|
|
Post by Fanny on Sept 28, 2017 9:38:50 GMT -8
The Handicapped Leaderboard, accessed by clicking the in-world Class Leaderboard or from the HotLaps website is an automated handicapping system using Goggle sheets and scripts that is totally hands-off. Click here to view the Handicapped Leaderboard. Here's the methodolgy: 1. Each HotLaps course runs for 7 days and can include 300-600 individual laps with 40 or more classes of boats. For each new week, the average lap time for each class is factored against the Nacra 17 base boat. The result is a weekly dynamic handicap for each class with the Nacra 17 having a handicap of one. 2. At the end of the week, each classe's handicap for that week is saved in history. 3. The trimmed mean for each class is then calculated for all weeks of saved history. Note: The trimmed mean is the average with the highest and lowest weeks averages removed to account for outliers. 4. The resulting trimmed mean is then used as the final dynamic handicap for each class. Could the handicaps be improved? Possibly by adding factors such as: Wind speed Points of sail and percent of time in each Heeling and reefing efficiency Rudder efficiency Others.... Although, what is the cost / benefit ratio of each and in the end, what is really necessary for SL racing? |
|
|
|
Post by burt on Sept 29, 2017 0:18:13 GMT -8
Do you also discard the top and bottom times for the weekly handicaps too? ( in case there are enough times....)
When you move the weekly to the historical calculations you should discard any handicaps that are calculated from only 1 or 2? laps ( sailors ) as they might not be real avarage times.
I would try to keep it as simple as possible.
|
|
|
|
Post by Fanny on Sept 29, 2017 1:45:59 GMT -8
Excellent thoughts Burt, I will augment the calculations.
|
|
|
|
Post by iteke on Sept 30, 2017 4:14:52 GMT -8
Why are the fastest times removed when calculating the trimmed mean ? The factor luck or chance has the least influence here and fast times show best what a boat is capable off.
|
|
|
|
Post by August on Sept 30, 2017 11:35:50 GMT -8
By keeping the top times in the handicaps they are skewed by the top skippers not a representative of the average skipper and boat. Same for lowest times
|
|
|
|
Post by Fanny on Oct 1, 2017 3:22:37 GMT -8
The purpose of the trimmed mean is to remove any excess fluctuations that could skew the average handicap. Typically, there is little or no difference between the trimmed mean and the simple average values. Click here for actual handicap examples that may help to clarify what is happening. |
|
joy
Developer
Posts: 4 
|
Post by joy on Nov 6, 2017 9:04:25 GMT -8
Hi. Might I just add a comment? (Im a very sporadic participant in the weekly races but am very interested in handicapping.
A couple of thoughts.
Id suggest that what 'trimming' is done depends on the goals, and on what concerns there may be about noise in the data. What Im trying to do for FIYC is to give everyone in a mixed fleet race "something to sail for", and in particular to adjust out the intrinsic difference in boat performance. So the crucial factor is the when ratioing mean times for boats, the means are on consistent basis.
I think a reasonable basis is to emulate "honest, good, but not perfect" skipper. Id get that by excluding the top 10%.
I'm thinking of excluding bottom 50% to remove effects of lag, bad sims, disproportionate number of newbies etc.
This is where there is a reasonable amount of data. Where there is some, but not a lot of data I suspect the least bad course is to just eyeball the data to see what might need excluding.
|
|
|
|
Post by Fanny on Nov 6, 2017 9:22:19 GMT -8
I'll change the algorithm to allow exclusion of different percentages for the high and low ends and test our history data to see the effect. Currently it's just a simple 30% off both the high and low ends.
|
|
|
|
Post by Andi Rain on Dec 1, 2017 16:51:42 GMT -8
Presumably chopping off the top 30% and bottom 30% is based on the " 68-95-99.7 rule", which seems a simple and reasonably fair approach. I'm guessing it would mean basing the handicap on data falling roughly within one standard-deviation of the mean? I don't think it's a good idea to build a handicap system around some subjectively determined "goal", especially if it's applied asymmetrically. It is not obvious why "bad" performances are more worthy of exclusion than "good" ones. Some classes of boat are much more difficult to sail well than others (the Flying Fizz, for example), and if that produces a bunch of "bad" times, there's no obvious reason why the difficulty of sailing the boat should not be factored into its handicap rating. One question that does occur to me, is the effect of "fleet size". I should declare an interest there, because I mainly hotlap in the RM Clever, which is not a popular choice for hotlappers (5 laps out of 208 this week as I type this). 30% of 5 is 1.5, so depending on whether you round up or down, I think either two runs or four runs would be cropped out of the five, assuming I understand the system correctly. |
|
