Brainstorm for a metric for the "interestingness" of a hit

Do you want to test the possibilities of this forum? Want to test out your new signature? You can do it here.

Moderators: Crazy Bob, Fons, avij, Phaseolus

User avatar
avij
Forum Moderator
Forum Moderator
Posts: 5611
Joined: Mon May 27, 2002 10:45 pm
Location: Helsinki Finland
Contact:

Re: Brainstorm for a metric for the "interestingness" of a h

Postby avij » Tue Feb 26, 2013 7:09 am

I had originally thought that the bundle hits would be scored so that the first hit in the bundle would get the "normal" score and the next hits in the bundle would get a significantly lower score, so that in the typical case their score would drop at least one "star" ranking. Considering a typical user browsing at * * *, this would be all right if the first hit was scored * * * and the others * *, but if the score of the first hit was * * * *, then the others hits might be * * *, bringing again "clutter" to the user browsing at * * *.

A bigger problem with the decreasing scores within a bundle is if any of those hits become a triple in the future. It can't be so that the score of that triple would depend on the position of the note in the first bundle. It kind of works now in the current setup, where the moderated hit in a bundle loses its moderation the moment it becomes a triple.

So, this brings us a new design goal -- all the hits in a bundle (or generally speaking, a group of hits) should have the same score. This in turn means that we'll need to bring in support for the grouping of hits. In practical terms this would mean the hit list would be modified to show "100 x :note-5:" in case of such a group hit. The hit report page would then show a list of all the notes in the group (or a summary of them), showing that these 100 x :note-5: travelled from place 1 to place 2. We could have a rather strict criteria for forming these groups, such as all hits having the same denomination, all entered at the same locations at roughly same times by the same users. If there are big enough differences, the other hits could end up in a different group.

We could also take this grouping into account when determining how often users get hits with each other. I'd say that if someone gets a 100 x :note-5: bundle from a bank that was dropped there by some other tracker, it should count as "one occurrence" when determining the hit frequencies. There are also people who would then say that this group of hits would actually count as one hit (approximately like it's now), but given the recent feedback about this I'd be inclined to count all those hits within a group as hits.

There are a lot of hairy implementation details for this scheme, which I've not described yet. Perhaps I'll do so at some later point if needed.
Money makes the world go round. We track how the money goes round the world.
EBT Tech WG leader. Do not PM me if your question is not related to Tech WG or the association.

User avatar
dserrano5
Euro-Master
Euro-Master
Posts: 3773
Joined: Sun Jan 31, 2010 11:59 am
Location: Madrid, ES

Re: Brainstorm for a metric for the "interestingness" of a h

Postby dserrano5 » Tue Feb 26, 2013 8:52 am

I'm slightly against this grouping, if only because it complicates things too much. We can start with a simple scoring system and maybe later add something about the bundles.

For an initial scoring of bundles, I'd suggest scoring them all individually, then keep the most interesting one as-is (e.g. there was some case in which a user requested unmoderation of a bundled :note-200: and moderation of the first note in the bundle, which was a regular fiver if my memory serves me well) and cap the rest at 50 points for example. Which means, all hits in the bundle should not have the same score, I think at least one note would have to be highlighted—it's a hit after all! Then if some of those becomes a triple in the future, it's just a matter of reassessing its score without affecting the other bundled notes and without having to do anything with groups.

OTOH I can see that grouping could be useful to determine hit frequencies between people. But again, I'd leave that for a later stage.

User avatar
avij
Forum Moderator
Forum Moderator
Posts: 5611
Joined: Mon May 27, 2002 10:45 pm
Location: Helsinki Finland
Contact:

Re: Brainstorm for a metric for the "interestingness" of a h

Postby avij » Tue Feb 26, 2013 1:20 pm

The grouping doesn't necessarily mean too much additional complexity. We could treat every hit as a group of hits, with the vast majority of the groups having only a single hit in the group.

The capping method could work as well.
Money makes the world go round. We track how the money goes round the world.
EBT Tech WG leader. Do not PM me if your question is not related to Tech WG or the association.

User avatar
CarlosManta
Euro-Master
Euro-Master
Posts: 3059
Joined: Mon Jan 07, 2008 6:11 pm
Location: Porto, Portugal
Contact:

Re: Brainstorm for a metric for the "interestingness" of a h

Postby CarlosManta » Tue Feb 26, 2013 3:22 pm

dserrano5 wrote:For an initial scoring of bundles, I'd suggest scoring them all individually, then keep the most interesting one as-is (e.g. there was some case in which a user requested unmoderation of a bundled :note-200: and moderation of the first note in the bundle, which was a regular fiver if my memory serves me well) and cap the rest at 50 points for example. Which means, all hits in the bundle should not have the same score, I think at least one note would have to be highlighted—it's a hit after all! Then if some of those becomes a triple in the future, it's just a matter of reassessing its score without affecting the other bundled notes and without having to do anything with groups.

