As far as I know, no one has figured out a reasonable scheme to score one-design sailboat races in a way that adjusts the race results for the personal skill levels of the boat and crew such that the less skilled can compete on a level basis with the more skilled sailors. Golf has a handicap system that enables people of different skill levels to compete on an equal basis. And, sailboat races that are scored based on corrected time around the course are able to use one of several personal performance handicap schemes that adjust the elapsed time and permit different skill levels to complete more equally. In fact, these systems often adjust the handicaps during the course of a series as the results point to changes in the crew’s skills.
There are several versions of personal performance (time) handicaps in use around the world. One source for more information on these schemes is from Australian Rod McCubbin of TopYacht, who has done a lot of thinking and writing on personal performance handicapping. Here is a link to some of his writing. The SailWave race scoring program that Fleet 27 uses for scoring can incorporate adaptable personal handicaps, but only for races using time as the basis of recording finish positions and scoring the results.
But, with sailboat races that are scored solely by relative finish position, there are no handicap scoring systems in use. For one thing, race committees undoubtedly don’t want to record elapsed time for one-design races; the boats often finish so close together that it would be hard to do correctly. And it further complicates one-design race management.
To try and fill this capability, I have put together a one-design handicap scoring system to be used alongside the traditional scoring approach for Fleet 27. Actually, this scoring system is more of an alternate scoring approach, but its intent is to give those crews not often at the top of the leaderboard an opportunity for a meaningful, measured reward. The philosophical basis of this scoring system is:
- Reward the boat that does the “best” relative to its own average finish position
- Reward the boat that is most consistent over the long run (i.e., don’t reward the boat that bangs the corner for an occasional great finish position compared to the boat that is always fighting it out in the middle).
- z = (x – x̄) / σ
- x is the finish position,
- x̄ is the average of prior finishes,
- σ is the standard deviation of the boat’s prior finishes.
I actually use the negative of the z-score; i.e., (x̄ – x), so the largest positive value is the best score – it is easier to explain the result when the largest positive number is the best. I include only actual race finishes. I skip over all races coded with DNC, DNS, DSQ, etc. Counting these would benefit a DNC’d boat unfairly by lowering the mean (but maybe also increasing the standard deviation). The OOD score for race committee participation is now scored as the average of finishes.
This method seems to reward all except the boats at the top of the traditionally-scored series, which is good. Anyway, the boats in the top of the fleet have the RRS Appendix A9 scoring system to rate their success. Boats at the top don’t have enough headroom above their mean score to do well under this system. {Although, the results from June 26 racing may dispel even this hypothesis, as Schadenfreude slipped into 3rd place.} But that is ok, I think.
Note: After this season (2018), I need to take back the previous paragraph. Vixen won both the regular fleet season series and the z-scoring for the complete season, which I had previously thought would be extremely unlikely. In retrospect, I think the z-scoring worked well for the individual races, but for the total season, it turned out that Vixen’s season was marked by a general improvement through the year which enabled it to lead in the z-scoring as well. Although, other middle fleet boats (D-Sports, Medusa) also did well under the z-score season total. Bottom line, some further tweaking may be in order.
This system is really measuring how a boat finishes relative to its own past finishes. So, in a way it becomes a competition against one’s self. Since this system can encounter numerical problems when a boat has many finishes at the start of the series that are exactly the same (the stddev is zero and dividing by zero causes problems), it takes a couple races to seed the system. So, the Z-Score is determined starting with each competitor’s 3rd completed race. Since we start the handicap calculations anew each season, it takes a couple races to seed a competitor’s handicap formula. Alternatively, we could use previous year(s) scores to continue the mean and standard deviation calculations and then be able to include the first 2 races. Or, we could use some variation on the golf handicap approach of best 10 of previous 20 scores.
- z-mod = {(x̄-x)^A} / (σ^B)
but nothing has jumped out at me, and certain values of A or B caused other numerical problems for various reasons. But the intent of using an exponent is to weigh more heavily either the standard deviation or the difference from the mean. Possibly more experience with this system will point to a way to use the exponents.
Finally, to account for the fact that each race has a different number of entrants, I have been normalizing the finishing positions (and thus the mean and standard deviation) based on the number of starters. This is not totally satisfying, but I think it works better than not doing this. As long as the competitors don’t change too much, it is meaningful. If the skill level distribution of the fleet changes radically, the statistics really won’t be valid.
I have been experimenting with this system to score individual races. I have keep a running total of the rank orders, as normally done, to provide a cumulative series result. But if there is some other scheme to be looked at for the series scoring, I would be interested in exploring that.
So that describes the scoring system that I am experimenting with. The arithmetic is still complicated. It takes about an hour with an Excel spreadsheet to get the Tuesday Evening Series scored and double checked. Ideally, this process can be put into a computer program to process all the data and format the results more rapidly.
On a second tack, I have been thinking about a totally different approach on scoring. This approach determines the pair-wise comparisons of each boat’s finish position against every other boat it competes against. Then it ranks the boats based on the pairwise comparisons – similar to the way Google determines page rankings for a search and it returns the best match. The approach is full of linear algebra, which is not my strong suit, so the development is slow-going. I am still thinking about a way to add a handicap to this approach, but the idea is that any boat’s handicap for each race then would adapt for that race based on whether the race is dominated by faster boats or full of tail-enders. I think the handicap horse racing guys use a scheme similar to this. I suspect this approach would be the holy grail of alternate scoring systems.
Thirdly, the local J-24 Fleet 43 has a “level” series handicap method that Carter White and others developed and that Fleet 43 used for the first time last year. They are using it again this year. Last year, Andrew won under both regular and level scoring systems, so I don’t think the objective was achieved. Just looking at the handicap Andrew received relative to Carter last year, you have to think that cannot be right. But, it is something else to look at.