Hacker News Reader

Comment by totalZero | Hacker News Reader

Dec 2020, second half:

  SETTLEMENT DATE|CUSIP|SYMBOL|QUANTITY(FAILS)|DESCRIPTION|PRICE

  20201215|36467W109|GME|170655|GAMESTOP CORP (HLDG CO) CL A|12.72

Not very rich data. I wonder how much of that comes from market-makers versus hedge funds.

View on HN

Wait, by my own calculation it seems GME had way more shares failed to deliver than that. I count over *14 million* in December using the [two data files provided by the SEC](https://www.sec.gov/data/foiadocsfailsdatahtm):

    $ cat cnsfails202012[ab].txt | csvgrep -d '|' -c SYMBOL -r '^GME$' | csvcut -c 'QUANTITY (FAILS)' | sed 1d | paste -sd+ | bc -l
    14276093

How many of those are the same shares failing to be delivered multiple times? Lots I don't understand...

The numbers you're looking at are aggregate data, so that includes past days I think, not just the given day. I think you have to subtract each number from the previous one.

> The figure is not a daily amount of fails, but a combined figure that includes both new fails on the reporting day as well as existing fails

Ah yes, they are aggregate indeed. The sentence just before the one you quoted also says:

> Fails to deliver on a given day are a cumulative number of all fails outstanding until that day, plus new fails that occur that day, less fails that settle that day.

So the last day we have data for is December 31, 2020:

  SETTLEMENT DATE,CUSIP,SYMBOL,QUANTITY (FAILS),DESCRIPTION,PRICE
  20201231,36467W109,GME,228358,GAMESTOP CORP (HLDG CO) CL A,19.26

As I understand it that means there were only 228358 shares failed to deliver at that point in time. This needs more analysis to understand how irregular it is compared to the other 12,000 symbols that had shares fail to deliver in December...

Reply on news.ycombinator.com