When you look at the latest proposals, the score already factors in the number of Hits between the users, and the number of Hits between the locations.
Let us day user A collect bills during the week, inserts them in EBT, and then deposited them on the Bank next Monday.
User B withdraws on Tuesday a bundle with several bills from user A.
With the score proposed, each bill gets an individual score, an they are all different.
The disadvantage is that the score received depends on the order the bills are inserted, i.e., the last bill inserted is the more penalized for more common Hits, which for me is somehow undesirable.

The way to go around this, in my opinion, is that for all the factors in the equation, these should be taken from day (D-1). If the Hits are registered on Tuesday, the score would be calculated based on the rations effective on Monday. Then it would make sense to cap the Score, with the most interesting bill keeping the score unaffeted, and the other bills getting a score S=S'*(1/Bd), where S' would be the individual score, and Bd the number of bills in the bundle.
If a hit becomes later on a triple, then the score is recalculated, wihtout this factor.

Note: Users may feel tempted to insert "one bill per day" from those bundles.

User avatar
avij
Forum Moderator
Forum Moderator
Posts: 5611
Joined: Mon May 27, 2002 10:45 pm
Location: Helsinki Finland
Contact:

Re: Brainstorm for a metric for the "interestingness" of a h

Postby avij » Wed Feb 27, 2013 3:40 pm

Just a quick note about the schedule. I have a deadline of 11th March for one study project of mine, and I don't see myself having much time for analyzing or commenting the published suggestions prior to that date. We can proceed to phase 2 after that date, depending on how much material we've gathered by then and if they still need some thinking. Thanks for your understanding.
Money makes the world go round. We track how the money goes round the world.
EBT Tech WG leader. Do not PM me if your question is not related to Tech WG or the association.

luckbox
Euro-Regular in Training
Euro-Regular in Training
Posts: 148
Joined: Fri Aug 05, 2011 10:20 am

Re: Brainstorm for a metric for the "interestingness" of a h

Postby luckbox » Wed Mar 13, 2013 11:39 am

For me a system with 3 ratings would be enough, I am thinking about the following:

1 star = A not very interesting hit between users who have most of their hits with each other (for example when I have 150 of my 1000 hits with the same user and also for hit bills which traveled together in a bundle)
2 stars = An average hit, this should be about 80% of all hits (for example a domestic hit or a quite common international hit like Belgium - Netherlands or Austria - Slovenia)
3 stars = All Multiple hits * and less common international hits (for example Belgium - Slovenia, Portugal - Italy or Netherlands - Finland)

In my opinion it would not be fair to rate a hit less interesting (for example 2 out of 5 stars) when a bill only traveled 20 km and the users don't have many hits together.

* For triple hits there can be an exception if 2 of the users have more than 15% of their hits with each other, these could for example be rated as 2 stars.

User avatar
tralla
Euro-Regular
Euro-Regular
Posts: 210
Joined: Wed Apr 27, 2005 5:23 pm

Re: Brainstorm for a metric for the "interestingness" of a h

Postby tralla » Wed Mar 13, 2013 2:50 pm

All hits should be marked "uninteresting" or be moderated that satisfy one or more of the following conditions:
- :note-5: Five Euro notes.
- One of the users in the chain has entered more than 100,000 notes.
- Hits inside Germany or Finland.
- Hints involving the capital city of a Euro country.

The threshold for condition 2 might be lowered to 30,000 notes; that would include all my hits, too.

See what I'm hinting at? ;-)
1 hit = 1 hit
Tralla
Meine EBTST-Statistik (danke, dserrano5)
Meine nigmm-Statistik (danke, tt.01)
My little gallery of stamps and scribbles on Euro notes.

User avatar
LITO
Euro-Master
Euro-Master
Posts: 3598
Joined: Fri Dec 28, 2007 12:09 am
Location: Bourg la Reine , France

Re: Hit interestingness scores (phase 2)

Postby LITO » Wed Mar 20, 2013 11:51 am

So now , how interesting is my hit :?: :?: :?: :?:
Image


Guys , I really do not think it is necessary to implement a formula to figure out how interesting a hit is , I do believe any hit is interesting or at least acceptable as long as it is fortuitous .
I know you are trying hard to improve the system but in that case I think you are going the wrong way.
Tracking notes is not a competition between users and there is no need to grade their tracking , tracking notes is a personal game which should not be judged by others .
Users are fully aware of the probabilities of their hits and get satisfaction in accordance , there is no need to frustrate any user by downgrading his hits, this has nothing to do with the purpose of €urobilltracker.
Every user should have the the right to be happy with his hits without any restriction.

[edit/avij: moved from the other topic to here]
Lito en dilettante
URL: http://my.flightmemory.com/airlito

User avatar
CarlosManta
Euro-Master
Euro-Master
Posts: 3059
Joined: Mon Jan 07, 2008 6:11 pm
Location: Porto, Portugal
Contact:

Re: Hit interestingness scores (phase 2)

Postby CarlosManta » Wed Mar 20, 2013 12:52 pm

LITO wrote:tracking notes is a personal game.
Every user should have the the right to be happy with his hits without any restriction.

