Jump to content

[FEATURE] Use Favorite Percentage, not raw count


ATXTracker

Recommended Posts

I just read the blog article on sorting by favorite points.

 

(http://blog.geocaching.com/2012/03/Groundspeak-weekly-newsletter-march-22-2012/)

 

IMO, sorting should be by the percentage of finders who favorited the cache, rather than the total number of favorites. I think the idea of favorites is to quantify the quality of a cache as perceived by premium members. Sorting by total number of favorites weights older caches over newer ones, which is a little redundant with the sort-by-date feature.

 

The percentage should be the default value shown everywhere that favorites is displayed, and hovering should show the total count. Percentage normalizes the stat and disregards finds and age.

 

In one search, I see a cache from 2001 with 44 (16%) favorites ranked above a cache from 2009 with 31 (48%).

 

(Yes, this is a re-post of a thread previous flagged as a SUGGESTION that was not picked up. My mistake.)

Link to comment

I disagree with the general assertion that the percentage is any more statistically valid than the raw count.

 

Many of the PMs that visited my caches long ago have dropped out of the games before favorites ever came along and very few that are still around bothered to go back and pick through all the older finds and award any favorite points. I know because I have talked to some of them.

 

Also - your method would show a brand new cache with 1 visit and 1 fav vote as a 100% or even 2 visits with 1 fav at 50% --> do you really think that is statistically valid??

 

I try not to look at them as a "ranking" above or below anything else but rather as an indication that it may be a cache I would enjoy. The percentage helps me to narrow down to the most awarded among the most visited.

Link to comment

I agree with both points.

 

1) Finders who dropped out or don't have the ability to favorite a cache for some reason should not be included in the denominator when calculating the percentage. This error does go to zero over time, however.

 

2) The bigger issue is caches which are very new. I don't have an immediate solution to this, but it does highlight the problem of of time-binding the metric. This is sort-of the same problem with old caches having an advantage due to their age. I still think that percentage is better than raw count, though perhaps very new caches need some filter as well.

Link to comment

I agree with both points.

 

1) Finders who dropped out or don't have the ability to favorite a cache for some reason should not be included in the denominator when calculating the percentage. This error does go to zero over time, however.

 

2) The bigger issue is caches which are very new. I don't have an immediate solution to this, but it does highlight the problem of of time-binding the metric. This is sort-of the same problem with old caches having an advantage due to their age. I still think that percentage is better than raw count, though perhaps very new caches need some filter as well.

 

There are a lot of factors that can impact the raw number. Another one, that can be quite significant is simply the relative cache density and the number of geocachers in which it's located. A cache located in a cache rich area, with a lot of local geocachers has the potential for a higher number of favorites than a cache located in a cache sparse area. The number of available favorite points that can be awarded is based on the number of finds one has. Those living in a cache dense area have more caches to find, thus are likely going to have a higher number of favorite points they can award. Those same areas are likely to have more local geocachers than a cache sparse area as well. Awhile back I compare two caches from different areas that were essentially similar (same hide style, placed around the same time). The one in the cache rich area (which had a lot more finds) had a *lot* more favorites than the one in the cache sparse area.

Link to comment

I think what's there now is sufficient - you can see a percentage if you click on it. A lackey explained that it's less server load to show a raw number then it is to show a percentage. It's there for those click on it. If it were default it would have to be calculated for every cache every time.

 

The percentage would be more accurate, but require more site scraping, if the percentage were calculated against Premium Members who have ever awarded a favorite point, as opposed to Premium Members who have found the cache. These are often not equivalent numbers.

 

Extreme caches tend to run high on percentage favorites, as those caches are rarely found by anyone who isn't inherently interested in extremes - whether it be a tough physical outing, or really demanding puzzle. I have a 100% favorited multi. It has 2 whole finds ;-)... I don't think that number tells you much, myself.

 

Old virts tend to have high numbers, and lower percentages. Another example where neither number says much.

Link to comment

Favourites are a pretty useless feature really.

 

I am a bit of a grouch mind.

