Detection PSM

[Ibr]

Hi,

I was making detection for an image of a Japanese document, the command was: tesseract image results_text -l jpn  --tessdata-dir ./tessdata -c preserve_interword_spaces=1 --oem 1 , and I noticed when I add the argument --psm 12 the accuracy is quite  better, as far as I know that the PSM is concerned with page segmentation meaning if there is a table in the image or to correct the orientation of the text in the image, and "12" is using the OSD (Orientation and Script Detection)  but I don't know why the detection was better.

so my question is , in general, is the detection always better with the argument psm 12, or its just a coincidence? keep in mind the image didn't contain any table, but skewed a little bit

Thanks