Number of boats to promote

Lester · Post by **Lester** » 16 Nov 2005, 10:20

An important topic for discussion is the number of boats to promote in a heat racing system. Currently we have 4, and several comments from the Mooloolaba Worlds and the 2005 Euros suggest this number is way too small.

The USA Nats used 6. Some specific feedback from that experience would be very helpful from participants and organisers -- Steve?

What plans do IOM organisers around the world have for using 6 or even 8 boats to promote in experiments that would help define some kind of "optimum"?

Olivier Cohen · Post by **Olivier Cohen** » 16 Nov 2005, 11:03

Be careful!

If you increase the number of boats to promote, you reduce the max number of boats at the event, or add heats...as there can't be more than 20 boats in a heat.

For example, in Mooloolaba, 84 boats in five heats (20 + 16 x 4)

With 6 boats to promote, it would have been 76 (20 + 14 x 4)

With 8 boats to promote, it would have been 68 (20 + 12 x 4)

...

Gilbert Louis · Post by **Gilbert Louis** » 16 Nov 2005, 16:30

Hi, only new to fleet racing but in my experience I found that it is easy to get "stuck" in a lower fleet for whatever reason where I should be in a higher fleet hence high points accumulated in the meantime. (it is then much harder if not impossible to recover)

More movement between fleets - ie more promotion and relegation should be at least tested to evaluate the difference.

I used to race in big dinghy fleet up to 450 boats at some events and they used International & National ranking to assign boats to the various groups and then they mixed the groups... it was a while ago so can't remember exactly how they did it but that was the gist of it.

don't see why we would need more groups though ?

Atb

Peter Allen · Post by **Peter Allen** » 17 Nov 2005, 02:51

I attended the USA nats,and found that 6 boats up and down was a much better deal than 4 having had that experience in Vancouver 2003.Look at the recovery Steve made from 16 to 4th at the end,and i think he fell back to B fleet while doing that,Craig on the other hand fell back to C fleet on the last day barley hanging in for the win,i myself was betwene A and B at least four times,plus several of the middle order would have had a similar experience.I personally enjoyed it more,but remember it's a bit easier to go up and also easier to go down,all in all it ends up with more action,i find it's a bit boring when stuck unless it's A fleet of course, also a good way at getting the leaders off that pedistal,i'd actually like to see a regatta try 8,forget decreasing the fleet size as it's only the worlds and the Europeans that command very large entries,what's 50-84 doing anyway other than making a financial contribution to the regatta.Peter

Steve Landeau · Post by **Steve Landeau** » 17 Nov 2005, 22:13

It is interesting how common it is to hear about "being stuck" in the bottom heats, and how difficult it is to get out (and up). While this is true without question, one must realize that when moving more boats, it makes it easier to go down!

I was a good example; I won my seeding heat to put me in A, but I finished 14th in the second race. A 14th would have kept me in A with only 4 movers, but with 6 movers, I went down and stayed down all day!

This certainly changed my regatta. I think it worked properly though, as I was not sailing well at the time, and did not deserve to stay in A without sailing better. I think this example proves that moving more boats will not only give more skippers a fair chance to do well, but will also pull down those that are sailing conservatively, simply to "stay in A". It is true, however, that it will reduce the amount of skippers in a full HMS format, and it also has the possibility of increasing your heat numbers. It worked perfectly for us, with 45 entries and 3 heats, we had 19 on the line every time. If we were to have 2 more entries, however, one would have to decide if you want 6 movers and 4 heats, or 4 movers and 3 heats. That decision would take some thought.

Roy Thompson · Post by **Roy Thompson** » 18 Nov 2005, 10:29

Moving more up and down also has the added 'benefit' that those in the A fleet will be under a pressure similar to the rest of the sailors. With only 4 up and down you only have a 20% chance of going up or down, bad if you in B, C, D or E, but great if you in A...that's not a fair system. With 8 up and down for example the odds are a bit closer and you have 40% chance (in a 20 boat fleet of course) of going down (worse than now for the A guys but fairer) and 40% chance of going up (better than now the others.)
All in all I think it may even out the playing field a bit....surely it's worth a tryat big event or international level no? What are the regulations on this? Can Nationals' organisers use a modification of the system without any problems from IOMICA?