 

Way back, I did a brilliant cache involving a 6 mile hike across moorland and had to investigate a crashed aeroplane for clues before climbing a steep rock face before finding an ammo can hidden where there are fantastic views (Close to the Edge - GCYQNM). It has been there nearly 6 years and has been found 31 times and now has 11 (mostly retro) favourite points. OK that IS 30%.

 

Where I live a film canister tossed into the bottom of a hollow tree by the side iof an urban road with a scrap of paper attracts favourite points at the drop of a hat.

 

We have a lot of new cachers in the area. I hope that they are able to have a great caching experience like Close to the Edge and don't think that film cans in trees by the side of an urban road ARE great caches.

 

PP

Link to comment

I actually agree to some extent with the OP. Recently whilst sorting by favourites and looking at the placed date, the top favourite scorers were those of tested time, some new caches pick up favourites QUICK, and have a high find to favourite ratio. In these circumstances they obviously have something special.

 

Obviously there must be systems in place to catch the "1 log / 1 favourite" phenomenon, but i think this could be done effectively, something along the lines of (and don't kill me for this) (favourites/finds)/time.

Edited by Z3ROIN
Link to comment

Please no un-favorite points. As these forums are proof, when semi-anonymous people are online, many tend to be more negative and mean then they would be in person. Negative incentive has proven to turn amiable communities in to nasty, competitive communities quickly. Why do you think it is so hard to dis-like something on facebook? Not voting is a nicer way to hate.

Link to comment

IMO sorting caches by the favorite percentage would be way more significant than the ranking of total FPs. As was already pointed out a very low (3-5) numer of blank finds (not official finds, but ones who didn´t favorite the cache and are used for the calculation of the percentage) would prevent very new caches with less than a handful of finds to be at the top of the list. In order to enable caches to gain a 100% FP/log ratio (or to show the true ratio anyway) the blank logs could be ignored after enough people have found a cache.

Of course there will always be examples where some caches are at a disadvantage with this system, but in general I think it would be easier to find very good caches if the sorting algorithm would work with the percentages. I absolutely would love to see this feature coming up!

Link to comment

Those who are asking for percentages instead of raw favorites are attempting to turn favorites into something it isn't meant to do. Favorites do not show that one cache is better than another. They simply indicate that for some number of cachers, this cache was in their top 10%. Caches that get a few favorite votes mean that a few cachers felt this cache was exceptional.

 

I understand that a cache that has been out for a while and in in an easily accessible location and perhaps there was a mega event nearby will have a lot more favorite votes than a relatively new cache in hard to get to location that perhaps involves solving a difficult puzzle. The raw number does not tell you which is better. But, neither does the percentage. The cache in the easy to get to location has attracted a broader cross section of geocachers. That means some will be cachers who don't particularly like to give out favorite points to this type of cache, even though they might have enjoyed it. They may save their vote for caches that are extra special for them. Or they may just not use all their favorite points. On the other hand, difficult to get to caches, and caches with difficult puzzles, tend to attract cachers who enjoy this particular kind of find. And since these are much rarer to find than easy to grab urban hides, they are more likely to get a higher percentage of favorites.

 

If there was such a thing as an "average" cacher, and I truly felt that my preferences were those of the "average" cacher, they I might believe that percentages would tell me something. But if there is an "average" cacher, I doubt very much that they like remote hard to get to caches that required solving a difficult puzzle. So, they would be disappointed to find so many caches with 80%, 90%, or 100% favorite votes that they have no desire to find.

 

What we have now allows you to search for the caches with the most favorite vote. You can look at the cache page to see how often this cache gets found and have some idea if the number of favorite votes is high or low for that type of cache. (The system even will show you the percentage of premium finders who gave this cache a favorite vote, though I'm not sure why this is needed). If the type of cache is a type of cache you might enjoy and the number of votes seems high for this type of cache and how old it is, then you'll probably find this to be a better cache, even if it isn't in your top 10%.

Link to comment

Favorites do not show that one cache is better than another.

 

In my experience they do, especially when it's heavily favourited.

 

I'm not saying percentage weighting should replace raw count, there's just been a few times when i've thought i'd like to search that way.

Edited by Z3ROIN
Link to comment

Favorites do not show that one cache is better than another.

 

In my experience they do, especially when it's heavily favourited.

