I'm leaving a question because I ran into a problem while implementing screen space reflection. The way I do it is by sampling the position and normal map saved with deferred rendering, changing it to the view point, using the reflect function to obtain the reflection vector, adding it little by little to the view position of the current pixel, and then using the ray marching method to determine the view position of the corresponding texture. We use a method of extracting color by comparing it with the depth value. However, reflections do not appear properly, colors appear in strange places, or are drawn overlapping multiple times. Even if you try both normal and world and view, it doesn't come out well, but world comes out more convincingly. Even if I try changing all the numbers, I don't get a good result. What's the problem?

screen shot 1 : normal view

screnn shot 2 : normal world

another best reflection screen shot