can simply reverse the flag if it turns out another way around.