 

I'm not saying percentage weighting should replace raw count, there's just been a few times when i've thought i'd like to search that way.

 

+1 - well said.

Link to comment

I think the suggestion to add the % of favorite points in the search list is an excellent one, and relativly easy to accomplish.

I would suggest to leave this measure blank for a chache that has less then f.e. 5 founds, to eliminate the issue of having a cache with 1 found and 1 favorite getting 100%; but even if you don't do that it would be a nice feature.

 

Many cachers now use the sorting-by-favorites as a way to discover the best caches in an area, this would be a much better way to do it.

 

If you don't like favorites, or don't like precentages of favorites, no problem, don't use it. For us that do like and use it, a % would simply work a lot better,

Link to comment

Those who are asking for percentages instead of raw favorites are attempting to turn favorites into something it isn't meant to do. Favorites do not show that one cache is better than another. They simply indicate that for some number of cachers, this cache was in their top 10%. Caches that get a few favorite votes mean that a few cachers felt this cache was exceptional.

 

Perhaps, but when favourites were introduced they were granted retrospectively. So someone with my find count (about 1500 at the time) suddenly had 150 points to give away. More active cachers would have had more favourite points than they are ever likely to use.

 

(cut for brevity)

 

What we have now allows you to search for the caches with the most favorite vote. You can look at the cache page to see how often this cache gets found and have some idea if the number of favorite votes is high or low for that type of cache. (The system even will show you the percentage of premium finders who gave this cache a favorite vote, though I'm not sure why this is needed). If the type of cache is a type of cache you might enjoy and the number of votes seems high for this type of cache and how old it is, then you'll probably find this to be a better cache, even if it isn't in your top 10%.

 

Personally I think the favourites system as it stands is a(nother) half-baked idea that could have been really good but turned out to be largely useless. As you've said (I just realised it's in the bit I cut for brevity), an easy virtual near a mega-event or in a large touristy town will chalk up a lot of favourite points. As a rule I dislike nano caches but every once in a while I find one that I give a favourite point to - usually because it was a puzzle and I enjoyed the puzzle.

 

It would make a lot more sense to me if I could search for caches that a particular user had tagged as a favourite. That would give me the ability to say, for instance, that I'm stopping in an area where a caching buddy has been before me so I want to see which caches he considered to be the best in the area. That would tell me a lot more than the fact that 38 people from a random bunch of strangers liked a particular cache, which could still be a film pot in an ivy covered tree but released for a mega event two years previously.

Link to comment

It would make a lot more sense to me if I could search for caches that a particular user had tagged as a favourite.

 

In some sense you can do that, but search might not be a proper term as the possibilities are so restricted and old-fashioned.

You can look only at the full list of favourites of each user who has one, but you cannot search within this list e.g. with respect to coordinates, keywords etc

 

Cezanne

Link to comment

It would make a lot more sense to me if I could search for caches that a particular user had tagged as a favourite.

 

In some sense you can do that, but search might not be a proper term as the possibilities are so restricted and old-fashioned.

You can look only at the full list of favourites of each user who has one, but you cannot search within this list e.g. with respect to coordinates, keywords etc

 

Cezanne

 

That's the trouble, something that could have been quite useful became a half-baked implementation that's little more than a gimmick.

Link to comment

I think what's there now is sufficient

 

No, it is absolutely not.

 

Proof:

 

http://www.geocaching.com/seek/cache_details.aspx?guid=33be88f2-324c-485d-81f0-e1aa7451f3ec

972 FP, 49%

Metal tube in the center of Dresden, attracts tons of tourists

 

http://www.geocaching.com/seek/cache_details.aspx?guid=33be88f2-324c-485d-81f0-e1aa7451f3ec

Touristic-Spot: Hackesche Höfe

908 FP, 51%

film canister in a sock in a crowded place in the center of Berlin

 

Both hides are funny, but quick and easy to find.

