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

[-STAR-]

I think we shouldn't give that importance to the number of days. As a 100km hit in 10 days is more interesting than a 10km 100 days. (IMO!!) What do you think?

Au contraire, Monsieur!

A long-distance hit can be relatively easy obtained, by someone bringing home notes from a foreign destination, e.g. from holidays.
A long-duration hit cannot be forced in any way; you just have to wait until it happens -- or not.

Similarly, I think we should give less importance to the users themselves. What if a user has a hit when travelling with a person whith whom (s)he has previously had a hit? is the hit for that reason less interesting than if it would have occurred with a different user.

In my opinion: Yes.

The good thing of this formula is that it will give us a more close range of values, although I think that what we actually need is a formula which considerabely punish the non interesting hits and "rises up" the super-interesting hits.

Objection! No hit shall ever be "punished"; the more linearly the scale is, the better.

Another reason why I do not like the "n" is that: if both users do not want to share the "hit partners" statistics... you are betrayed. I suppouse that info is stored somewhere, but the "interestingness factor" wouldn't be fully transparent.

No problem at all - the server does know all the correct numbers. The stats that are shown at the users' pages are just filtered from the "real stuff".

BTW: this example gave me a hint... Should we consider the denomination of a note for this interestingness stuff?
I think a hit with a or or or are more or less within the same interestingness sphere. but and and specially are more interesting... as these notes are rather unfrequent. (At least for me)

Absolutely! I was about to mention this factor just yet, but you must have read my mind!
Best example is a hit I had tonight, that I'm rather proud of, because you will not get a Madrid - Wien-Hit all that often, and especially not, if it's with a Euro bill.

However, this lead us into a very complicated formula, not only because we should introduce a new factor in there, but also because... Why not consider then if the note is a R/X or something more strange M/M or whatever...

Why not? That's one of the main reasons I'd favor a one-shot solution, i.e. calculate the interestingness of a hit as it happens and not on-the-fly:
It does not hurt the performance of the server too much compared to re-iterating the formula again and again.

Rgds, Franz

[Nerzhul]

Don't worry too much about implementation details (and its impact on server performance), that's my job

Greetings from Cambodia,
Nerzhul

[R/J]

Just brainstorming here and adding my "two cents".

What makes one hit more interesting than the other and in which way (linear, less than linear, more than linear)?

Days: I would make it linear.
Distance: I think this would be less than linear.
Multipleness: (I just notice, that except for the quintuple hits, the ratio (hits with n users/hits with n+1 users) is always more or less 58)
So definitely more than linear, possibly even 58^(users involved).

Users involved: In my opinion this does not add anything to the interestingness of a hit. So I would not add it to the formula.
Countries involved: I think this is the most tricky one. We all know that international hits are more interesting than others, but how exactly. I would put this, I am unsure of the dependency of interestingness on countries. I would play safe and go for linear.
Outside the eurozone: I miss this one in previous suggestions, but I think it's a huge factor in the interestingness. We also have them in "countries involved" so they would be counted double in a way, but I think there's extra interestingness in this one. Also I would make it more than linear, as hits between two non-Euro countries are add even more to the interestingness.

Hit ratios of locations or otherwise: If using this, I would use country hit ratios not cities. I feel that a hit in a low hit ratio city in a high hit ratio country would be getting too much credit. Using them other than linear would be too messy.

So if I put all this together I get: (To be safe, multiply with a constant)
Interestingness = Constant * Days * sqrt(Distance) * 58^(Multipleness) * Countries involved * (Non-Euro Countries involved)^2 / SUM(Country Hit Ratios)

Perhaps, we would have to use sqrt(Distance+1), as in this formula, all 0 kilometer hits have always zero interestingness.

So, that is my opinion.

[De-Ker]

any news on this?

[Jes]

any news on this?

Last days I had been thinking about a new formula with complex numbers (ie: the sqr root of -1) but I think it might have been quite complicated for most people and at the end it may not be that useful.

[CarlosManta]

My 2 cents.

I am "stealing" the work from Claudio with minor changes

Id = sqrt (b) log( d t ) : [ 1 + log( n c ) ]

Multiple Hits: I = sqrt ( m - 1 ) sum ( Id )

m = multiplicity: 2 for normal hits, 3 for triples, 4 for quadruples...
b = countries involved (outside the euro-zone, countries count as 2)
d = 1 + distance in km
t = 1 + time in days

n = number of hit involving the users
c = number of hit involving the locations

log = decimal logarithm (because Claudio said so)
sqrt = square root

---

Main Changes:
- I added (+1) to the day count and Kilometer count, this only skews the metric slightly and avoids the "zero problem"
- I handled multiplicity in a different way. The issue is that defining "n" and "c" in triples and quadraples will often lead to both being "1". You can of course rewrite the formula including sqrt ( m - 1 ) in the first formula, as Claudio had it, but this way I was trying to make the point clear.