I do not agree at all.
Tracking notes is everything but a personal game.
It is intrinsically a game based on interactions. Without interactions, the tracking is not achieved, you only get a list of bills inserted.
If you talk about collections, such as coin collections, then it is more of a personal game, I agree.
However when you consider Hits, it is everything but individual, it is exactly about others, is is about connections with others.

And look as it has been said before, adding a score does not remove any information about the Hits. A user can choose to ignore the information about the score if he/she wants to. I even agree it should be an option in the website to show or hide the score (like the advanced form to enter hits)

On the contrary, the system we have today effectively "removes" Hits, through the moderation process. Again, the user can still see them, but they are not public.
If you see it, it is actually a formula: "If interesting, score=1, else score=0"

To end up, having a score is not about putting restrictions, it is about offering more information.

User avatar
Phaseolus
Forum Moderator
Forum Moderator
Posts: 13109
Joined: Fri Feb 11, 2005 2:16 pm
Location: Quelque part ou même ailleurs !

Re: Brainstorm for a metric for the "interestingness" of a h

Postby Phaseolus » Wed Mar 20, 2013 2:06 pm

I would prefer a system based on the Facebook features "I like" -"I don't like"

Just keep the table of hits as it stands nowadays and add one column with the possibility for each user to click either "I like" or "I don't like".

Then you give the possible to search through the database through the hits having gethered the most "Like"...
Ceci est une signature.

User avatar
dserrano5
Euro-Master
Euro-Master
Posts: 3773
Joined: Sun Jan 31, 2010 11:59 am
Location: Madrid, ES

Re: Brainstorm for a metric for the "interestingness" of a h

Postby dserrano5 » Wed Mar 20, 2013 3:19 pm

I wonder how many people sees the list of hits regularly. I, for one, hardly do, which means that I would "Like" or "Dislike" only a few hits. This is a problem that an automated built-in system doesn't have.

User avatar
eddydevries
Forum Moderator
Forum Moderator
Posts: 18869
Joined: Sat Feb 18, 2006 5:49 am
Location: Den Haag (NL)
Contact:

Re: Hit interestingness scores (phase 2)

Postby eddydevries » Thu Dec 12, 2013 9:09 am

Does someone have some more ideas?
Eddy - Den Haag - NL
weblog - EBT-stats 24-05-2013 - hits: 1590 +56 triples +1 quadruple = 1647.

Facebook-profile + hyves-profile + postcrossing

User avatar
LArdennais
Euro-Master
Euro-Master
Posts: 12038
Joined: Sat Jun 12, 2010 11:58 am
Location: perdu dans le Condroz belge

Re: Hit interestingness scores (phase 2)

Postby LArdennais » Sun Nov 02, 2014 10:15 am

Into EBT-website, where is this idea ? Abandonned ? "In reflexion" ? "In creation / development" ?

Thanks in advance for the info's ! ;)
http://liste.eurobillets.free.fr/ : don't miss it every day ! à zieuter tous les jours ! kijk elke dag ! :D

User avatar
Phaseolus
Forum Moderator
Forum Moderator
Posts: 13109
Joined: Fri Feb 11, 2005 2:16 pm
Location: Quelque part ou même ailleurs !

Re: Hit interestingness scores (phase 2)

Postby Phaseolus » Tue Nov 18, 2014 6:07 pm

seeing the last discussions, I guess it would be nice to be able to click on a link "I like" for any hit appearing on the EBT hit list. Doing as such, we could come up with a list of "EBT best liked hits" ; this system would also lead to a situation where uninteresting hits would not be liked at all, thus would be less ranked.

I have a second suggestion that I witnessed on wheresgeorge.com ; whilst tracking dollars, I got involved with a triple hit with 2 users that has already more than 20 hits together. This triple hit was no longer called "interesting hit", it had another denomination and it was not shown publicly, although it exists. The hit notification was referring to a website explaining that when two users have more than 20 hits together, their hits, although legitimate are no longer shown publicly nor do their count for the George Scoring system.

I think we could learn from this experience in USA and create a third hit category between "interesting hits" and "moderated hits".

I hope this contribution helps.
Ceci est une signature.

User avatar
lmviterbo
Euro-Master
Euro-Master
Posts: 4099
Joined: Thu Aug 21, 2003 5:23 pm
Location: Lisboa, Portugal
Contact:

Re: Hit interestingness scores (phase 2)

Postby lmviterbo » Tue Nov 18, 2014 6:54 pm

Although I was a firm supporter of the interestingness score and still contributed a little with ideas, I realized that finding a sound scoring system was proving to be extremely morose and difficult, given the huge amount of parameters to take into consideration, not to mention the weights we would have to give to each one of them.

I still believe that ending the dichotomy intersting / moderated would be a very positive thing.

Surrending the score to collective wisdom may indeed prove to be a good idea. But if there's a Like button I believe there would have to be a Dislike one, and maybe also a Report one, just as on YouTube. I, for one, would immediately Unlike if not Report the last 4 hits I had (which I actually reported under the Suspicious notes and users thread).