 

http://www.geocaching.com/seek/cache_details.aspx?wp=GC2KD4Z

8 FP, 57%

great multi with a station which is worked out extensively

 

http://www.geocaching.com/seek/cache_details.aspx?wp=GC3H09K

8 FP, 42%

10 mile hike through nature, dared only a few

 

A systematic way of finding caches like the latter quick and disthing them from caches like

http://www.geocaching.com/seek/cache_details.aspx?wp=GC3D5XB (11 FP, 3%, film canister) is required.

Link to comment

Many cachers now use the sorting-by-favorites as a way to discover the best caches in an area,

 

Indeed. I now have found the caches with lots of FP in my home city Leipzig and most of them in Seattle and Berlin.

How can I find the caches with high % and few finds?

Why does Groundspeak not help with this?

Edited by Der Wald-Pirat
Link to comment
an easy virtual near a mega-event

 

Indeed.

 

http://www.geocaching.com/seek/nearest.aspx?u=Matze1955&sortdir=desc&sort=fav

 

Matze1955 is an owner who made several great caches in Meißen near Dresden who received many FP.

 

His best cache has 1311 FP, the next one 434 FP. What is the difference?

 

In the summer of 2012, these caches had about 200 FP vs. 100 FP.

So, Matze19555 took his best cache http://www.geocaching.com/seek/cache_details.aspx?guid=7d636f41-0ee3-4056-9f08-c7d66f6240eb to the mega event in Leipzig where 3000 cachers came and the cache was exposed and could be found and logged.

While all them logged within a month after the mega, the number of FP for the cache rose

from 200 to 1000. It has gained 300 FP since, same as his number 2.

% has not changed significantly.

 

Several other caches in the north and center of Leipzig gained the overwhelming majority of their FP just from the

3000 mega visitors:

http://www.geocaching.com/seek/cache_details.aspx?guid=50130805-acb4-40f9-8149-92ff7aa7ee82

http://www.geocaching.com/seek/cache_details.aspx?guid=8a9f806d-d7fb-4a2e-91f3-8dd44d88111c

http://www.geocaching.com/seek/cache_details.aspx?guid=34f7faa7-5b68-4227-b163-284bacdc4777

http://www.geocaching.com/seek/cache_details.aspx?guid=061f41a6-ca72-49a1-b0c6-02d5529a310b

Link to comment

Nobody ever claimed that a cache with more favorite points is better than one with fewer. The purpose of the favorite point is to indicate that some geocacher liked the cache enough to give it a favorite point. But, as has been noted, a cache that gets found more often is likely to have more favorite points than cache that is rarely found. Similarly an older cache that has been found as many times as some newer cache may have fewer favorite points because people who found it in the past may be less likely to go back an give it a favorite point than someone who found the newer cache more recently.

 

Since the number of favorite points is not a good indicator of "goodness" of cache, some people have argued that the ratio of favorite points to finders or of favorite points to premium finders is a better indicator.

 

I happen to believe that the ratio is no better an indicator of "goodness" than the raw count and may even be a worse indicator.

 

Certain caches self select for finders - for example a difficult puzzle or a long hike. Not everyone will do these caches and those who do are more likely to enjoy them. So these caches will tend to have a low number of favorite point but a high ratio of favorites to number of finds. If you are not interested in tough puzzles, long hikes, or caching at night in cemeteries, you might find some caches with high ratios that you are not going to enjoy.

 

New caches may also have higher ratios that a similar older cache, because those geocachers who first go to a new cache are more likely the ones to give favorite points. After the regulars have found the cache, a lower ratio of finders will give out points. Does this make the older cache less "good" than the newer one?

 

I find a far better way to use favorites is to find a few geocachers who have given favorite points to caches that I also liked. Then I can see what other caches they liked and use these as recommendations of what to find.

 

The one problem with this is when visiting a different city. There I may not recognize anyone, so I have no way to judge their favorite lists. It doesn't surprise me that some of the caches with the highest number of favorite points are ones in city areas frequently visited by tourists. Most geocachers have limited time to search for caches when visiting someplace. They may only find a few caches and these will be ones easy to get to, near their hotel or near the train station. Caches in city centers with a high number of favorite votes probably stick out as the ones for visitors without much time to go after.

