Ranking

[euphoria]

I don't see a reason why a long race should be weighted differently from a short race

Thank you. This was a mistake by me. One of the intetions was to value longer races more, since it's more likely that the race order will reflect the ranking order. But if a stronger skipper beat a weaker skipper on a long course, it should give less points than on a short course, since the probability to win is higher on a long course. But if a weaker skipper beat a stronger skipper on a long course, it should ideally give more points than a victory on a short course. Therefore this length parameter has to be included in the probability to win function. I will look into that. The other intetion of this factor was to give more points for longer races since it's takes more time than shorter. Instead of playing only short games to progress fast in the rating, this would stimulate to also play longer races. But I must admit this isn't very relevant as soon as you rating has stabalized at your skill level.

So if you want Harald, i can send you the code, database etc.
to play around with. Only needs to be ported to ASP and or Msql i think is what the vsk servers run on.

Jean-Pierre (Sothis) is actually working on a huge database with course records, tournaments and rankings. He will implement this rating formula there. He will very soon open for beta tests for ITBYC members.

Harald

[admiral 1]

If the intention is to reach the correct rating fast. eg. with a new skipper joining. K should be dependant on the number of races aswell

i propose to make x's K dependant on something like
[delta_in_finishtime_to_y /x's_finishtime]
(i presume delta = 0 doesn't exist) determines how good x is compared to y in this particular race, and to make it change faster (like when x is new to the game) multiply by highest_nr_of_races_of_any_individual_skipper / x's_nr_of_races

[delta_in_finishtime_to_y /x's_finishtime] takes in account the length of the race, course wise but also weather wise.

K for x then would include something like
(Dft_y / Ft_x ) . (NrRc_Max / NrRc_x)

IMO the probability for a win (1) remains indpendant of anything except a skippers record of performance before the start of any race.



Edited By admiral 1 on 1046256171

[euphoria]

IMO the probability for a win (1) remains indpendant of anything except a skippers record of performance before the start of any race.

Do I interprent you correct by assuming you say that skipper A has the same probability to beat skipper B, regardless of the laps/length? 1 short lap vs 4 very long laps (length difference of 24)?

If it was a speed race with no other boats on the water, I would have agreed. But imagine 8 boats on a short 1 lap course... Your performance is much more depentant on other boats actions (mistakes...) than on a long course where you have more time to recover.

The start is a key factor, as the starting line is identical for all course lengths (varies with No. of boats). And the length to first mark is important. The longer course, the more stretch in the fleet, and less probability that you will be "disturbed" by other boats.

It's more important to get a good start on a short course than a longer, which again makes the "random" factor greater at short course starts. The bigger the random factor is, the more likely it is for a weaker skipper to beat a stronger skipper.

Harald

[admiral 1]

the odds is a statiscal evaluation of who is better to date.
depending on what you are ranking (overall, only melges, only very long races) within a ranking however all parameters are equall for all

Now odds evaluates the probabiltay of a win 'W' (1 if won 0 if losst)

If in a certain Ranking (overall) A's rank is higher then B's the probability of A to score a win is higher then B's.

This means that if you pick a venue by chance from short, medium, long, very, long, melges, offshore, acc, multyhull, f3, f4,f5,f6 and f7 the probability for A to win is bigger then for B.

If B's is a master at starting tactics and timing, and thus may have a better skill at winning short races, why then should he be 'penalized' for that in the way his rating is calculated.
(his rating would no longer reflect his starting skill!)

If A is a lousy starter (like me coming late all the time) but has good boat speed (I still can do 30+ knots down wind in the acc at 2) then on average in doing an equal number of short and long races the score should be balanced.
(the rating reflects skill)

Iportant is that conditions are equal to all, so over an infinite number of races in every variation the ranking is only influenced by skill, all other factors then are balanced.

Would be a thing if for the ISAF matchracing ranking, NZL venues scores would be penalized because they have more frickle winds compared with european venues :)

Taking in account error's, connection speed, operating system or Internet Eplorer version is ofcourse all possible but highly impractical. Unless one can proof that over an infinite amount of races in every variation these factors have a determining meaning.

The same goes for matchracing a different game you say...
okay then you can't compare skippers that matchrace with those who don't. If you want to... how do you want to weigh and why? The only solution i came up with is that 1 matchrace equals 8/8 of an 8-boat fleet race equals 7/7 of an 7-boat fleetrace etc.

Unfortunatly not all skippers race melges, offshore, acc, multyull, short, intermediate, long, verylong, f3, f4, f5,f6 and f7 in equal amounts. Now if processing power allows a rating may be calculated for every and all variation..... and combination of those........
(that's why it is important to collect all data :)



Edited By admiral 1 on 1046278416

[admiral 1]

To bad i didn't save the mail thread from a discussion about 3 years ago about the same topic.

I remember someone (could be bifrosty) put the idea forward to set the ranking of new skippers on hold (calculate but not publicized) untill a minimum number of races is sailed. So that upon publication the ranking would be more or less an acurate reflection of skill, balancing 'accidents' like bargers and the likes.

