Topic: Merit Data from 315 Bitcoin Paying Signature Campaign Participants (Read 311 times)

I think you are a prized cunt and contribute Very little to this place

I think the large difference between the average and the median number of merits received is an indication that most participants receive either no, or very little merits for their posts.

HabBear

I agree, no conclusions can be drawn without more investigating...

Interesting that the person who sent 18 to the sig person is just a member and is receiving merits like this:

Sent in the last 120 days

    January 31, 2018, 04:47:00 AM: 8 to dvillier for Re: encourage others to bitcoin
    January 25, 2018, 10:20:09 AM: 5 to dvillier for Re: Good news!
    January 25, 2018, 10:19:44 AM: 5 to dvillier for Re: Good news!

Received in the last 120 days

The sender  merited this jewel twice for a total of 10:

"That is nice to read. I certainly hope the market will recover because of this good news. I have been on this for a while that I observed when there is a negative news about bitcoin it is gravely affected and vice versa. but if bitcoin price drops the alts drops too and if btc price goes up the alts goes down too. So if you have some crypto now other than bitcoin I suggest you get out for a while. "

So, no conclusions, but the fish smell increases

OP, i also apologize, I did not intend to take this turn

This could be a dangerous request. What hypothesis are you trying to prove with this analysis?

High quality posters are going to give merit to other high quality posters. So there should be a lot of people with high received merit giving merit to others who've received a lot of merit.

Are you trying to isolate specific member names for some reason?


You can't infer collusion between members simply by identifying who's giving merit to whom and the fact that they're both in signature campaigns. There's, literally, not data in the comparison that could lead you to prove a causal relationship.

Theoretically, what's more likely to prove collusion is how much of the merit sent from high quality posters in signature campaigns is sent to members not in signature campaigns and the percentage that the merit sent to those accounts of the total merit available to send by the member.

Summary:

• Collusion is more likely to happen from high profile accounts to low profile accounts (i.e., alternatives)
• You can only infer collusion if there's a high amount of merit sent from the high profile account to a low profile account (i.e., squandering)

Toughit raises an interesting point. Receiving 50 merit from 1 user and receiving 1 merit from 50 users are two entirely different cases, yet appear the same when the data is collected like this. I think receiving merit from a larger number of users is more indicative of high post quality than simply having a high total, perhaps you will agree.

Since you seem to like stats, maybe you could run a similar script that goes through each user's merit page and tabulates the instances where the word "from" appears, since each one of these represents a time they received merit. I suppose this would still count user's who have received merit from the same user multiple times, but it will at least make the metric a bit more robust.

I found the link to your raw data, thanks for loading it.

I picked the first one that had a large amount of merit received (20 is a lot to me).

Turns out the 20 received came from just 3 people: 1 merit, 1 merit and 18 merits.

So it seems the receiver of the 18 merits had sent 6 merits to that person 4 days earlier and then got 18 merits back.

I did not look at the posts to see if they were worth 6 or 18, but seems fishy already from
a sample size of 1.


Data:
Sent in the last 120 days

    February 15, 2018, 12:41:04 PM: 6 to piter66 for Re: Bitcoin can be hacked?
    February 15, 2018, 05:56:32 AM: 2 to mkettler for Re: what is the predicted value for BTC in end of 2018???
    February 15, 2018, 05:55:19 AM: 2 to zzxswqq for Re: What is the easiest way to help grow Bitcoin?
    January 29, 2018, 01:20:14 AM: 6 to dvillier for Re: Bitcoin in the future

Received in the last 120 days

    February 09, 2018, 12:40:43 PM: 1 from Napole0n for Re: what's your strategy for bitcoin now
    February 03, 2018, 10:42:09 PM: 1 from amla for Re: What happens to your Bitcoin when you Die?
    February 03, 2018, 10:35:42 PM: 18 from dvillier for (Deleted/Off-limits/Ignored)

You did a great job but I think this is too early to draw conclusions. This data involves a lot of suspicious use of merits like  this:
In the bitvest campaign, there is a member:
https://bitcointalksearch.org/user/shezu007-1038148

As per his merit summary, https://bitcointalk.org/index.php?action=merit;u=1038148 , he got 18 from https://bitcointalksearch.org/user/dvillier-1223911 and then sent 6 to him. The post he gave 6 merits was: https://bitcointalksearch.org/topic/--2099052

There will be a lot of such cases in this data.

