New Posts New Posts RSS Feed: New PY Website
  FAQ FAQ  Forum Search   Register Register  Login Login

New PY Website

 Post Reply Post Reply Page  <12345 14>
Author
AlexM View Drop Down
Really should get out more
Really should get out more


Joined: 10 Jan 06
Online Status: Offline
Posts: 857
Post Options Post Options   Quote AlexM Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 9:02am
Hi Chris,
I think itís an IE issue because it doesnít seem to happen on Firefox or Safari.
When does it change to medium confidence? I think itís pretty hard to convince people that we should be changing the number if it says low against it.

Alex
Back to Top
chrisg View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 23 Mar 07
Location: United Kingdom
Online Status: Offline
Posts: 886
Post Options Post Options   Quote chrisg Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 10:02am
Not sure Alex, I'll ask the question. I only have some classes on 0.4(low), then others jump to 0.7 (high). I thought I had some in between but haven't. It may be that we need to rethink the wording of low, medium and high. Any thoughts on what would be better? Perhaps 0.1-0.2 - low, 0.3-0.4 - reasonably, 0.5-0.6 - fairly, 0.7 and above - highly.

I use firefox and i definitely get the problem of not all classes showing up sometimes. I just logged in to check my confidence factors and had to hit refresh about 4 times before they all appeared. Definitely something that needs investigating.
Something I forgot to mention earlier is that when you upload, I think the PYS site does the race claculations immediately but only updates the report for the classes at regular intervals. So if you upload and want to check immediately it may not have updated the report page, thats what the floating refresh button is for on the report page.

Chris


Edited by chrisg - 30 Mar 12 at 10:12am
Back to Top
marke View Drop Down
Far too distracted from work
Far too distracted from work


Joined: 16 Jun 08
Location: United Kingdom
Online Status: Offline
Posts: 209
Post Options Post Options   Quote marke Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 10:58am
< ="text/">p, li { white-space: pre-wrap; }

Chris/Alex

The confidence factor is a composite factor looking at sample size, the distribution shape, and the historical trend. Each of these is a useful measure of the validity of the data. The tricky bit is joining the factors together to give a simple to understand single factor. As Andy said in his email, this aspect of the CF will need to be tuned when the PYS has suficient data to be able to get some meaningful data out of it.


The issues that you and AlexM are seeing look to me like an issue with the individual factor scaling. When the pressure is off Simon to get the site released I'll take a look at some specific datasets and see if I can spot any problems with the algorithm or the implementation.


Absolutely endorse Chris' comments about using it as a relative factor until the RYA has had chance to tune it a little - looking at the the data shown, even a low factor is better than no data at all, or murmurings in the bar Wink


Mark

Back to Top
Simon Lovesey View Drop Down
Far too distracted from work
Far too distracted from work
Avatar

Joined: 30 Nov 04
Location: United Kingdom
Online Status: Offline
Posts: 317
Post Options Post Options   Quote Simon Lovesey Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 11:17am
Pleased people are starting to find the new PYS site useful,  as aready stated we are very much in early stages of release and refining the system as we get a better understanding of the vast number of variables at play here.

With regard to the reports,  these are built in advance  (there are many more automatic calculations than V1) -  this means if you are jumping around you are likely to see a lag in the data -  again this is all part of the refinement process and we are working on improvements in this area.

New logins are being released on a batch basis -  this allows us to manage and take on board comments received in a practical manner,  rather than be flooded with the same comment.  The staggered release also allows us to manage server loading.  The good news is once you have received your login it takes no time at all to upload and process several years of data -  so thank you for your patience.
www.sailracer.org
Online Sailing Results, GPS Tracking & Event Management
Back to Top
AlexM View Drop Down
Really should get out more
Really should get out more


Joined: 10 Jan 06
Online Status: Offline
Posts: 857
Post Options Post Options   Quote AlexM Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 1:02pm
Thanks for all the replies and the hard work!

I think i'm going to go with everyone else and anything over the 0.3 we'll change and assess at the end of each series.

