As a "veteran" competitor and judge since the mid 1950s, including competition at the Nats from as early as 1959 and 1967 (the year Bart won, Dick was second and Bob finished third with scores in the high 400s/low 500s) and then nearly continuous Nats from 1974 until, IIRC, 2006 I've been exposed to a lot of changes in scores in addition to seeing how the patterns appeared to earn those scores. Believe me, the guys back in the early days flew some pretty decent stuff but it did tend to be more open and less aggressive styles that, given the rules then and now, probably influenced the judges to some degree (I've no opinion regarding the Navy judges in those two years I flew before them. I'm sure they did their best but, given a lifetime of exposure to the very best fliers of their eras I don't feel those very low scores were representative of what would have been given by judges steeped in the genre.)
Post Navy hosted Nats scores trended higher but were still mired in the low to middle 500s as noted by other posters here. Come the 2000s and on scores escalated noticeably until we now see low 600s reasonably regularly. Yes, the top fliers are the best there has been in my life long exposure to the best. No question about it. Are, however, the flights receiving those scores that remarkably superior to those of the early big bore and tuned pipe era top guns?
Better yes...but not that much better. Here's an important part of what I think happened in those transitional years from the 540s to the 600s.
For many years the "favored" method of judging was based on deviations from an "average" maneuver and judges were taught/inclined/tended to develop scores based on the assumption that an "average" maneuver was worth 25 point (halfway between the minimum of ten and the maximum of 40) and the ultimate score for any maneuver was based on adding/subtracting points from 25 based on perceived degree excellence (up to 40) down to all but "unrecognizable" (the rule book basis for scoring a maneuver as worth 10 points). IMHO, one of the results of such a system was that something had to jump out as "super" in some manner or another in order to add to that "magic" average 25 maneuver but any perceivable flaw would result in deductions.
On the surface, this would appear to be a logical basis on which to evaluate performance and a great many contests were adjudicated in that manner for many years.
Just one problem that, if considered rationally, obviously makes such a system worthless for assigning numbers to maneuvers that have a rational basis. No doubt the ranking (first through last) of fliers in most contests was generally as accurate as our current contests. The numbers used to determine those outcomes, however, were often based on a flawed predicate. That assumption was that the declaration of an average maneuver valuation of some random number between 10 and 40 had no foundation with regard to the rule book descriptions of the maneuvers being assessed.
The only description we have of "correct" maneuvers is the rule book description of perfectly flown ones...round, square, 45 degrees, sharp corners (yeah, right five foot radii), etc.
Based on that rule book description the only rational way for a judge to determine the value of a given maneuver is to deduct "value" for flaws he/she actually "observes" during those few to several seconds it takes to perform them. Every maneuver must be evaluated based on its accuracy with respect to the rule book's description of that maneuver and the 40 point maximum reduced. Plain, simple and straight forward...but not easy!
Once the "average maneuver" concept is properly abandoned and judges concentrate on degrading from perfection there is no longer an "artificial half way to unrecognizable" starting value to each maneuver. Judges must, to the best of their ability, decide how much errors they perceive must "decrease" the value assigned to the maneuver and not let the number on the score sheet represent pluses and minuses from an artificial starting point.
Our top fliers--pretty much from the development of the tuned pipe power trains (and later by the even more predictable electric power trains)--make damn few errors even under less than ideal conditions. Properly powered and trimmed stunt ships of appropriate proportions and design flown by those top pilots are perfectly capable of flying maneuvers which will contain very few errors on difficult maneuvers and few if any detectable errors (within the ranges allowed) on easier ones and can do so even under conditions once thought to be "tough" to fly in!
The bottom line to all this is that once the concept of average maneuvers was/is abandoned as a "starting" point for evaluation of maneuvers for which we have only descriptions of "perfect" ones it becomes much harder to degrade very good ones from excellent ones and the bottom line numbers will predictably rise accordingly.
Note that the competitions still remain extremely competitive with top fliers scores within fractional parts of percentages. I contend that the numbers we see now reflect the reality as to how keenly and correctly human judges can discern and quantify the errors of very good to great fliers and the scores we now see are appropriate for the basis of evaluation provided by the rule book.
....Or not...