(Hey guys! I just registered for this account after being a daily lurker for more than half a year. Even though I feel like I'm out of the speaker market myself (I got an amazing deal on some Technics SB-R1), there's always a buddy or relative that needs advice. More generally, the entire audio world is in dire need of some clarity.)
The proximate reason I created this account is because of the wishes for an easily digestible and hierarchically rankable numeric value for each speaker measured (comparable to the electronics SINAD metric). Now I understand that there is no single such metric that could encompass the totality of a speakers performance - especially because of the differing preferences regarding narrow/broad listening windows among buyers, the varying rooms those speakers will have to fit into and the willingness of buyers to EQ.
However, if I understood Amirs explanations of the various plots correctly, there seems to be one metric which could be transformed into a simple numeric value and which would be unambiguously desired by all prospective buyers open to the idea of equalizing the speakers (which I suspect is a sizeable and ever-expanding percentage of ASR readers): the relation of the amplitude-frequency-curve within the listening window and the amplitude-frequency-curve of the early reflections.
Imagine you sample both curves n times at corresponding intervals.
Then you subtract the sampled values of the early reflections from the sampled values of the listening window at the same frequencies.
You get n numer of difference values.
(Ideally, you'd want all those difference values to be equal: No matter how crooked the curve within the listening window is, if the early reflections curve tracks it precisely (by whatever offset), you can then fix it in EQ.)
So then you calculate the inequality among all the difference values. I'm not familiar enough with the mathematics to know which measure would be most suitable here. Perhaps a Gini coefficient or similar measurement could be adapted to produce one simple to understand number that would then tell us how amenable to speakers are to equalization - irrespective of how fucked up the unequalized curves look. You could call it the FBEQ-metric (fixable by equalization) and design a nice hierarchical ranking using it.
Going beyond this suggestion... while one value cannot encompass the totality of a speaker's performance, maybe a set of five or six such values could? For example: How amenable is it to eq? How linear is it withing the listening window without eq? What's the -6db point in the low frequencies? etc. If we could come up with a set of such metrics that could individually be transformed into numeric values, we could then build radar charts that would enable at-a-glance assessments of the general outlines of a speakers performance.