Benalla

On Wednesday, January 18, 2017 at 2:30:06 PM UTC-8, Tom Claffey wrote:
At 18:02 18 January 2017, JS wrote:
Flying again today?
Jim


Yep!

Thanks, Tom!
From what I've seen, don't think Sammy The Snail will be going home with you. I know Kerrie doesn't like him.
Jim

Well, the difference between 66% and 75% is 9%, so it's a matter of degree. I've seen speeds less than half the winners on occasion (and complaints about the resulting score). Lowering the percent does focus on relatively small numbers of points between pilots who have already dropped 250 point to the winners. Dropping 340 points to the winners instead of 250 so you can claim a few dozen points against another pilot who is waaaay out of the money seems like it might be worthy of re-thinking as it only solves an ego problem with very slow finishers wanting to feel like they did better than a long landout or another very slow finisher. You pay a price in more important parts of the scoring formula (because math!) to create points spread at the bottom of the daily scoresheet. We need to look at all of this holistically. You can't push on the ballon on one side and expect it not to bulge out somewhere else.

A little adjustment to 75% doesn't matter as much as the assymetry of devaluation that BB describes, but it does create tension for other considerations.

9B

On Wednesday, 18 January 2017 21:44:13 UTC+2, Andy Blackburn wrote:
Well, the difference between 66% and 75% is 9%, so it's a matter of degree.

No it isn't. If winner flies 100 km/h, you score speed points from speed range of 66-100 km/h. That is 34 km/h range. Diminishing that to 25 km/h means 26% reduction, not 9%. So you are in effect taking away quarter of point spread awarded to fastest "lone wolf".

There is simple scoring formula that is understood by most, the GP scoring. If that is what pilots like, it is available.

Sure, you can take a percentage of a percentage, but it kind of misses the bigger picture about how scoring math works and what the tradeoffs are. You can't just talk about one objective in isolation. You have to check snider how they play off against each other.

If your goals a

1) Not penalize landouts so much that coming up 2km short knocks you out of the contest - especially if you are unlucky enough to be the only landout for the day. This is the case being discussed where scoring encourages staying in gaggles most of the flight.

2) Award points roughly proportional to speed/winner's speed.

3) Not have long landouts score more than the slowest finishers.

4) Allow for wide variations in finishers speeds and still have proportional scoring as in #2.

You have a problem that is constrained such that you can't maximize all the constraints at once. My personal view is id rather compromise on #4 a bit and somewhat on #3 in order to optimize more around #1 and somewhat on #2. The reason is because doing that helps prevent a bit of bad luck from knocking a pilot totally out of contention (the gaggle incentive problem).

Increasing weighting on #3 and #4 might be good if you are optimizing for score differentiation in the bottom half of the scoresheet, but they also contribute to separating the bottom half from the top half because slow finishers and landouts get even fewer points.

In the US rules we used to have 400 max points for a land out which allowed finishers to get points differentiation down to 40% of the winner's speed. We've upped it to 60% now.

I don't know if 75% is too much, but it's pretty clear that if you want to reduce incentives for taking risks flying away from the gaggle, reducing the penalty if it doesn't work out would help - as was demonstrated to clearly by Sean's situation, which was exacerbated by how IGC devaluation works where a lone landout gets 300-plus points. Miss the last thermal getting home and drop from 3rd to middle of the pack. It's no wonder people elect to stay in the glider tornado even though they mostly claim that they hate it.

One person's perspective FWIW.

9B

"check snider" -- "consider"

(autocorrect)

We see a lot of that these days - knowing the rules going in and
complaining about the results later. If you land out, you're not gonna
win the contest. Note: I did NOT hear Sean complain about this and do
not intend to imply such.

On 1/18/2017 12:44 PM, Andy Blackburn wrote:
Well, the difference between 66% and 75% is 9%, so it's a matter of degree. I've seen speeds less than half the winners on occasion (and complaints about the resulting score). Lowering the percent does focus on relatively small numbers of points between pilots who have already dropped 250 point to the winners. Dropping 340 points to the winners instead of 250 so you can claim a few dozen points against another pilot who is waaaay out of the money seems like it might be worthy of re-thinking as it only solves an ego problem with very slow finishers wanting to feel like they did better than a long landout or another very slow finisher. You pay a price in more important parts of the scoring formula (because math!) to create points spread at the bottom of the daily scoresheet. We need to look at all of this holistically. You can't push on the ballon on one side and expect it not to bulge out somewhere else.