Peter Allen · Post by **Peter Allen** » 18 Nov 2005, 14:09

Seems like most agree,i know from experience it's fairly easy to stay in A fleet if prevaling conditions are good and your boat is sailing well.I just like the idea of a gang coming to get ya from B fleet so you better start sailing agressively to stay there,as i said more action.Peter

ralph kelley · Post by **ralph kelley** » 19 Nov 2005, 16:53

One must remember that the HMS system is designed to form a classification system in which the best sailors sail with the best (hopefully getting the best competition) and those of lesser skills/equipment sail with their peers. This generates "groups" of sailors within a larger regatta that don't intermingle with the others in the regatta.

I, and I know some others, do not like the HMS system for this very reason. We like to sail with everyone and get the social interaction that this provides.

So if we must stay with the HMS approach, 6 up and down is better than 4 and 8 up and down would give a lot of sailing for those "in the bubble" that bounces between groups.

Ralph

Lester · Post by **Lester** » 19 Nov 2005, 19:18

ralph kelley wrote:One must remember that the HMS system is designed to form a classification system

Hi Ralph

As far as I understand it, the RSD (and MYA) HMS system wasn't actually explicitly designed to do this, though many think it was.

Lester · Post by **Lester** » 22 Dec 2005, 14:30

I'd like to start a project to simulate the outcome of various promotion/relegation places in an event, to try and see if there is some kind of "optimum" number. The idea is form a fleet of, say, N = 80 boats in the computer, and give each boat an "ability" rating on a scale of, say, 0 to 99 (call this ability rating A). The fleet is then roughly divided into five seeding heats (roughly, that is, so that part of the heat assignment is based upon the ability rating, but part is based on some quite large random factor (call this initial randon seeding factor S). Then, the computer runs off a whole bunch of heats and races, maybe races (R) = 24, promoting and relegating the first P boats (P = 4 for one set of runs, then P = 6 for another set, = 8 for another, and so on), and at the end we see how the simulated regatta results match the ability scores given originally.

So the question at the heart of the simulation is, how to simulate the outcome of a particular heat, given the n = 20 or so boats in that heat? My thinking is to say that, being simple if not simplistic, the expected heat position (POS) for a boat depends upon two random factors: boat's expected finishing position based upon ability, (call it EFP), and involvement in incidents during the heat (call this I), so the boat's final position in the heat is something like POS = EFP - I.

Specifically, we first score the expected finishing position EFP of the boat by saying this is equal to a random number between 1 and n (n is the number of boats in the heat), multiplied by the boat's ability:

EFP = (100 - A) * n

Then we score the effect of an incident of the boat, by first generating a random probability that it loses X places in an incident, multiplied again by the boat's ability to recover somewhat. The probability of an incident gets larger as we get lower in the fleet, perhaps, so we generate a random number between 0 and X, ran#, and multiply it by the fleet number f#, where f# is 1 for A fleet, 1.5 for B, 2 for C, 1.5 for D, and 1 for E. This says the prob of an incident in E fleet is about the same as one in A, while in C fleet it is twice that...

probI = ran# * f#,

so

I = (100 - A) * probI,

and finally

POS = EFP - I.

The POS results for the n = 20 boats in the heat are ranked in order, the top P boats are promoted, and the computer runs the next heat simulation...

What do you think of such a model?

Andy Stevenson · Post by **Andy Stevenson** » 22 Dec 2005, 16:47

Lester,

Sounds like an interesting project. Assuming the objective is to study the difference a different number of promotions / demotions will make and not an attempt to predict actual results

Iâ€™d probably be inclined to add a random "Performed above ability" factor for each race.

If you fancy a joint effort with the coding Iâ€™d be happy to pitch in.

Bruce Andersen · Post by **Bruce Andersen** » 23 Dec 2005, 00:51

Sounds great! perhaps we can determine a "skill factor" for each skipper, then virtual race all winter long. Do you think you can simulate the outcome of protests also? (note: my tongue is firmly planted in my cheek!) Happy Holidays

awallin · Post by **awallin** » 23 Dec 2005, 21:38

Hi Lester,

A nice idea !

being an experimentalist myself I would suggest trying anchor the model in reality. (do away with any more or less artificial model for incidents, protests etc.)

For example, look at the scores from vancouver, arcos, and mooloolaba. For each competitor, compute the standard deviation of their scores in the series. I bet the standard deviation will show significant correlation with capability (final placing). Fit this to a line or parabola and hopefully find a general relationship between ability and standard deviation.

Then in your model, the random variation in performance around the skippers average should be normally distributed with a standard deviation found by the method above. This way, incidents, protests, broken keels etc. are taken care of by the (pseudo)random number generator (I have C-code for the Mersenne Twister if you are interested, ventured into MonteCarlo simulations during the fall).