Comments:
- I included the sqrt to try and contain the dispersion of values. I am ok also with taking the sqrt out

---

As for the issue of Dynamic vs Static, my vote is for Static. But this is a weak preference. It is just more "intuitive" to me that my first Hit with Leofer40 was more Interesting than the one I got last week. Having the score floating over time is also strange. On the other hand, if the implementation is static, then any change to the formula would mean recalculating all the existing Hits...

[avij]

Here's something which I used as some sort of a "test exercise" for the moderators some time ago.
link to a pic
How would you calculate the scores in this particular case?

edit: removed attachment, added link, should work better.

edit2: another try at the attachment:

ebthitsexample2.png (178.34 KiB) Viewed 1583 times

[Jes]

Here's something which I used as some sort of a "test exercise" for the moderators some time ago.

ebthitsexample.png

How would you calculate the scores in this particular case?

All I can say is...

You do not have the required permissions to view the files attached to this post.

[avij]

Continued from here.

Thanks to those good wills who try to improve the system by implementing a formula for interesting hits, but seriously why bother create a complicated gaz factory ?

Because of your request:

To avoid this kind of problem the similar enough rule must be clearly defined and not left to the appreciation of each moderator

Once there are clearly defined rules for hit interestingness, we can pick some sort of a threshold for an interesting hit. Say, on a scale from 0 to 100, all hits which score higher than 50 are classified as interesting. The moderators would then no longer need to worry about hit interestingness, they'd only need to weed out hits that didn't really happen (intentionally incorrect locations, say). If someone has an accidental hit, say, a meeting hit, those users can request moderation of those hits via the support system. The scoring system would guarantee that the human factor would be taken out of the equation, just like you requested.

Things usually work better when they are simple , and the simplest way is to consider every hit as interesting unless it is friendly.

How does an outsider determine that a hit is "friendly" (an ambiguous term in itself)? Would that decision be made by a moderator or the users involved in that hit?

If someone gets a bundle of already entered fivers from a bank and gets 100 hits, do you consider all of them interesting? If you answer "no", you're already on the slippery path of exceptions. Once you have one exception to the "simple" rule, you'll eventually end up coming up with more and more exceptions to the "simple" rule. Been there, done that.

[LITO]

Sorry you did not understand me , my request is not for a complex system which defines whether a hit is interesting or not , no I just want something simple that works the same in every country .

If a user gets 100 hits from a bundle in a bank , I shall not decide if this is interesting or not , it has to be the hit-maker's decision to consider them uninteresting , this is not my business I am not competing against him or any other , I just want to know where my notes come from and go to, that is why I came here on EBT.

[avij]

Sorry you did not understand me , my request is not for a complex system which defines whether a hit is interesting or not , no I just want something simple that works the same in every country.

Yes, I understood that you'd like to have a simple system. If you have such a simple system in mind, please describe it with sufficient detail so that any potential flaws in the system could be taken into account. Once all those potential flaws have been taken into account, I'm afraid the "simple system" will have grown into a somewhat complex system.

If a user gets 100 hits from a bundle in a bank , I shall not decide if this is interesting or not , it has to be the hit-maker's decision to consider them uninteresting

That's a novel approach. By definition, all hits have at least two users who have entered the note, let's call them user A and user B. What happens if user A thinks the hits are uninteresting, but user B insists that they're all interesting? In that situation someone else will need to make that decision. So I think we're back to square one with this. Try again.

this is not my business I am not competing against him or any other , I just want to know where my notes come from and go to, that is why I came here on EBT.

Oh. In that case I'm happy to inform you that even if your hits are moderated, the path your notes have taken can still be seen on EBT. Moderating hits only removes the hits from the interesting hits listing.

[Bollivierke]

this is not my business I am not competing against him or any other , I just want to know where my notes come from and go to, that is why I came here on EBT.

Oh. In that case I'm happy to inform you that even if your hits are moderated, the path your notes have taken can still be seen on EBT. Moderating hits only removes the hits from the interesting hits listing.

Why aren't these 2 (moderated) hits of mine not on my hit map then?
https://nl.eurobilltracker.com/notes/?id=114680639" onclick="window.open(this.href);return false;
https://nl.eurobilltracker.com/notes/?id=114680674" onclick="window.open(this.href);return false;

[avij]

Why aren't these 2 (moderated) hits of mine not on my hit map then?
https://nl.eurobilltracker.com/notes/?id=114680639" onclick="window.open(this.href);return false;
https://nl.eurobilltracker.com/notes/?id=114680674" onclick="window.open(this.href);return false;

Hm, that's a good point. The user-specific hitmaps are a subset of the global hitmaps. As the moderation is also used for cleaning the obviously fake hits out of sight, I think it makes sense to not include such moderated hits on the global hitmap.