A little adjustment to 75% doesn't matter as much as the assymetry of devaluation that BB describes, but it does create tension for other considerations.

9B

--
Dan, 5J

Ok, let's imagine score formula that gives 900 points to all finishers and remaining 100 points are awarded according to speed. No more gaggles, problems solved?

Wrong. Nothing changes. Same pilots will win and others loose. All we change is point spread between pilots. If winner of the whole competition gets 10000 points and last one 6000, new formula gives 10000 to winner and 9500 to last one. Point spread is very small, but it is as difficult to make any difference by flying as before. Next we start calculating decimals.

On Tuesday, January 17, 2017 at 8:12:30 PM UTC-5, John Cochrane wrote:
7T: I'm sorry to hear of your heartbreaking landout, especially from such a good position. Time to show great sports psychology and go win tomorrow!

It's a great example of the IGC rules issue: Only landout, a few km short: 330 points. Only finisher, everyone else a few km short: they get 999 points. Now you know why "stick with the gaggle" is so vital in IGC scoring. Better to land out with the gaggle than to take any risk in order to be the only finisher.

US scoring isn't perfect either. It also switches from speed to distance points in a complex way depending on landouts.

Heres my current best suggestion:

points = (day devaluation factor) max ( 1000 x speed / winner's speed, 750 x distance / winner distance)

the day devaluation factor doesn't matter here. What matters is speed vs. distance points. And the key -- they are fixed, irrespective of the number of landouts.

So, only finisher gets 1000, 1 km landout gets 750. Only landout gets 750.. The incentive to be lone wolf goes way up.

It works as now if you're really slow. If you are below 75% of the winner's speed, you get 750 points for finishing. You always get the better of speed and distance.

And real simple too. People might (gasp) actually understand their score! And strategy. No more need for team captains to report landouts to tell you if it's a speed day or distance day.

it gets rid of some other IGC idiosyncrasies too, like the occasional incentive to deliberately land out.

John Cochrane

But I want to have to solve non-linear partial differential equations in my head while on task! Why do you want to take all the fun out of it?

Make racing great again!

QT (tongue firmly in cheek)

On Tuesday, January 17, 2017 at 6:51:07 PM UTC-6, wrote:
Yeah sorry guys, blew it yesterday. Long story. Was almost at 133kph, needed a 3 knot climb to maintain that for 1500 feet. Anything higher would have increased speed so a 900 day was close but...no cigar. These guys here are incredible. Just happy to be here and experience the pace they set.

My point on rules is that we need to be careful. Our rules in this environment would not diffuse gaggles much (IMO). That said, I agree with John C, John G and Rick Scheppe that our rule system is more inviting to not flying in gaggles.

My example yesterday (only land out) was a great example of why FAI scoring rewards reducing risk by staying in touch with the gaggle or others. I flew almost entirely alone and was isolated at the end of the flight. Even though I made it to the mountains which are the best late day lift source, they were. It working and I was doomed. Also, now behind, few points scenarios allow a major catch up.

So I want US scoring to make it into FAI. But not the tasking!

Focus on the current race, Sean. Plenty of time for the US System distraction when you get home!

And, yeah. Those guys you are racing are REALLY GOOD!

Steve Leonard

Winning a day against those guys is quite an accomplishment.

Racing is making decisions based on estimated probabilities. As BB points out, scoring affects those decisions because it skews the upside/downside payoffs. Limited upside and huge downside equals risk averse behavior to maximize total score. Landouts and devaluation are one way this happens but speed points are another.

Example - The prior race day I stayed up late watching the final leg. Early on that leg you were basically tied for fastest speed with the pilot who ultimately won the day - and about 10 km behind him. It seems that you were both ahead of the main gaggle. It wasn't clear if there were tracker-less gliders with either of you - you looked to be alone. The ultimate winner seemed to connect with better lift and maintained speed while you lost about 6kph. In a straight proportional speed scoring system you'd have scored 940 points, but IGC scoring spreads the speed points out about 2x versus straight pro-rata points allocation so you ended up with 888 I think, or almost double the points gap.

Another reason why people are well-advised to stick with the gaggle unless they are highly confident (high estimated probability) they can maintain any "lone wolf" advantage.

People back here are pulling for all you guys.

9B