What about a Wilson score for owner stats?

Question

Well we already have the Wilson score for caches and this works great!  For owners we have several lists but it would be an interesting list to see owners sorted by Wilson score!

The Wilson algorithm (for caches) is using percentage (as a quality indicator) and number of favorite points and both indicators are available for owners too so probably an owner Wilson list could be easy to be implemented?

For me I'm always looking for interesting areas for holiday spots. Of course the Wilson lists for caches is a great help but it would also be interesting if an owner is working in such an area with providing highly rated caches.

The cache list will show me the "must see hides" but of course I'm spontaneously searching for other caches in the holiday area too and being on a trip at site, just checking the live map and must decide between two tradi caches with both 20 fps it could be a helpful information if I remember that one of the two owners has a good Wilson score (which I would have checked in preparation of the stay)?

Sometimes there are owners with powertrails or many "standard" caches earning lots of fps and others having just a few caches but with high percentage so the owner by fp list isn't helpful in that case...

Answer 1

Something similar has been asked before. The problem what you have noticed in the end of your post.
It can be reformulated that averaging wilson on all caches would discourage publication of new caches by someone to keep a high score.
And it would be as you noticed a bad messurement. If you owned only very god cache with 100% FP and you published a simple trad on place to stop along a road with a nice view and tables to eat you food at. That cache would likey get a few FP but not a high %. Would that make you a worse CO? I suspect most people would say no but the stats would show something different.
There is a reason that no hide stat is and average. FP is there but not average FP per cache.

If you use the pgc live maps you can filter by wilson or click on a cache and see the wilson and %fp score directly

And I would argue that it is better to use the pgc functions and look at the the different FP stats then to look by CO

Comments

Well perhaps I'm just searching for caches in another way. Before holidays of course I check the area by Wilson list for caches and I will definitely visit the must see hides. But on top of these highlights I'm not focused on high quality caches. I'm even not using PGC on the road, just for preparation. I'm using my preferred app, check if there's a cache close to me when I've got some time beside of the family business. I would see the total fav numbers in the app but coming back to your example if there are 2 caches without many favs but I remember the cache owner with a great highlight I would probably visit the one of that owner...or even more curious to decide to not visit any of the both... :D
And I would think that an owner with almost pretty high rated caches would rarely drop a standard cache without any reason! So in your case if he puts a petling at a rest area with great view and this cache didn't get much fps the cache would be still worth a visit for me...

Why avering the Wilson score of all caches from an owner? The profile of each cacher at pgc contains fp percentage and total number of favs!? Wouldn't that be enough for the algorithm?

---

The problem is not to calculate some average FP for a hider.
The problem its that top list like that would not promote hiding caches.

Read the post from Ganja1447 in this thread
http://project-gc.com/qa/?qa=9151/total-favorite-points-received-number-of-hidden-caches&show=9158#a9158

Do you want to know who the best hiders with som measurement are in an area.
Take the top wilson list and  add all above som value to the vpgs.
Then group the VGPS by owner and is is quite obvious who has most good caches


The problem of creating a pgc top list is that is should be a good measurement.
Avragin FP/willson or calculating wilson on all hides is not a good measurement.
If you can create an algoritm to create a good measurement feel free to suggest it and it might be added

My crude litst idea above is not good for a topl list with hard cutoff etc

---

Thanks for pointing me to the discussion about "favorite points / number of owned caches".

I understand your point of view (which is also Groundspeak's in regards of the challenge guidelines) that such a list would not promote hiding caches but I do not agree! Perhaps such a list would encourage cache owners to not only hide a new cache but even more hide an interesting cache?

And that's exactly my personal(!) goal when I'm on the road and I've got time for just one quick finding. If you, "target.", would have a honorable Wilson score as an owner and I remembered that fact when I'm in touch of one of your caches looking at my app, I probably would give it a try. When I never heard about the owner, the cache has no other interesting attributes, listing etc. I would probably decide to not go for any cache at all and stay with my family.
 
We have millions of caches without a single fp - so do you really think that owners might get discouraged by _any_ owner quality indicator? :o ;)

For me a highly rated cache is a good target while for another person a spot with many caches is the preferrable target. All ways of playing the game are fine with me. My question is not about the "better" owner. Every owner is a good owner since he supports the game. For me he's doing his job even better when he takes care of his cache for not just hiding it but keep it maintained (even without a favorite point).

As long as we accept fps in general and provide such Wilson and fp lists (beside others) to make it easier for people to find their individual cache targets or preferred spots they want to go for, I don't understand why that should be negative at once for owner lists?

