Sparky wrote:The biggest problem with most mouse tests I've seen is that they only look at what comes out of the mouse, they don't compare it to a more accurate position measurement. For example, you could use a linear encoder from DRO(a digital position readout for machining), and directly compare it against the mouse output.
Statistical analysis of mouse output can actually reveal a surprising lot of data. (granularity behaviours, rounding-behaviours, erratic behaviours). It won't be as perfect as a full DRO setup but a software app that does some basic statistical analysis or visualizations on position readouts, is what I'm interested in.
Initially I'll just do visualization (e.g. graphing deltas, graphing acceleration, etc) and letting people interpret the color-coded output. If there's waviness, sawtoothing, zigzagging, erraticness in the data, rolling-averages of varying depths, etc -- imperfections can shows up like a christmas tree in graph visualizations. So let's begin with easy. Reference mice (mice that eSports players love) will have their particular data look-and-feel in color coded graph visualizations.
Mice that does 'too much guessing' or 'averages too much' or is generally very erratic, will show either unusual smoothness (independent of mousepads: Rough-vs-smooth mousepads will show instantaneous acceleration spikes, unless the mouse intentionally averages the data to data-smooth this). Or color coded spikes in graphs (obvious overshoot artifacts from interpolation and extrapolation errors etc), etc.
I already started coding something, almost finished. It's very, very rudimentary, and requires users to interpret data (or just post screenshots). But at least it is a quick test. It does one thing most other testers don't currently do, at least in a user-friendly way: Realtime color-coded graph visualizations, with multiple concurrent simultaneous graphs for different data points simultaneously. They already do for some datapoints, just not all of them.
Users can keep using other testers, and simply use the Blur Busters Mouse Rate test application as yet another datapoint (to fill in the missing gaps).
I'm currently mainly interested in visualizing the accuracy of position reports. Statistical analysis can come later (since visualizations are really useful!).