I'm not totaly sure of how you imagine the scoring happening but using the system outlined above all skippers in a heat would be given first their intrinsic ability "points", and then a random component, either negative or positive, and then all skippers in the heat would be ordered from highest to lowest (brake ties by chance).

Another related thing: I've heard more than once during an international event (by Mr. anonymous of course!) "why do we have to go on sailing and sailing for a whole week ? 10-15 races is enough!".
I've thought about trying to prove or disprove this by looking at the scores (list of who is 1st, 2nd, 3rd etc) after each race. Then you could compare (R^2 correlation or similar) this list to the final list. This would be a number that would start out quite low and go to 1 at the end of the series.
I think I did this for one of our national championships... will see if I can dig it out.

If you are using Matlab for the simulation I can help, otherwise I don't have time to learn a new syntax (sorry)

Merry Christmas to everyone !

Anders

Dick Carver · Post by **Dick Carver** » 25 Dec 2005, 07:41

This is very interesting guys, although I think modeling something as complex as a sailing regatta will be very difficult.

A couple questions about the basics. Lester, how do we define an "optimum" number of boats to promote/relegate ?
At the US Nationals in San Diego, we transfered 6 boats instead of the usual 4. I think the majority of skippers came away thinking that 6 was better than 4. But if one was to ask them why they favored more transfers, I think you would get some pretty vague answers like, "it was more fun", or there was more mixing of competitors.

If the latter is true, perhaps an optimum number of transfers could be defined as a number that allows a greater probability of winning the regatta for a greater number of boats. Sort of undoing the effect that Anders points out, where the eventual winner often seems to be determined very early on.

Also, on the probability distribution of an incident. Just an opinion here, but it seems to me that the probability of incurring a penality is pretty much the same regardless of which heat one is in. Getting in trouble seems to be more likely if one is sailing in the mid-heat pack, and less likely if one is out in front or trailing the pack.
What do you think of making the "incident" factor in a bell shaped distribution for each heat, maybe slightly greater for the D & E heats, rather that bell shaped accross all heats ?

awallin · Post by **awallin** » 26 Dec 2005, 10:48

this is the kind of analysis I was talking about:

scores are from mooloolaba and vancouver. no discards when calculating the standard deviation.

the R^2 numbers are not great but there does seem to be a trend: the top 20 or so skippers are more consistent, stuck in A, and the bottom 10 or so skippers are stuck in E.

I would repeat that if you/we do not have a very good understanding of the mechanisms of incidents, protests etc. then it will be hard or impossible to model them accurately. a random variability in preformance, based on empirical data (above), is my suggestion.

Anders

(email me for the excel file if you want it)

Lester · Post by **Lester** » 26 Dec 2005, 12:28

Dick Carver wrote:how do we define an "optimum" number of boats to promote/relegate ?

Hi Dick

Well, the model starts with allocating each boat an "ability" score between 0 and 99. So my feeling is that the optimum promotion number is whatever gives a final regatta result that correlates highest with these scores. A boat with the highest ability, 99, should on average return in first place, and so on.

perhaps an optimum number of transfers could be defined as a number that allows a greater probability of winning the regatta for a greater number of boats

That is certainly one approach, and would be best achieved by some kind of handicapping system I suppose.

it seems to me that the probability of incurring a penality is pretty much the same regardless of which heat one is in. Getting in trouble seems to be more likely if one is sailing in the mid-heat pack, and less likely if one is out in front or trailing the pack.

Good points. I think I might have confused "number of incidents" with "amount of aggression on display" (smile).

What do you think of making the "incident" factor in a bell shaped distribution for each heat

Yes, this would be better.

Lester · Post by **Lester** » 26 Dec 2005, 12:41

VCinfocomms wrote:this is the kind of analysis I was talking about

Hi Anders

Fascinating! The consistency right at the front, st.dev around 13 or 14, is not that different from the consistency shown in C (15) or even B (16) heats... And the consistency in D and E fleet is even better, somewhere around 11 or 12.

scores are from mooloolaba and vancouver. no discards when calculating the standard deviation.

The discards system could be thought of as a method of eliminating outliers -- "rogue" results which cloud, rather than clarify, the picture. What graph do you get with discards *allowed*?

awallin · Post by **awallin** » 26 Dec 2005, 21:00

I've now included the 2004 euro results and also plotted the standard deviations I get by excluding the discarded scores when calculating the stdev. see pic below.

