Averaging the band powers across channels will throw away spatial information, for one thing.
When you average the signals from all the electrodes, you may end up amplifying whatever information is common to all of them, which is the common-mode noise, something you want to ignore. Since some electrode locations exhibit more activity in certain bands than others, averaging will reduce their individual influence. For example, the ear electrodes tend to pick up more alpha than the forehead electrodes, so averaging them all together would just downplay the presence of alpha overall, and increase the presence of certain types of noise.
In fact, an average of all electrodes or some subset of electrodes is commonly used as a reference in standard EEG methodology. In other words, an average of a bunch of electrodes is considered in the literature to be representative of the common-mode noise.
If it’s an overall single measure you’re looking for, you might try using the experimental “concentration” and “mellow” outputs. Over time we will be adding more and more black-box style algorithms to the SDK that consist of just a single value. We’ll keep you posted!
Also, you may find some of the info I posted in this topic useful: http://forum.choosemuse.com/forum/de…ds-and-muselab