Seems to be a stupid article. Both ways to compare are very wrong either way. The correct way is to divide, and then mpg vs pgm does not matter. It just happens, that subtracting in pgm is perceived closer to the real rate you get by division.
It's not just a gripe over imprecise language... This is a serious logic error about wrong language that many people make, and it conveys absolutely the wrong information.
There was a highly commented article here a month ago which said a person bought litecoin at $67 and had achieved 570% gain at $387.
That's absurdly wrong due to a similar misunderstanding about math. Many people didn't understand _why_ that is wrong - even after a couple people jumped in to help explain the reason.
There can well be a serious logical error somewhere, but I think it most likely to be the authors fixation on speed rather than the interesting quantity in all examples: time.
If the newest build of the game loads in one tenth of the time it took an older build, I've eliminated 90% of the previous wait time and it is completely coherent to call the new build 90% faster (in wall time to be pedantic).
I think that fundamentally the author insists that "faster" can only mean "higher speed" and not "shorter time", so that when even when what he reports is a reduction in render time he only does it indirectly: by stating the change in speed...
> There can well be a serious logical error somewhere, but I think it most likely to be the authors fixation on speed rather than the interesting quantity in all examples: time.
There is a serious logical error.
> If the newest build of the game loads in one tenth of the time it took an older build, I've eliminated 90% of the previous wait time and it is completely coherent to call the new build 90% faster (in wall time to be pedantic).
That's not the algebra that most people use. It is, however, the exact same logical error demonstrated in the article. This is from the absolute basics of algebra.
Imagine if a doctor instructed a nurse similarly about how to change the amount of critical life-saving medicine from an IV. If the nurse or doctor didn't understand algebra or ratios the patient might just die from this error of 10x dosage difference.
Look, most people don't consciously use algebra at all! When they do intentionally use algebra I would be very surprised indeed if they did use something else than the standard field of real numbers.
I honestly have no idea what meaning you put into the word "algebra"; on reflection it's pretty weird to bring it to a discussion of percentages and their interpretation.
In the same vein, what do you wish to prove by bringing up the example of the nurse and the doctor? They seem to do the sensible thing and communicate in absolute quantities rather than rations. Still patients do die to mistakes in the drug administration every now and then, and often from the rather banal mistake of missing a decimal point.
Let's go a level deeper: why do people so often write "90% faster" to describe something that is "10 times as fast". If they were using "90% faster" to describe something that was "1.9 times faster", the likely explanation is that they are trying to deceive their audience while remaining technically correct. But why would someone bragging about a genuine improvement choose a wording that undercuts their message?
My guess would be that some authors feel that "10 times faster" is low brow, unscientific, and therefore unconvincing, while using a percentage comes across as more precise, authoritative, and thus convincing. As others copy this, the social effect strengthens. Are they possibly right about this? Is claiming "90% faster" perhaps more effective marketing despite being wrong?
This is why in realtime applications the correct measures are x times faster than realtime (relative throughput) or total runtime (absolute throughput) and a full histogram of response times. (absolute latency)
Relative latency would be used in games perhaps. (x times faster response than frame time)
