Good exponential algorithm for mouse acceleration

[Jezze]

Hi,

I'm trying to find a good algorithm for mouse acceleration during mouse movement. Any good ideas?

[bluemoon]

Do you consider threshold method good?

[Jezze]

I haven't tried it. Sounds like it could work. Previously I've tried to apply different algorithms on the entire span from slowest movement to largest and noticed that it didn't really feel very good, it was either to slow in high speed or to quick in low speed. I never thought about having a threshold where if the mouse reports a movement larger than the threshold it will start to accelerate.

[bluemoon]

AFAIK DOS, Windows and probably mac os uses threshold and user configurable speed (2x, 3x, 4x, etc).

[Nessphoro]

From a gamer's perspective:
Acceleration is total trash

[Owen]

From a gamer's perspective:
Acceleration is total trash

From everyone else's perspective, (and many gamers, e.g. those who play RTS games) acceleration is a great thing.

But, then again, game types which don't want acceleration are those for which there shouldn't be a mouse pointer to accelerate anyway (e.g. an FPS should be listening for raw mouse events, not pointer movements.

(and the merits of pointer acceleration in FPS and similar games are still somewhat debatable)

[Nessphoro]

I am afraid I must disagree.
In FPS I prefer having the acceleration on as it allows for a faster twitch response.
In RTS, however, I prefer to have it off. For example in Starcraft, microing units with acceleration on would result in many miss-clicks.
And I'm sure that a big portion of the community( TeamLiquid that is, not OSDev) will agree.

[bewing]

Engineers have done countless studies on human feedback. The basic rule is that feedback (and mouse acceleration is a form of feedback) should always be logartihmic -- because that provides the best "feel". That is why I suggested in the mouse wiki article to use the BSR opcode to calculate an approximate log base 2 of (deltax + deltay).

[bluemoon]

You realize "acceleration off" is a subset of threshold set to 255 (the max value of a movement packet, or otherwise the max movement of a sampling period)? The proper implementation should include threshold value, acceleration factor and perhaps acceleration function (linear, log, etc).