 

Certainly one can argue that a visitor who plans to spend more time geocaching might want some help in selecting the kinds of caches they would like to find. If they like puzzles, which ones should they look at and try to solve? If they like hiking, are there any trails with some exceptional caches to visit? If they enjoy history, what caches are pointing out interesting historic locations? It is silly to expect that favorite points alone or as a ratio of finds will help solve this problem. Instead, I would suggest trying to find caches that you might like some other way. Then if you still need to narrow down which ones you will find, you can look at the favorites. It may be that once you get to this level, that a ratio might provide a better estimate of which caches you will enjoy more. I'd argue that once you get to this level, just knowing which caches got more than one or two favorite votes is already enough to find caches you are likely going to enjoy.

Link to comment

The fact remains, that FPs mean different things to different people.

 

The result is, that as a guide of any sort they are pretty useless since you have no idea what prompted the awarder to award the point. Was it awarded because it could be grabbed without straying far from the car and another smiley was earned for minimal effort or because it required half a day trekking through stunning countryside?

 

Pointless.

Link to comment

I'd argue that once you get to this level, just knowing which caches got more than one or two favorite votes is already enough to find caches you are likely going to enjoy.

 

Certainly not true in countries like Germany and Austria where FPs are used much more frequently than e.g. in the US.

 

It should be no problem at all to offer a choice according to which number (total number of FPs or ratio) someone wants to obtain results, but once again Groundspeak offers a weak service which is not at the height of what is possible nowadays. There is so much more that could be done including automatic recommendations of the type "here is a list of caches that you might enjoy as others who have a similar favourite pattern than you have enjoyed them" etc

 

Cezanne

Link to comment

I'd argue that once you get to this level, just knowing which caches got more than one or two favorite votes is already enough to find caches you are likely going to enjoy.

 

Certainly not true in countries like Germany and Austria where FPs are used much more frequently than e.g. in the US.

I don't see how this invalidates what I said.

 

It may be true that in Germany and Austria favorites are used more, but your own posts elsewhere point out the while many caches are popular and get lots of visitors, others are set up to appeal to only select groups of geocachers and may get very few visits.

 

You seem to be looking for scale on which to compare caches the have limit appeal to ones that have many. I don't believe that you can do this. If anything a ratio will unfairly benefit the caches with narrow appeal as these caches attract only finders who are likely to award them favorite points.

 

Of course if you are using some other method to preselect a certain class of caches - say multi-caches that take you on a long trek of many stages - then you may find that within this class ratios work better for you. I contend they are still inaccurate as older caches are penalized since finders may not go back and award favorites to caches they found in the past.

 

My comment was that once you have preselected caches, you will likely have more fun on any of these than you might on an urban hide, and that any of these caches that have just a few favorite points are already among the best of this category. There is really no need to pick the absolute best cache - even if you have only time for one cache. And the cache with the highest ratio may not even be the best for you.

 

It should be no problem at all to offer a choice according to which number (total number of FPs or ratio) someone wants to obtain results, but once again Groundspeak offers a weak service which is not at the height of what is possible nowadays. There is so much more that could be done including automatic recommendations of the type "here is a list of caches that you might enjoy as others who have a similar favourite pattern than you have enjoyed them" etc

 

Cezanne

It turns out that sorting by ratio is a problem. It involves doing a calculation based on two numbers - favorite points and number of finds (or possible number of finds by premium members - which makes it three variables instead of two).

 

There are ways to compute the ratio and using it to sort data. However determining what method to use will depend a lot on how this will be used.

 

One scenario is that Groundspeak spends programming resources to implement the OP's idea. And once in place, the OP and others find that this isn't as helpful as they though it would be. Now they ask for a different number; already it's been suggested that caches with less than 10 finds don't have a ratio. Instead, let's understand better the motivation for requesting this and whether it makes sense to implement it.

Edited by tozainamboku
Link to comment
if a cache is good

 

And that is whole reason why this whole debate and the idea of favourite points is flawed and pointless.

 

One man's good, is another man's pain in the arse.

 

Smiley monsters are not likely to think that a cache that takes 2 hours to complete is good.

 

Those of us who are thrilled to find an ammo can half way up a mountain after trekking through streams and bogs, won't think that a film cannister chucked into the bottom of an urban tree is good.

