I disagree - I think the perfect solution is a combination of both. Use a measurement based system to establish initial ratings or modification factors and to eliminate the [brackets] of uncertainty in the Portsmouth ratings but continue to use performance data to fine tune the numbers. I'll admit to being astonished at how accurate the measurement based systems are but there are still issues and there will continue to be new issues in the future with new performance gaining design features that are really difficult to quantify...canted boards and/or lifting foils for an example. How much performance to you attribute to a canted daggerboard at, say 15 degrees? what about 45 degrees? Over the last several years, I've sailed a Nacra 20 against F18s and I've seen F18's become considerably faster in the moderate and light breezes though none of their parameters have changed. A measurement system just can't have enough complexity and be manageable to compensate for the differences between a box rule development class and a strict one design class that hasn't seen a significant sail shape / performance design change in 15 years.

My point is that where the measurement based system gets really complex and begins to falter, the performance based system can take over quite easily to refine the figures.

Wind range parameters in the Portsmouth system make the system very accurate when they are used (which is less often than it should be).

There are some other issues with performance based systems with rating creep as boats become older and sailed less by the experienced sailors...but that can be managed by establishing a lower ceiling for the rating based on the measurement rating of the platform.