Good measurement? We all would agree that the Wilson list does show way better results than the top list with just total number of fp, don't we? I don't see why a cache owner should get more frustrated by a Wilson owner list than he already might get frustrated by the top list with total number of fps, total number of received logs or images received etc.? We already have a fp list for owners - so why not use the better measurement?

> It can be reformulated that averaging wilson on all caches would discourage publication of new caches by someone to keep a high score.

That's absolutely fine with me if this owner decides to not hide an additional cache just to keep his ranking - though I think's it's more likely that he would take the success from the first cache as a motivation for building another great new cache! 3 millions of geocaches and you're worried about losing a new hiding? ;)

But I don't want to insist on that suggestion...though I don't agree to the reason why you declined it.

P.S.: Just as a follow up for Ganja and his example of 2 cache owners with one cache: Wilson top list is just showing caches with more than 10 fps. One easy thing could help in Ganja's example: Just show owners with at least 2 or more  hidden caches...but I would be fine with showing all owners.

---

I am not involved in the development but ganja1447 is the developer and founder of the site.
What i wrote is based my opinion that is the same as he have said in response to other similar questions.
In ganjas example the relevant is not that it is one or tow hides. It is an example with with extrem values to show the problem in an obvious way

If I am not misstanken there is only one stat on the website that are the average over many finds/caches and that is the the bade gen badge "The Author". And I suspect that is a reason for long copy and paste logs on many caches. I think the result of the badge is negative for geocaching. It is not original from pgc but from the GSAK script.

All other stats are per cache or total of X


I don't think that kind of stats would stop people hiding bad caches. What I suspect is that someone that creates really good caches might chose not publish an cache that are fare above the average of a all caches but below his/her average. I would miss that hide.

This problem is two parted. One part is to encourage hiders to create good caches and the second part is to find good caches.
I think average FP stat will hurt the first part and be useless in the second

In the first part doing averages on hides would result in fewer hides for the users that create good caches and change nothing for users with bad hides.
Only doing sums will not discourage hides.
Creating average stats will not improve the situation.

The second part of finding good caches.
Your example with two caches with similar FP and you know that one hide has better caches. I done thins that use case is worth the potential problem above.
And if you have to look it up it is better to lock at the wilson or %FP score.
An average score would be useless.

Taking the best caches and look what name is in the list will provide informations like that you ask for and you can do it now.

The improvement that would be useful would be for groundspeek to export the %FP or even better FP,FP for current premium caches and  Total current premium finds.
The values already exist and is used for their %FP on the cache pages and can be used to calculate wilson score.
But %FP has never been used by gc.com It is listed on caches but cant be used on search

I am not sure what "owners sorted by Wilson score" means mathematically.

The "Lower bound of Wilson score confidence interval for a Bernoulli parameter" is the lower value of the intervall we can assume the FP% are in with 95%(If a remembered correctly) confidence level the error.
I think the wilson on all finds and FP will give the average FP for a hide and that is not that useful. A high value will indicate something. But a lower will not necessary say something about the hide.
And you can't calculate a wilson score on the wilson value since is for binomial distributions and the wison score has more then two values

I am not that good at statistics but if there is a good to determine the best cache owner it might be added. I don't know if there is a good metric to use


The short conclusion is that a top list like that will not show the best cache owner and would be useless. And it will discourage good hides.
And the only existing average stat give bad results. Long meaningless copy and paste loge. And with it was created it was likely intended to get people to write long and good unique logs for their find.
I suspect a average on cache FP will also have bad effect.

There is a big difference with average and the other top list. On the other everything you do will increase your result. With average it will not be like that.

If you look at gc.com everything is positive if you do something. There are no negative votes on a cache. Challenges that will stop you to log a caches is not allowed etc

---

I don't want to bore people with an endless discussion but you contradicted yourself:
"{Wilson owner list}...it will discourage good hides." as you mentioned before: it will discourage BAD hides so that the good hider will not lose his ranking. (Ganja's example; of course I know who he is).

I still don't believe that a good hider will skip a new cache due to potentially losing his ranking position. More likely he will put some more effort into the hide and try to climb up. But that's the question about the glass of beer being half full or half empty... Think positive! ;)

And the example of the average log length for the author badge completely misses the point: Any owner could improve his ranking in an averaged owner list just by launching additional good hides instead of any logger who can climb up to a diamond author badge by producing tons of copy & paste waste logs filled with nonsense...

Sorry I'm still not convinced. ;)
Owners providing "average" caches have no mettle for a good ranking in such a list and owners providing quite good caches, having a good ranking in that list are exactly what I'm looking for and if he/she really would decide to not publish a new cache which is crap because it would potentially downgrade his ranking - that's fine with me. I can pick up tons of superflous caches everywhere. The lack is at good hides...and that would be supported by such a list...