I certainly agree with this, but I'll offer a deeper historical perspective:
The original dial form is a product of sundials that cast a shadow from a central gnomon onto a roughly circular arrangement of hour markers (classically Roman numerals) arranged around the periphery of the dial. Clock dials followed this familiar tradition, which happily also was the most sensible mechanical indicator from the geared mechanism - and we can note the wonderful collection of gear-driven pointers and dials on the remarkable ancient-Greek Antikythera mechanism.
Some of my antique Volvos actually have a bar-graph or "ribbon" speedometer; these were somewhat popular in the fifties and sixties but I'm not sure when and where they first appeared. Some of these did use a ribbon I think, but the older Volvo / VDO ones were created by replacing the dial pointer with a horizontal cylinder laid on its side behind the cluster display. The horizontal cylinder had a colored helical section on a black (but vertically striped) background, all viewed through a horizontal-bar window mask. So as the car speeds up, you see a colored bar-graph speed indication extending from the leftmost zero position, and the stripes to the right of the red portion serve as x-axis tick marks for the number scale.
(This video bench-test was grabbed from the web and shows a European version)
A later variant replaced the colored markings with a helical slot so that a moving pointer could ride along under the number scale. There were also some designs that used a cord-and-pointer mechanism similar to a radio tuning dial. I'm not sure which this is, but this was the effect:
View attachment 750867
Having said all of the above, I'm very happy with a simple large numeric readout and dispensing with an analog dial or bar-graph simulation altogether. For things like SOC where the current fraction of max capacity is significant, the bar-graph makes sense. Personally I don't mind just a numerical percentage reading even there, but for me it needs to be larger than what Tesla likes to give us!
Regarding the earlier comments about lag and hysteresis effect, I believe there are some pretty simple numerical filtering algorithms that can perform the smoothing without noticeable lag-error on acceleration or deceleration. These can take into account the first or second derivatives of velocity ( i.e. the acceleration and rate of change thereof) to create a simple predictor function which is also filtered for noise.