Lets see what the usual meta contributors think about this one, thank you don DarkStar_leone for spending time on this, but merit system doesn't work, Bitcoin paying campaigns were always accepting people with average post quality.

Merit system would work if all the bounty beggars with low post quality stop posting low quality, they couldn't, because an illiterate street beggar doesn't know how to not be an illiterate street beggar. if ranking up from a member into a full member is impossible for a beggar, they could join a bounty with 2 member accounts. if becoming a member is impossible for the same beggar, they could join a bounty with 4 jr accounts.

Your 3 points aren't able to be easily automated, so I don't have the data. Upper right corner isn't an actual data point, I think it's just showing the top of the scale on the other side. It was put there automatically.

1) Of the 315 people can you do a table showing how many sent merit to any of the others on that list.
2) I think math tells me 5112 merits were received
3) of those 5112 merits, how many transactions were there (if every merit awarded was just 1 merit, there would be 5112 transactions)

do you have a table that shows how many smerits each level received to award?  As a full member I think I got 5

Thanks!

(and yes, my answers would show collusion or not among the 315)

edit: and what is that outlier in  the upper right of the chart?

I scraped the earned merit data of 315 people participating in bitcoin paying signature campaigns, just for fun and to have a comparison versus my earlier dataset of 503 altcoin paying signature campaign participants. The raw data is available in the format Username | Earned Merit | Profile URL. The data is separated by campaign name, so you could compare between campaigns if you wish.

---

Data set

Data was obtained from 15 campaigns. I aimed to scrape all of them, but was not able to due to formatting, or copying restrictions. Some campaigns blocked copying, so I respected their wishes and didn't take their data. Those campaigns were:

 - Crypto-Games.net (oddly, all other Lutpin run campaigns were scrape-able)
 - NitrogenSports
 - Coinpayments

Other campaigns did not have a public dataset:

 - YoBit.net
 - Coinroll

Finally, some campaigns had weird formatting, making it too time consuming for me to grab the profile data. Those campaigns were:

 - BitDice
 - OneHash
 - FortuneJack


There are potential errors in the data, especially for campaigns where "Profile links may be incorrect" is noted on the spreadsheet. Campaigns without that string had their profile links taken directly from what the manager had, while campaigns with the string did not have links^[1]. Those without links had their profile link scraped^[2], which could result in errors.

---

Statistics

 - Average of 16.2 merits each
 - Median of 4 merits
 - Excluding the 75 people with no merits, the average person earned 21.3 merits

---

Possible Interpretations

• People participating in Bitcoin paying campaigns generally have a better post quality than those without a campaign or in altcoin bounties, as managers are usually very picky about what they accept. You could infer that the merit system is working, as people with good posts are getting many merits.
  

---

How?

I'm scraping merit data using Google Sheets, with this formula:

It grabs their merits, subtracts how much merit they would have been given at the start of the system based on current rank, and returns the result. Does not work if they have ranked up, ranked down, are Hero Members who were given 1000 initally or with Copper Members. You need to subtract/add the correction manually, and it's pretty easy to spot errors. This formula grabs their current merit count:


Be warned: You will get rate limited eventually. My original plan was to grab 1000 users worth of data, but it was getting really slow.

The D2 in the formula is a link to the user's profile. Data will automatically update as long as the formula is present.

---

^[1] Legit question, why? Seems kinda annoying to do post counting with that
^[2] Scraped using a DuckDuckGo^[3] with the below formula. A3 represents the username.
^[3] Google wouldn't let me scrape Google search, so DuckDuckGo it is! Side note: Bing search is absolute trash.^[4]
[4] I literally searched for "View the profile of XXXX", in quotations so that it returns exact phrase matches, yet I still got completely unrelated results very often