Zezo forum:

Post by Alexblond2005 | 2016-11-24 | 08:38:21

Hello Master Cvetan.
I invite everyone to give something to your site, which is very helpful. You certainly are able to know how many players on VDG2016 are routed by zezo.
I note a difference between speed in first infobox and my real speed when I change direction in the game, ex 15.3 vs 15.1 in the same conditions. What could be the reason: polars? Grib? Game engine aleas?
I'm certain that you have an idea.
Best regards
AB

6 Comments

Post by zezo | 2016-11-24 | 12:58:03

The polars should be OK, but there are small differences in wind speed and direction because of the details of the implementation of the interpolation algorithm.

I'm still investigating the following points :

- The game transfers the wind data in some specific way that creates some rounding error

- The interface downloads two sets of GRIB files with 6 hours difference - "old" and "new" forecast. Not sure how exactly they are used and if those affect the "current" weather or only the predicted track.

- I'm probably doing the temporal interpolation in a different way

In general in more stable conditions we see smaller error, but in dynamic weather the error becomes bigger. Should not affect the general strategy, but may be unreliable for minute-for-minute routing.

I already have better idea how things work and will be improving the accuracy in the future.

Post by Pennard-Bed-BZH- | 2016-11-24 | 13:06:52

Hi all,

About the 2nd point you mentionned Cvetan in your previous post, I read this from the club house (weather folder), wrote by VR Addmin :

Voici comment fonctionnent les vents sur Virtual Regatta Offshore :

● Les vents sont interpol�s c�est � dire qu�ils sont �volutifs spatialement. Les vents visibles sur l�interface restent valables pendant 12 heures. Le bateau utilise la force de vent et l'angle au vent du point g�ographique o� il se trouve pour calculer son segment d�avancement de 10 minutes. Au point d�arriv�e suivant, il utilise la force et l�angle du vent pour calculer le prochain segment et ainsi de suite.

● L��volution du vent, sa vitesse et optionnellement la direction peuvent �tre directement lues sur la carte en survolant les fl�ches de vent avec la souris.

● Les pr�visions m�t�o sont fournies par la NOAA et refl�tent les vents qui sont pr�vus sur la zone.

● La m�t�o est actualis�e deux fois par jour et active sur le jeu � 8h00 et 20h00 (UTC+2 en heure d'�t� et UTC+1 en heure d'hiver). En dehors de ces actualisations, les pr�visions � +12h et au del� sont mises � jour 4 fois par jour, � 0h, 6h, 12h et 18h (Heure de Paris).

● Les vents � 12, 24, 36, 48, 3J, 4J, 5J, 6J et 7J sont �galement fournis par VR pour vous donner une id�e de ce qu'ils pourraient devenir et vous permettre de pr�voir une route � moyen terme. Les vents � 12 h deviennent certains d�s 6h00 et 18h00. En effet, � partir de ces heures, les vents affich�s seront effectivement ceux qui seront utilis�s lors de la bascule.

Le vent donn� peut changer brutalement lors de la bascule. Il est donc fortement recommand� d'�tre pr�sent derri�re votre �cran � ces deux moments cruciaux de la journ�e.

Post by zezo | 2016-11-24 | 21:32:44

Moving it into separate because it's quite important.

WTF? I thought they have implemented proper wind interpolation when there is no fixed weather for 12 hours and folks in Brazil are not at disadvantage because of it.

And because of the local time instead of UTC the switch does not even correspond to GFS model output.

Post by guigui | 2016-11-24 | 22:27:33

Hi there,

The system described by Pennard-Bed-BZH from VR admin is from April 18th.

It was asked and answered about it yesterday and confirmed to be out of date.
--------
Rilwen
Community Manager
Rilwen
Administrators

172 posts
Posted yesterday at 4:04 PM � Report post
Bonjour,

Le post de Crew du 18 avril n'est plus d'actualit�.

Lire notre aide en ligne � ce sujet:

http://www.virtualregatta.com/faq.php?p=2729#p2729

http://www.virtualregatta.com/faq.php?p=53#p53

Bonne journ�e!
-------------------
The links do not offer any precise explanation of how the new system works.

I can't remember where I read (maybe here ?) they use the 3h / 1� stuff.
The hard part is to know how they mix it. I have no clue myself, sorry.

Post by joebar | 2016-11-25 | 02:08:00

Hi all,

From what I understood, winds interpolation works as follows:
- for a given point M=(long,lat) and a time T, first calculate the round coordinates (around M)
-> this gives you a cell defined with 4 points on a 1� grid. let's say (x,y),(x+1,y),(x,y+1),(x+1,y+1).

- for each of these points (ABCD), read the wind in the available wind datasets around the time T considered, and linearly interpolate (in time dimension) the u,v values for each of the 4 points.
- we now have 4 (u,v) pairs around our point. (Ui,Vi)=>(Speed_i,Dir_i)
- bilinear interpolation of those (u,v) pairs gives an interpolated (um,vm) pair at point M, from which the wind angle can be computed (dir=atan2(ym,um)).

- the norm (speed) is more complex:
- take S1 = norm(|um,vm|)
- take S2 = bilinearinterpolation at M(Speed_i)
- then use the formula that used to be present in iwind.js (see post http://zezo.org/forum.pl?tid=5716)
- final_speed = S2 * Pow(S1/S2, 1.0 - Pow(coeff, 0.7))

where coeff is based on a division of the cell in 4 even quadrants (using the centroid point of the cell). According to the quadrant M belongs to.

e.g.
..(X,Y) A --- AB --- B (X+1,Y)
........| .... | ....|
........| .... | ....|
.......AC ---- G --- BD
........| .....| .M .|
........| .....| ....|
(X,Y+1) C --- CD --- D (X+1,Y+1)

coeff = bilinearinterp on pos M in subcell (
G = norm(mean(Ui),mean(Vi))/mean(Speed_i),
BD = abs(sin(dir_B-dir_D)),
CD = abs(sin(dir_C-dir_D)),
D = 1)

Sorry for the poor explanations, but this what I reversed engineered so far.
I hope it helps.
Good luck with that ;-)