Live data from Hacker News

The MtGox 500

bitcoin.stamen.com

21–30 of 43 posts

Re: The MtGox 500

#21

Users 1 and 15's charts make no sense - they have to be special system accounts of some sort. My guess is that #15 is the account that receives trading fees, and #1 represents MtGox itself (or some specific aspect of MtGox, such as its cold-storage). There're still a lot of points on these plots that don't make sense, though; generally they look like vertical stripes labelled as sets of small sell orders, both far be…

They look like bid/ask offers.

Re: The MtGox 500

#23
post #22

This is incredibly well done. Did you just download all the data and analyze it with some crazy R skills?

Thanks! The data was released in a big zip file with a mix of malicious content (malware that stole your btc, personal data on MtGox CEO Mark Karpeles). Trade data was in monthly csvs, and also included some duplicate data which we removed.

Analysis happened in sqlite, python/pandas, csvkit; charts were built in d3.js and exported with phantom.js

Re: The MtGox 500

#24
post #7

Earlier quoted context omitted.

Yes, you used to be able to trade via dark pool in large amounts (at the time, that would be 5+ kbtc) to avoid upsetting the market price. Not sure if they still supported it at time of closure. This is actually rather scary. Enough people have logged data that these can be dereferenced despite the feeble attempt at depersonalization, but I guess the names will come out in the legal proceedings anyway, depending on j…

I don't think the (publicly) leaked database contained any usernames/emails.

You're right, it didn't. All of the trades data included internal User IDs. More recent data (as of April 2012) included User ID hashes, which Gox users have access to (from their password recovery emails for example).

Re: The MtGox 500

#25
post #7
post #2

I’m not sure I understand the difference between the green Selling BitCoins (to get Dollars, presumably) and Buying Dollars (by selling BitCoins, also presumably). I’m even more puzzled by the rates out of trade: did MtGox allowed traders to agree one-on-one on their own rates? That would allow a lot more laundering than any theft. Missing too are relative size of the traders.

Yes, you used to be able to trade via dark pool in large amounts (at the time, that would be 5+ kbtc) to avoid upsetting the market price. Not sure if they still supported it at time of closure. This is actually rather scary. Enough people have logged data that these can be dereferenced despite the feeble attempt at depersonalization, but I guess the names will come out in the legal proceedings anyway, depending on j…

> depersonalization

I think you're maybe looking for bowlderization or anonymization or something similar?

Re: The MtGox 500

#27
post #20
post #12

The biggest take away from this is that bots were driving the price up, and preventing its fall. "Bots" offer a lot of stability to the market because they don't react to bad news. Though an error can cause "flash crashes" As happened in May of 2010 on the stock market. http://en.wikipedia.org/wiki/2010_Flash_Crash

How exactly did these charts lead you to that conclusion?

Look at several places the bots started binge buying when the price started to fall out.

Re: The MtGox 500

#28

Users 1 and 15's charts make no sense - they have to be special system accounts of some sort. My guess is that #15 is the account that receives trading fees, and #1 represents MtGox itself (or some specific aspect of MtGox, such as its cold-storage). There're still a lot of points on these plots that don't make sense, though; generally they look like vertical stripes labelled as sets of small sell orders, both far be…

Each dot is a trade. I pulled the columns from the db we generated with the raw info of that one big trade, so you can see what the structure of the source data looks like. Removed the user id of the buyer, but it's trivial to find if you have the dataset.

File|Trade_Id|Date|User_Id|Type|Currency|Bitcoins|Money|Money_JPY|Money_Fee|Bitcoin_Fee

2013-11_mtgox_japan.csv|1384974682548514|2013-11-20 19:11:22|*|buy|USD|1000|610000|60941267.96|0|3

2013-11_mtgox_japan.csv|1384974682548514|2013-11-20 19:11:22|THK|sell|USD|1000|610000|60941267.96|0|0

Re: The MtGox 500

#29

Fun stuff. Bryan if this is your site can you add to the plot in the corner a dot with net value change? Assume that all bitcoin "held" are currently worthless, so a holding of 10 btc at the end would be -10*last sale price recorded on the exchange. That would give an interesting idea of which traders "won" the game and which "lost."

We started doing this analysis and have the data per user but didn't want to include it in this goaround. Still much to do and investigate with this data set! Thanks for the feedback.

Re: The MtGox 500

#30

Users 1 and 15's charts make no sense - they have to be special system accounts of some sort. My guess is that #15 is the account that receives trading fees, and #1 represents MtGox itself (or some specific aspect of MtGox, such as its cold-storage). There're still a lot of points on these plots that don't make sense, though; generally they look like vertical stripes labelled as sets of small sell orders, both far be…

From "About":

>The special user TIBANNE_LIMITED_HK is not represented in the MtGox 500 since its trades are not in JPY. You can read a conjecture on the role of this user[1]. The other special user, THK, shows up as #1 in the MtGox 500.

[1] http://7u83.cauwersin.com/?p=12

Post reply on HN