Handicap Analysis

Handicap Analysis

PHRF Handicap Analysis on Lake Ontario – A Perspective
By Steve Corona (author of present Race Analysis Program)

Historical Background

The performance handicapping movement for boat racing actually started on the West Coast over 50 years ago but was organized, as we know it about 25 years ago as a response to the performance, high cost and complexity of measurement handicap systems.  It was originally not meant for “Grand Prix” racing events but rather was designed to allow individual clubs to easily and inexpensively rate and race boats of different sizes and shapes within their own organization.  Over the years the popularity of the system has expanded (wisely or not) to some of the largest sailing events.  PHRF is expected to be a “waterfall” system; handicaps are determined at the lowest level (generally the club or small geographical area) where performance can be continuously observed and handicaps adjusted in response to that performance.  Elevation to “higher authority” was necessary only to coordinate handicaps in larger geographical areas for interclub racings. Quantification of boat performance, in the form of “on the water” observations or race data analysis, is implied in both the name and spirit of the system.

The PHRF system was adopted on Lake Ontario about 20 years ago.  Prior to that time, a variety of measurement-based systems were used on the Lake.  The Lake Ontario Rule (LOR) enjoyed popularity for several years and along with IOR (International Offshore Rule), Measurement Handicap System (MHS), and Midget Ocean Racing Class (MORC).  These measurement systems formed the basis for the initial handicap assignments on Lake Ontario.  Changes from those initial “borrowed” ratings were generally laboriously made with hand calculations of race performance and subjective judgment of the handicappers.  Initially the “rating holes” in the measurement rules provided some boats a distinct advantage but these anomalies were patched over time yielding a performance based rating system that is considered improved over the measurement rules.   Another positive outcome of PHRF is that PHRF has allowed an improved range of boat types to be developed which are unconstrained by the hull bumps and funny shapes that provided an advantage in certain measurement systems.  Older boats have enjoyed an almost infinite racing life since, at least theoretically, PHRF neutralizes any advantage of newer faster boats.

The PHRF Goal

The fundamental goal of PHRF is to provide a handicap system where equally sailed boats of equal preparation can compete equitably over the long term. While the goal of PHRF is clear, implementation of that goal can be difficult for a number of reasons:

  1. It is very difficult to assess the skill level at which a boat is being sailed. Additionally the range of boat performance associated with skill level differences is very large.  Graham Moss, one of the initial PHRF-LO handicappers, offered a study that showed that the range of boat performance associated with the “skill level” of crew, and boat preparation on Lake Ontario was approximately 120 sec/mile.  This number dwarfs any adjustment one might make to an individual boats rating.  Obviously it would be nearly impossible to correctly handicap a boat without somehow considering the skill factor in any handicap adjustment either by subjective assessment or averaging a shear volume of performance data.  If there are very few boats of a class being raced and skill level can’t be assessed than the confidence level of the results is always suspect because of the influence of the skill factor. The subjective judgment of the local handicappers becomes very important in these instances.
  2. Individual boat performance varies by sailing conditions and course type.  There is no such thing as an equitable race for every boat in a mixed boat type fleet.  One or more of the boats will be advantaged or disadvantaged by the particular circumstance of the race.  Therefore any attempt to evaluate individual boat performance must be done over many races (perhaps even a season or two).
  3. Many races are unsuitable for race analysis.  A race is not like a controlled test tank; there are many reasons why a boat can finish ahead of another other than the actual boat speed difference.  Wind gradients, current, effects of competition, or other events can drastically increase or decrease a boats performance.
  4. Handicappers assessing performance are human and therefore, to an extent, inherently biased.  It takes an open minded, observant, and inherently fair individual to be a good handicapper.  More important is the handicapper’s realization of his or her own limitations and the inherent difficulty in objectively assessing observed performance.
  5. There is also a problem comparing performance of boats of widely different construction and intended use.   Can a cruising boat’s performance really be directly compared to a dry-sailed one-design boat?  A one design racing boat enjoys a single purpose and is generally sailed by individuals dedicated to racing not multipurpose boating.  One design competition provides a means to optimally tune the performance of a boat to a level not obtainable in a cruising fleet.  As an example, it is reasonable to assume that the average J24 is sailed to a higher level of performance than a Beneteau 361 with roller furling and shoal draft keel.

Assessing Performance

In the early stages of PHRF-LO it was recognized that data collected from as many races and from as many individual clubs as possible would be very helpful in setting and adjusting handicaps.  However, initial attempts to determine performance level from actual race data was disappointing.  By trial and error it was found that culling the race results of suspicious or spurious data improved the ability to predict boat performance (actually it was the only way it was possible).  This concept was reduced to simple criteria built into the analysis program. The analysis algorithm automatically extracts only the race data and individual boat performances that meet a “data quality” criteria.  For example: since it is assumed that the inherent residual handicap uncertainty of any boat is 6 to 12 seconds a mile, boats that exhibit a performance level over a set limit from the other boats in the race (i.e. 50 sec/mile) are extracted from the analysis.  The underlying assumption is that such a large difference in performance is due to factors other than any handicap rating error which might be present.  Also boats that exhibit drastic performance improvement chronologically (as determined by regression analysis), are similarly extracted since it’s not possible to evaluate performance when the boat and crew are on a steep learning curve.  Only relatively stable performance can be used to evaluate boat performance.

Individual races that exhibit a wide range of corrected time are also suspect and removed since the conditions on the course, for what ever reason, may have provided an advantage to one or more boats independent of the boats performance potential.  Because of this data removal it is quite possible that some good races are discarded, but that is infinitely better than leaving suspect races in the final data set.  Since the data removal is automatic, any handicapper bias for data acceptance is negated.

It was never expected that the PHRF-LO performance analysis of race results would be the sole method to set handicaps.  Rather the performance analysis was to be used as one, among many, tools to evaluate performance level of a boat class.  Handicapper interpretation of the data is still required to validate a change. An “Alert” of a particular boat is just an indication that further analysis and consideration is required.  Factors such as crew and preparation level, other analysis, and comparison to the rating in other geographical areas must also be considered.  Perhaps passing the “Alert” list information down to the affected clubs and regions for consideration might be a preferred way to handle a potential rating adjustment.  When in doubt it is best “to do no harm”, taking a conservative approach to the handicapping process, with the only caveat being that the fleet should never be compromised by a gross handicap error of an individual boat.   Any potential gross rating error should favor the fleet not an individual boat.