So it's

RS400 from 949 -> 940
Solo from 1153 -> 1147

Seem reasonable

Alex
Back to Top
craiggo View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 01 Apr 04
Location: United Kingdom
Online Status: Offline
Posts: 1723
Post Options Post Options   Quote craiggo Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 7:52pm
I'm a bit gutted that none of the historic data has been ported across. Are we honestly expected to go back and reload all historic data?
Back to Top
chrisg View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 23 Mar 07
Location: United Kingdom
Online Status: Offline
Posts: 886
Post Options Post Options   Quote chrisg Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 10:12pm
Craiggo,

I thought originally that the site was going to cut old numbers off and add new in to it, but as it works the numbers out using a different method I think its best just to start again. However, I can see this causing problems to any clubs that had already adjusted numbers based on the old website. Although not sure there were many to be honest.

I reuploaded every single race I had in the old system (500+) to the new site in about an hour and a half. With the new upload from sailwave file (not uploading from inside sailwave) you can do a whole series at a time and it runs the calcs on every race immediately. The whole site is much less labour intensive. Let us know how you get on.

Chris
Back to Top
craiggo View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 01 Apr 04
Location: United Kingdom
Online Status: Offline
Posts: 1723
Post Options Post Options   Quote craiggo Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 10:33pm
Thanks Chris,

The issue is trying to find all the old .blw files and then making sure that the boats spelling are consistent in all series.

I'll give it a go and let you know if I come across any issues.

Paul
Back to Top
chrisg View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 23 Mar 07
Location: United Kingdom
Online Status: Offline
Posts: 886
Post Options Post Options   Quote chrisg Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 12 at 11:23pm
Ahh yes, hope you can find them. It is worth opening the file and just ahving a quick glance down the list of classes to check youve called them all the same thing as its easy to lose some. You need 9 data points I think it is before boats appear in the report page so if you accidentally misname one, it may never appear...

Chris
Back to Top
sargesail View Drop Down
Really should get out more
Really should get out more
Avatar

Joined: 14 Jan 06
Location: United Kingdom
Online Status: Offline
Posts: 1186
Post Options Post Options   Quote sargesail Quote  Post ReplyReply Direct Link To This Post Posted: 02 Apr 12 at 8:28pm
Originally posted by slop_idol

We have discussed changing at 0.3cf due to it being a good amount of data for the sample size. When we can pool like venues in to the system, and the cf goes up we will use a higher cf before we change, but as it is the boats are currently performing to these numbers and as such should be changed. Now if it turns out to be too harsh the data will go the other way and we will change it again. Its great that the system now reflects this, but I think that having a pre defined low medium and high value against the cf number to be like the old style of defining the yardsticks with primary secondary numbers etc which will scare people off changing numbers.

Yes - but on the other hand it would be useful to ask for those options, to see how they varied at different cfs.

As for the blaze moving as much as it does, I have to say that it is probably along with the fireball the most suited boat to our stretch of water, therefore it will perform above its percieved performance range for most clubs.  There hasn't been a lot of all in races at the club where either a blaze or fireball hasn't won or dominated as fleet the results in the last few years, the Musto is another one that when conditions suit can't be touched on handicap, so every dog does have its day I guess.

And this is where I struggle with using cfs as low as 0.3.  If the Blaze is 50 odd PY points better then you ought to see a similar change in the 300.  Now I'll accept that there aren't enough races yet for the 300, but you are, of course, changing the Blaze relative to everything else - if that's actually because you're water suits the Blaze then by not changing similar boats you are creating an anomalous situation.  How much is Crew Skill Factor too. 

As it stands I think this new system is much better and with perhaps a few minor tweaks and the proposed planned additions it maybe as perfect as handicapping can possibly get.  The trick is not to be scared to change numbers if it looks like they need changing, just make sure everyone knows how you plan to implement changes and that the rules used to apply them are the same for all classes.

Yes I'm all for experiment, as long as there's a control, and the experimenters understand what they're doing.  The irony in all of the PY work is that you could make statistically perfect fair racing and still annoy just about everyone!
Back to Top
 Post Reply Post Reply Page  <12345 14>

Forum Jump Forum Permissions View Drop Down

Bulletin Board Software by Web Wiz Forums® version 9.665y
Copyright ©2001-2010 Web Wiz