If you want to model the actual performance of a skipper then I would think that you want the discards to count towards a higher standard deviation.
Excluding discards naturally lowers the stdev and gives a higher correlation between stdev and final placing.

also, the variability in performance we see in the graph below may be false/contaminated by the way it was "measured", i.e. with the HMS system using 4 up/down.
Thus I would re-phrase myself: any simulation you run should reproduce these graphs, more or less, _when_run_with_4up/down_. The 'true' variability in performance that you input to the program might be different than what we see here. Hopefully I said that clearly enough ?

Lester · Post by **Lester** » 27 Dec 2005, 01:36

VCinfocomms wrote:any simulation you run should reproduce these graphs, more or less, _when_run_with_4up/down_. Hopefully I said that clearly enough?

Hi Anders

Sure!

Lester · Post by **Lester** » 31 Dec 2005, 19:55

Well, what else does one do at Xmas? So here is a beta version of an event simulator:

http://www.onemetre.net/OtherTopics/Eve ... entsim.htm

Early results tend to suggest that there is little objective benefit from increasing the number of boats promoted and relegated per heat, though the subjective benefits could of course be considerable. Also, increasing the number of discards seems to have very little effect.

Bruce Andersen · Post by **Bruce Andersen** » 01 Jan 2006, 08:24

Merry Christmas

Not being one to knock scientific analysis, if the good guys are still going to win, why not increase the number of promotions? It sure makes those of us in the back of the pack feel like we have a better chance of getting up into A fleet (even if science tells us that we're still chumps)!

Happy New Year (or Hippo Gnu Deer if you prefer wildlife)

ralph kelley · Post by **ralph kelley** » 01 Jan 2006, 20:41

Lester's data is interesting, particularly his last graph.

His obversation that there is "little objective benefit" to increasing the number of promitions from 4 (which might have been an initial arbritary value) to a higher number depends on where you stand on the benefit issue. We need to define benefits.

Personally, I think there are benefits to having a larger number of promotions to get more sailing with a larger group of skippers and others probably like the current system that groups the best sailors into the A fleet for the tighter competition in the championship. I prefer the increased social and sailing interaction with the entire fleet, others will prefer different objectives.

I routinely hear complaints about the 4 up/4 down system we have been using. Conversely I have not heard of any negative feedback from the USA Nationals, and I think this is a clear sign of approval of the selection of 6 boats

Ralph.

Steve Landeau · Post by **Steve Landeau** » 02 Jan 2006, 01:02

I would agree that the changing the number of boats moving will most likely have no effect on who the top 3 to 5 boats would be, but I think it does make a difference, say between top 10 to 20 in a full fleet, or top 5 to 10 in an example of our nationals. The difference between 14th and 17th might not seem significant to a skipper that went in with serious goals to win, but the positive impact on a B heat average skipper making it into A more often gives them more initiative to strive for the next level. This would obviously expand down the the lower heats as well.
I think it makes the mid A fleet skipper work harder to "earn" the spot, since that skipper will likely be headed down a bit more often, so would have to show a bit more consistency to stay up. Even though there would be more boats moving up, it would still be harder to move up than go down.
As Ralph mentioned, I too have not heard anyone complain about moving 6. Even if it would not make a difference in the placing outcome of any boat, knowing that the skippers were happier in the end should be enough to make the change more often.

RoyL · Post by **RoyL** » 03 Jan 2006, 02:07

Have to second Steve Landeau's statement on how positive everyone felt about moving six people up and down at the US Nationals.

It made a major subjective change in people's feelings and the nature of the sailing. There was not that desparate attempt to not finish at the back since it was likely that everyone would be moving down eventually nor was there that similar feeling of despair when you were not one of the first few "breakaway" boats at the start of a heat because more than just the very top boats would be moving up.

As to the statistical models, not exactly sure what they show other than that the most skillful sailors tend to be at the top of the fleet. On the other hand if you guys can really actually predict how sporting events will come out, I could use your help with a few local bookmakers....

Bruce Andersen · Post by **Bruce Andersen** » 03 Jan 2006, 06:35

Roy

I realize that Doug is no longer with us, but could you dust off your famous "Fun-O-Meter" to add more data points to Lester's simulation?

Lester · Post by **Lester** » 03 Jan 2006, 23:31

Have added a couple more pages. Some of the interesting findings include "discards don't matter nearly as much as promotions", and "if equal weight is to be given to all boats regardless of ability, then the more promotions per heat the better".

IOMICA Forum

Number of boats to promote

Number of boats to promote

promotion

Re: promotion

promotions