OK, I've been working on my Alpha calc tonight, but there must be some criteria I'm missing. Apart from the run being under 500m, and the course being more than 90 degrees off the original, what other criteria is there for an Alpha 500?

In my test data, I am pulling short fast alphas that are faster than the longer alphas detected by RealSpeed and GPSResults. They tend to be the faster segments of the tracks that RS and GPSResults pick. Is there a minimum length for an Alpha? How about the angle, should it be 180 instead of 90?

AFAIK there is no min length for a 500m alpha. For the 1000m Alpha it is over 500m I think. Come to think of it there may be a 50m minimum for the 500m and 250m but Mal would have to answer that one

I tried a 50m minimum and still got the same results. However, when I tried a 100m minimum, suddenly all my results are spot on with RealSpeed.

Of course, none of the test tracks I have seem to have any Alphas under 100m, but that could be a coincidence. Until I get confirmation that there is a minimum distance, I'm not 100% certain.

Anyway, looks good. I'll publish Beta 4 tonight and we'll have a full set of calculations available.

Then I just have to work out what is causing the minor variations in 1hr speeds.

Dylan.

Dylan the other criteria is of course the separation of start and finish points, (50m in our case)
90deg sounds too small as well, but I don't think it's as high as 180.
I've seen some very high number alphas, produced by catapults, that weren't 180deg.
And were shorter than 50m, be lucky of they were more than 5m.
Haven't seen one lately so not sure where to go looking for an example, could take ages, not even sure if it was one of mine.
Nebs, does this ring any bells with you???

I've been back through all the .sbn files in my test set, and the shortest distance I have for an Alpha 500 is 111m.

This leads me to guess that Realspeed looks for a minimum distance travelled of 100m for an Alpha 500 result.

Once I put the minimum distance travelled to 100m, the angle is fine at 90 degrees, or up to about 150 degrees, so I've left it at 90 for the moment.

Select to expand quote

Dylan72 said...

I've been back through all the .sbn files in my test set, and the shortest distance I have for an Alpha 500 is 111m.

This leads me to guess that Realspeed looks for a minimum distance travelled of 100m for an Alpha 500 result.

Once I put the minimum distance travelled to 100m, the angle is fine at 90 degrees, or up to about 150 degrees, so I've left it at 90 for the moment.

How many alphas are you looking for in each file? Realspeed normally only gives the top 5, these are going to be the longer ones, so you may not see the slower but valid shorter ones.
I'm off now to see if I can find any short ones with GpsarPro.

I give up, I've been thru 3 months of files and only got as low as 250m looking for a 100 results per session.
But none of them included a catapult!

GPSResults, on the other hand, has Alphas down as low as 61m long in my sample data. I think it must be more generous about total length of the Alpha track.

I'd love to see a complete definition of what each program considers to be an Alpha. Since GPSResults allows such short tracks, it's inevitable that occasionally it will let a faster result through than the other programs.

Select to expand quote

Dylan72 said...

>>>> Since GPSResults allows such short tracks, it's inevitable that occasionally it will let a faster result through than the other programs.

Don't think so Dylan, the shorter the alpha, the slower it will be as the proportion of slow gybe speed in the average is greater.
The only way it can be faster is if there's an invalid spike or a catapult.

Think I'll take my gps for a walk and see how short an alpha I can make.

Hmm looks like you're right Dylan,

just jogged a short alpha.
Gpsar gives no result for alpha 500 but gives an alpha 250 167m long.
And a very strange looking thing it is, includes all 3 alphas I tried, one long 2 short.
Now no wise comments please!

You know I hate to disagree, but I have a file with a top Alpha 500 of 73.9m according to GPSResults.


The next four Alphas are all high 400's in length, and the next one under 100 is number 8.


There are only a couple of files like that in my ever-expanding library, but this Alpha is fine, no spikes or anything else, completely valid, and comes from Sandy Point in August. It happens to be one of my test files, which is why I had trouble proving my Alpha algorithm was good.

So it does happen occasionally.

Select to expand quote

decrepit said...
Now no wise comments please!

How many knots is that?
And are they all bunny ears?

3.36, but I stopped in between each loop. Seems very strange that 3 180deg changes indirection are included in the one alpha?
Next time I'm in windows I'll run it thru R.S. and see what that does.

Well, it obviously wasn't windy in WA today!

Select to expand quote

decrepit said...

3.36, but I stopped in between each loop. Seems very strange that 3 180deg changes indirection are included in the one alpha?

According to Mal's site, the "rules" for an alpha state that it can contain as many changes in direction as you like. Naturally, when sailing instead of running, straight lines are usually faster.