As for the user-specific hitmaps themselves, technically it wouldn't be a problem to modify the user-specific hitmaps to include moderated hits as well. However, I don't know if the majority of the users wants to include moderated hits on their hitmaps. If that change was made, 99,5% of users wouldn't care (as they either have no moderated hits at all, or they have also a non-moderated hit taking the same path as the moderated hit), 0,25% of users might be annoyed that their "accidental meeting hit" is now visible on the map and 0,25% would be happier to see their moderated hit on their hitmap as well. But this is not the correct topic for such discussion. You can file an enhancement request on ebt.uservoice.com and if it gathers enough support, it can be implemented.

[Jes]

Here's something which I used as some sort of a "test exercise" for the moderators some time ago.
link to a pic
How would you calculate the scores in this particular case?

edit: removed attachment, added link, should work better.

Thanks. Now I could see it.

I think that it might be usefull some kind of "depreciation" of the hit, whenever it comes among the same users.

eg:

1 x DonChori - jonman interestingness = high (initially)

But

10 x DonChori - jonman interestingness = low

BTW: it is curious that the hits are allways DonChori - jonman and there are not "inversed" hits (ie: jonman - DonChori).

If it was the issue that you wanted to point out, I think it can be solved by implementing some sort of negative exponential factor in the formula. Some kind of:


EXP[-(number of hits with the same user - 1)/(factor greater than 1)]

The factor greater than 1 may be 10, for instance.

In that case, if an user has only 1 hit with his hitpartner, then the previous factor is 1. However, when these 2 users have 2 hits in common, the interestingness of the hit is decreased by a factor of:

e^(-(2-1)/10) = 0.9048
That is: the 2 hits become 10% less interesting. (-10% of interestingness)

Whenever these hitpartners have 3 hits in common, the interestingness of the 3 hits is decreased by a factor of:

e^(-(3-1)/10) = 0.819
So: these 3 hits become 18% less interesting.

And so on...
e^(-(10-1)/10) = 0.41 10 hits with the same user mean -60% of interestingness
e^(-(15-1)/10) = 0.24 15 hits with the same user mean -76% of interestingness

It would be easy to implement into the formula just by adding the factor "φ = [EXP[-(number of hits with the same user - 1)/(factor greater than 1)]]" at the end of the "chorizo".

Now, after all this long paragraph... I wonder if it was your point anyway, avij

[jonman]

just because you asked, jes, on that day i had exactly 200 hits with donchori. i accidentally ran into a bank where he had deposited his bundles from erfurt just before that. of these 200 hits of course only the 15 you can see on the pic escaped automatical moderation. after that donchori and i agreed on an operation mode avoiding any of us getting a bundle the other one had had before that. (and btw, there are also 25 moderated hits the other way round, from me to donchori.)
from what i can see there are two different meanings of the interestingness of a hit that are being thrown together. the first is that obviously some hits “seem” more interesting than others, based on the many factors people were already trying to put into a formula. until now ebt doesn’t know any differentiation between the interestingness of different hits, but everybody agrees that local helmbrechts hits are less interesting than an ireland italy hit and so on. also i can see very much how people would find several hits between the same users at the same time (due to bundle trackers who accidentally visited the same bank shortly after each other) are less interesting, and actually i would like it if this was (somehow) a factor of the final formula, should it be implemented.
i didn’t know that there had already been deliberations on measuring the interestingness of a hit back in 2009, and i find this idea actually quite nice.
however, there are two points i would like to have considered. the first is that this “interestingness” shouldn’t be confused with the other “interestingness” which is the opposite of “moderatedness”. there are several good reasons why a hit should not at all be included in thoughts of whether they are interesting or not. (i.e. obvious fake entries, friendly hits, hits with a time difference of 0 days, multiple hits with the same number of days and kms) from what i have understood in my 2 and a little bit years in ebt these are the only cases in which rigorous moderations of hits are being effected. these are very clear regulations and every hit can be assessed according to these categories (of course sometimes it takes some investigation to identify friendly hits, but theoretically speaking), so every hit can be put either in the moderated or interesting category from an objective point of view. as the first is the total opposite of the latter, it is impossible that there are degrees of “interestingness” in this meaning.
i still like the idea of giving degrees of “interestingness” (in its first meaning) on the hit report, but my second point is about the possibility for users to still be able to access statistics without this value. i could imagine (i am just writing what pops up in my head now) that there are two ways of having statistics shown to the user. first: hit lists according to the interestingness (first understanding) of the hits, when maybe more exotic hits are featured higher up in the list or whatever. second: hit lists according to the interestingness (second understanding, only the values 0 and 1 are allowed) of the hits, where you can still see plain old tables as they look like today

apart from thinking about the future, if i apply this to the cases mentioned before, i would say that both my 15 hits with donchori and also lito’s and beaweb’s hits should not be moderated, even if they’re not the best example of how exciting hits on ebt can be