[cubenviper]

do u guys think seriously that a skipper's talent could be judged ( in acc series ) according to an short or medium race with eight boats on the field? same as playing soccer in the bathroom !! if u want to consider that tactics and strategy count as much as only a good starboard start at comitee, then this becomes obvious...i do not pay any attention to short or medium races (acc) results... ???

[admiral 1]

there's no judging of talent in this or that condition, there is only comparison of talent in equal conditions.

Anything else would mean to discard with the short and medium races. Think of measuring talent? Then don't do any ranking at all and do some real sailing on real water..... it still is only a computer game, maybe a damm good one but still a game...

[NZL Huricane]

i totaly agree. medium,short and even long races are too short, and create a battle field and make the game increasingly frustrating, remove the rank system and we all have a good days sailing.

i personly love the very long and the real amc course, you get the chance to have some great tacking duels, time to pick the puffs and lulls, use the streams to your best advantage.

the shorter courses are 2 or 3 tacks at maximum which is not a real boat race, these courses should be for power boats and sprints, these are the rank hungry dreamers course.

loose the rank completly, use the game more and more, youll learn whos hard to beat and whos not.

[euphoria]

If B's is a master at starting tactics and timing, and thus may have a better skill at winning short races, why then should he be 'penalized' for that in the way his rating is calculated.

He is not penalized for his starting skills... If he's good at short courses, and doing alot of these races, he will actually be benefited by my proposal. Simply because the rating initially gives him less chance of winning and thus more points if he wins.

So no one is penalized by this reduced winning probability. It's just that you are not penalized that hard by events you are not master over (bargers, rammers etc). So called random events...

The same goes for matchracing a different game you say...
okay then you can't compare skippers that matchrace with those who don't. If you want to... how do you want to weigh and why?

I do not want compare skippers who only match race with skippers who only fleet race. It's different games and different rankings.

Unfortunatly not all skippers race melges, offshore, acc, multyull, short, intermediate, long, verylong, f3, f4, f5,f6 and f7 in equal amounts. Now if processing power allows a rating may be calculated for every and all variation..... and combination of those........

That's maybe an overkill, and we would loose the overview. 8 different rankings is maybe what we should keep focus on? But ideally, it should be possible to look up any rating by entering parameters to a database...

(that's why it is important to collect all data :)

That will be an awful lot data... I think this type of rating is great since you actually don't have to keep track of all old results. Just record the current rating (for all possible signifficant game settings).

set the ranking of new skippers on hold (calculate but not publicized) untill a minimum number of races is sailed

Maybe also an overkill? But very simple... Let's say your performance in the first 10 races will not contribute to your or the opponents rating. But based on your performance against other skippers with known ranking, it's a simple formula to back calculate your own rating. But again, this is not very important. If you loose a few points because a new skipper is better than his initial 1500 points, then you will win them back soon again.

do u guys think seriously that a skipper's talent could be judged ( in acc series ) according to an short or medium race with eight boats on the field? same as playing soccer in the bathroom !!

Good comparison ;) Just like short races, the result is much more dependant on luck in these situations.

Harald

[admiral 1]

Same as playing soccer in the bathroom !!

I disagree with the conclusion but soccer is indeed a very good example. Except that it's called indoor soccer, where tecnically gifted players stand much better...... played on half a soccer field i believe, with sideboards to keep the ball in play and heat up the action.

Short races can't be won if you mess up the start. 'accidents' can't be controlled by a formula, not even estimated. If you are going to be 'hunted' by a crowed without regard to the rules and the judge can't protect you, yes then you loose.

However if everybody adheres to the rules then a win is a win.
If you want to take in account how much time one spend on being better, average for race length, and use a delta / finishtime ratio.

you win with a 10 seconds delta in a 5 minute race =
you win with a 30 seconds delta in a 15 minute race

Instead of taking in account the number of races to you can take in account the sum of all finishing times (time spent racing)

1 VOR would then astute hundreds of AC's

IMO one wants to stay away from fussel factors or the like, that are prone to discussion.

IMO the only parameters to be taking in account are those describing the ranking list (all conditions equal for every particular ranking list) and those
describing skill. (those changed thrue actions of the skipper)
Wich would be acuratly calculatable, and not be based on some estimates.

If there are 'accidents' the only real measure would be to cancel result.

Unless i misunderstand what is proposed, it would'nt pay as much to do short races as much as it would to do long races, dependant on a factor estimated to account for 'accidents'.


recording all data

actually the data recording races would be only
1 race_id + description
maximum of 8 x skipper_id + finishtime
for every race.

~8000?? skipper records should be already in the database
at 1 Kb per record = 8 MB

witch comes much less 1Kb / race

6 boats per race

~ 8000 skipper
~ 10 minutes / race

10 hrs / week / skipper = 60 races / week/ skipper
with 6 skippers per race (host 1 out of 6) is 10 race records
per skipper / week
520 per skipper / year at 1Kb / race
that makes 520 . 8000 . 1 KB = 4 GB / year
so 20 G more or less should surfive the livespan of vsk2

I'm sure vsk knows better accurate numbers.



Edited By admiral 1 on 1046344056