Bump with some content:
GladiOol wrote:
10 notches is more than enough.
Actually, that's a fencepost error... there are 11 notches, because 0 is included.
Except for CS, which has 6 notches... and the thing about a straight fence 6 posts long is that it has 5 rails, and 6 + 5 = 11! So adding in all the halves for CS will make it equal to the other parameters.
awp wrote:
it's kind of like how we used to have 10 different circle sizes, but that ended up being a bad idea, so we scaled it back.
This would be like that scenario, except in reverse.
The initial change was due to the range being too large. This increases granularity, which is a different reason. Having 11 different circle sizes itself isn't bad... if they had originally had the current range, they probably would have been left alone.
Arguably, if maintaining 11 notches is so important, the other parameters could have similar changes, lose a couple points that are rarely used and won't be missed much, while increasing the granularity in a range that's highly used. Sure that might not leave things linear (ie removing 0 and 1 to add two half levels near the top), but the perception of things like AR isn't linear either. The ratio between adjacent AR values gets larger as you go up... the difference between 0 and 1 hardly matters, the difference between 8 and 9 and 10 and above is much more noticeable, because each is actually a bigger step percentage wise (it follows 1/x, but backwards (ie heading to zero from right to left)). Which is why it feels like there is more room at high AR for in between values, there actually is more room at higher ARs because the ratio difference gets more and more pronounced until you hit 0ms and things feel infinite. To feel more linear (ie up to the point where humans perception can still tell there's a difference) the ratio between adjacent values should be constant (ie each is 3/4 the length of the previous) which results in a logarithmic curve to the ms values.