Is SailWave capable of, or willing to incorporate, the following one-design scoring alternative? The idea is to have a score that represents a boat’s finish in a race relative to its own historical average over a long series. Also, the score should reward consistent finishes (e.g., low standard deviation) and penalize wide variations in finishes (shooting the corners for the random victory or last place).
Here are some features of this scoring approach:
- Usurp the statistical parameter “standard score” (a.k.a. “z-score”) as the boat’s score. (Maybe the mathematicians in the community can suggest a better parameter to use here.)
- Z-score = (average - finish)/sigma (where: sigma = standard deviation of finishes)
- The z-score measures the number of standard deviations the finish is above a boat’s average finish
- Do not include in the statistics any DNS, DNC, RET etc. races. Boat must start and finish the race to have the race counted in the z-score series. When a bunch of DNS scores fill up the score card, the z-score results are not as meaningful for that boat, I think. Better to not use the incomplete races, but some may want to include these.
- Do include in the boat’s statistics any percentage penalty, OOD (race committee average) and similar races. Any DSQs probably should be counted, as well. Not sure how to treat DSQ’s in a normalized manner as discussed further below. Maybe it will work out fine using a normalized score value a little greater than 1 for DSQ.
- Probably don’t need to use the concept of throwouts here (a throwout improves the average but also improves the standard deviation - so the net result on the z-score might be somewhat of a wash).
- Use a “large” number of races to establish the statistics; say the previous 20 races, possibly bridging back to the previous year(s) series. So, this is best for club racing with season/year-long series. Probably not so useful for typical 5 or 8 race championship.
- For boats without prior race history, just start building up the statistics from the first race. Need logic to avoid dividing by sigma=0.0 during the first race or two. Often, in the case when sigma=0.0, then z-score=0.0 as well, but this is something that may need to be refined.
- Normalize the finish scores used in the z-score statistics formulas (say 0.0 = 1st, 1.0 = last) so that the mean and standard deviation are meaningful over a series with varying numbers of boats sailing in each of the races.
- Also, for a total season (or series) “winner,” determine the trend through the series of a boat’s finishes (e.g., perform a linear regression on finishes (or cumulative average finishes) and use the slope of the regression line as a score). The best average improvement in finish position over the year is award winner.
This approach turns the series into more of a competition against one’s self and provides a mechanism for boats in the middle of the fleet to have a chance for a meaningful award. The boat that beats its own average by the most places, but also with the most consistency (e.g., has the smallest standard deviation) scores best in this approach. This can happen at just about any position up and down the fleet (maybe not so much at the top of the fleet, but the top boats already get awards).