On the subject of Jake and myself and being passionately opinionated, I remember no conflict between us at the time of NMBR and I feel the project was abbandonned for other reasons. Additionally, I much prefer a good fight over a "group hug and love another" all the time. I have no problems what so ever to working with Jake or others again. I understand that my personality can be a little bit abrassive (probably more then a little) but I must admit that the same character flaws also allow me to work through (compex) projects like NMBR very quickly. Two sides of the same coin.
On the topic of Yardstick systems (Performance based systems). In theory these can work well but only if the "system identification algorithms and supplied data" are well developped. This is actually part of my field of expertise as a Control Systems engineer. Having a well developped sys. ident. system is alot more involved then what current Yardstick systems are able to achieve. The quality fo the results, Ratings, are accordingly. We must take care not to assign a confidence in the yardstick systems that may be justified theoretically under ideal assumptions but are not supported by real life (sub ideal) facts. Mark Schneider touches on these point quite well in his posts.
With respect to measurement based systems. It is not forbidden or even impossible to use a Performance based (Yardstick alike) calibration mechanism to fine tune the (simple) describing function (i.e. regression). I know for a fact that Texel does indeed "Back check" the system using new race data and analyses whether a modification or adjustement is in order. Texel is not a static system and its revisions over the years have proved that. In fact, making a marriage between the measured based core and a performance based adjustment is ver much possible and adviced. NMBR was indeed developped using such a calibration method, although this is not documented on the website.
The more fundamental difference between a measurement based system and a yardstick system is that in the first you use ALL the data of all classes to fine-tune the describing function, where in yardstick systems you devided the total data set in significantly smaller sets to fine-tune each class rating individually. This is one reason why I favour measurement based systems so much. You make a lot better use of ALL the information that is included in the data set whereas yardstick system ignore significantly parts of it.
An example :
The perfromance of the occasional Prindle 16 that participates in a race in largely dominated by the very same limits as the Hobie 16's that participate. Systems like Texel use info milked from the H16 data to influence the P16 rating as well, as the describing function aims to approximate all designs accurately with only a handful of parameters. However, the Yardstick system suffers from dead-boat-society problems as only 1 P16 provides insufficient data of statistically poor quality to have the rating converge on. Afterall, who is sailing the P16, is he a good sailor, did he get caught in a windhole on the course etc.
This is one reason why new designs are always more accurately rated in a measurement based system then a yardstick system. No new design completely breaks with the performance trends as established by the other boats and the describing function. Typically, the adhere almost completely to these trends albeit at a slightly different location (coordinates). The describing function covers all coordinates whereas a yardstick system only covers a limited number of point that all coincide with existing designs.
No again, I underscore the fact that the describing function is not a fixed quantity. It can be adjusted on new race data if it turns out that by adjusting it coefficients it can become more accurate. In such a way a measurement based system can grow with the times and adjust itself to new developments. In some cases this means adding a new measurement to keep accurately rate old designs that don't go with the times. NMBR has done that and did indeed introduce a few additional measurements related to out dated designs.
So race result reporting is still much appreciated although much less important then under a Yardstick system. The system identification approach with a model structure to it (describing functions) requires much less data and can operate well with data of significant less quality. It is actually much less sensitive to measurement noise. I will leave the science stuff to this point.
Mark underscores a few good points. It is technically not difficult to make a rating system that does what we want it to do. Making is practical enough and attractive enough to have RC's use it is much less succes guaranteed. One lesson learned from Texe, when it too used two wind dependent rating numbers as NMBR had done, was that a mutany amng RC's killed that modification. Now we are back at single number handicap ratings again. Texel committee understands that this is less accurate but the real life won't swallow the revised system otherwise.
Therefor a good number of drawbacks assigned to measurement based system is not inherent to these system but rather forced upon the system by some groups with lots of power to make or break a system. I feel USA offers a good opportunity to break this particular dead lock as you guys are actually favouring wind dependent ratings. The USA could provide a succeful counter example that would eventually pressure RC's in EU to follows suit. By that Time I'm sure Texel and probably SCHRS will follow as well. Maybe that is the main beneft of NMBR, not a 3rd system competing with the others but as a crowbar breaking open the entrenched positions we find in EU and maybe the world at large.
Wouter