 

You don't need favourite points to tell you which is which.

 

PP.

Link to comment
if a cache is good

 

And that is whole reason why this whole debate and the idea of favourite points is flawed and pointless.

 

One man's good, is another man's pain in the arse.

 

Smiley monsters are not likely to think that a cache that takes 2 hours to complete is good.

 

Those of us who are thrilled to find an ammo can half way up a mountain after trekking through streams and bogs, won't think that a film cannister chucked into the bottom of an urban tree is good.

 

You don't need favourite points to tell you which is which.

 

PP.

 

I think Fav Votes are the best thing to happen to GC since PQs. Sure it's not perfect. Most of the favs in my area are off limits for me - high terrain ratings, tree climbs, boats, kayaks required. Throw a film can 15 feet up a tree in the middle of a small river island and people will go ga-ga. No problem really. I have found I like the caches that get a few votes. A percentage of around 20 to 35%. They tend to be quality swag size containers, well maintained in a nice location and usually have terrain ratings of 3.5 or under. Before FPs it was so hard to find them among the chaff.

 

I like the current system. But perhaps Groundspeak could offer both options - sort by Raw FPs or sort by Percentage.

Link to comment

Percentage is a very useful mathematical concept :)

Percentage is a very useful mathemtical concept when used properly. Percentage (or more generally a ratio) tells you something about the relationship between two measurements. In this case it would be about the relationship between number of finds and number of favorites. It provides a way to estimate for a partcular cache how many favorite points it would have if it got a certain number of visits. This can give a way to take two different caches and ask the question "If these two caches were visited by the same number of people, how many favorite points would each have."

 

But you need to know this is an estimate and that the estimator has certain biases.

 

If a cache has very few visitor because it appeals only to a few geocachers it may get more (or sometimes fewer) favorite votes than a cache that has broad appeal. It has also been pointed out that many older caches have a lower number of favorite votes than expected because finders of the cache in the past may not go back and award their old favorite vote. And there are difference caused by geocachers not always giving out all of their favorite points and by cache density where the favorite votes that are awarded end up being split among more caches.

 

These biases occur not only between caches but even over time they effect a single cache. A cache with 90% favorites today may get visited by hundreds of mega-event attendees tommorow and see the ratio drop to 50%, even though many of the new visitors might give a favorite vote.

Link to comment

Percentage is a very useful mathematical concept :)

Percentage is a very useful mathemtical concept when used properly. Percentage (or more generally a ratio) tells you something about the relationship between two measurements. In this case it would be about the relationship between number of finds and number of favorites. It provides a way to estimate for a partcular cache how many favorite points it would have if it got a certain number of visits. This can give a way to take two different caches and ask the question "If these two caches were visited by the same number of people, how many favorite points would each have."

 

But you need to know this is an estimate and that the estimator has certain biases.

 

If a cache has very few visitor because it appeals only to a few geocachers it may get more (or sometimes fewer) favorite votes than a cache that has broad appeal. It has also been pointed out that many older caches have a lower number of favorite votes than expected because finders of the cache in the past may not go back and award their old favorite vote. And there are difference caused by geocachers not always giving out all of their favorite points and by cache density where the favorite votes that are awarded end up being split among more caches.

 

These biases occur not only between caches but even over time they effect a single cache. A cache with 90% favorites today may get visited by hundreds of mega-event attendees tommorow and see the ratio drop to 50%, even though many of the new visitors might give a favorite vote.

 

Thanks tozainamboku for the well written and insightful critique. I appreciate that you made your arguments without belittling my suggestion, as I sometimes see on this and other forums.

 

I agree completely with the complexity you explain. These considerations are what I alluded to when I noted 'No one system is perfect.' I still think for most/many caches, the percentage metric is less flawed than the raw count, and it would be great if we could have search functionality to find caches with high favorite percentages. Obviously, I would support an effort by Groundspeak to normalize for the age of the cache, density of the area, and other factors, to try and better suggest great caches. In the past I've also proposed a recommendation engine that could take in to consideration which caches you find and favorite, and recommend a customized list of caches to each cacher. There are lots of great opportunities to leverage favorite points.

 

Happy Caching!

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...