DPI_from_forum

For the sake of discussion, let's say there are 3 types of mouse "aimings" in quake:

1. Move the mouse quickly from one point to another and stop as accurately as possible. The main usage for this is aiming with the rocket launcher, some SG aiming, some rail aiming and some trick jumps.

2. Move the mouse as smoothly as possible, at varying speeds in a sweeping motion for lightning gun/machinegun/chaingun/some rail aiming, and some strafe jumps.

3. Move the mouse smoothly in a very tiny space for long range lightning gun aiming, long range rail, or just about any long range aiming.

For the first type of aiming, which in my opinion is the most common, lower CPI (800 or less) is preferred. This is because your hand doesn't operate at such a high precision as to "stop on a count" at values like 7 micrometers per count. The precision your hand actually operates with is closer to the 800 and 400 CPI values, ~32 micrometer and ~64 micrometers respectively. While you can learn to stop more precisely (usually by lowering your average mousing speed), it isn't as "natural" to play this way.

For the second type of aiming, higher CPI seems to help. The question is whether values like 400 and 800 are already "high" enough for this purpose.

If you were say to use a very low CPI value like 10 CPI, you would find it very difficult to aim with the lightning gun. At a low 50 cm/360 sens each count would turn 1.8 degrees. At 800x600, fov 100, there's 2880 pixels in a 360, so 1.8 degrees would "jump" 200 pixels every mouse count. So that even at such a low sensitivity as 50 cm/360 you'd have a hell of a time getting a reasonable LG accuracy since at the most common distances the player could easily dodge in less than 200 pixels of your screen.

Your only ability to compensate would be by strafing but you'd be forced into a certain dodging pattern and the speed of your strafes wouldn't be able to cope with the enemy unless you matched his dodging pattern exactly.

Strafe jumps would be a nightmare because you'd need to just be lucky to have the correct vector available. Spontaneous bridge to rail jumps on dm6 would be out of the question :-P

So there's definitely a "too low" here. What about 400 CPI? Certainly it can handle 50 cm/360 without issue, so let's try a high 10 cm/360 to see if we can find a problem. At 10 cm/360 you've got ~0.2 degrees per mouse count. At 800x600, fov 100, each mouse count would turn the screen ~1.6 pixels. No such thing as 1.6 pixels, so it will be just 1 pixel. It would appear that this is good enough to hit 100 % LG should you have the skill to do so, however in practice there's a problem.

It isn't natural to move your hand very slowly, and very steadily, especially when other operations that may have occurred a split second before involved quick motions that featured many direction changes. So there's this gigantic difference in mouse delta needed, and your hand/body/mind don't want to do this. My theory is that this isn't just an argument against high sens (mid sens = less mouse delta to adapt to situations), it's also an argument for higher CPI.

While you struggle to move slowly with that 10 cm/360, you probably end up moving so quickly as to jump many of those single pixels mentioned above. Whether it's by a factor of 5, 10, or 20 I have no idea, but here's where higher CPI might make things easier. When the game receives a count, it must change the screen instantly by the amount of degrees/pixels associated with it by the sensitivity. If this change is large, some visual information may be lost. Higher CPI reduces this by inserting more counts "in between", thereby making the same motion appear (and feel) smoother. The smoothness is helpful to hitscan prediction

The lower the sensitivity the less of an effect this has, since the changes in degrees per count are smaller.

It might be (I haven't done the math) that the speed of the enemy dodging, and the frames per second the game updates indicate that there cannot exist an advantage here mathematically, but this goes against my own observation so I strongly suspect that the math does enable this possibility.

On the other hand, with hitscans, especially the LG, you often need to change direction quickly ala that first type of aiming, where it helps to be able to "stop on a pixel" and head the other direction. So while higher CPI may help, it may also hinder. All things in moderation.

For the third type of aiming, which I find to be the least common, I can definitely see higher CPI being helpful, especially in the case of high sens.

Injx gave me this handy formula a while back for determining the size of the player in pixels in the center of the screen:

k = H / ( 2 z tan[ F / 2 ] )

where,
H = screen resolution (horizontal)
z = distance from observer to object
F = fov (horizontal)
k = conversion factor (to be multiplied with the size of player's hitbox)

I'll examine the case of long range LG'ing. The LG's range is 768 units, so let's place the player 750 units away such that the LG is barely touching him. The width of the player's bounding box used to be 30 units, it has since been switched to a cylinder, and I am not sure of the width of the cylinder but I imagine it is at least near 30 units so I will use that number for now.

At 750 units, 800 horizontal pixels, 100 horizontal fov, the width of the player's cylinder would appear as ~13 pixels (perhaps even less since the enemy model is inside the hit cylinder and may not "fill it out" completely). If he's going to be dodging, you can see how easily he's going to step out of such a small hit window, and the importance of being able to turn very accurately to keep hitting him. I can definitely see 800 CPI being easier to hit with than 400 here, but 3600 is probably just overkill.

Even at 10 cm/360, 3600 CPI, 0.025 degrees/count, 800x600 res, you'd still have 0.2 pixels/count which is overkill in assuring that you won't lose any visual information. You're bound to do more harm to your type 1 aiming this way, while giving an unneeded level of precision to type 3 aiming.

There's another issue that affects all three of these, and that's the friction of the mouse on the pad. If you have a high static friction (resistance to start moving), and a low kinetic friction (resistance when moving), I think high CPI would murder your aim. This is because the mouse being a bit "sticky" and the slightly erratic speed going from stationary to motion would all be picked up by the high CPI.

I think that a lot of cloth pads are like this, although not necessarily to problematic degrees. The converse would also be true, but I can't imagine such a mouse and pad setup (40 gram mouse resting on the worn part of a sandpaper pad?).

Therefore, if I had to choose between 800 and 3600 CPI, I would choose 800.