Yeah, so I used the FINRA "Monthly Short Sale Transaction Files", which can be accessed at:
These are monthly releases on a 1 month delay - the September 2019 data is available, the October one not yet. The September short sales transaction log is a large 450 MB file:
The 2018 September one I used for the figure above is:
These files are huge text files when uncompressed: the 2018/09 one is 2.5 GB. All lines that contain 'TSLA' are the Tesla transaction log, which have a fairly simple, easy to parse format:
Code:
Q|TSLA|20180918|08:00:02|S|100|296.300000||
Q|TSLA|20180918|08:00:02|S|47|296.500000||
Q|TSLA|20180918|08:00:02|S|10|296.176000||
Q|TSLA|20180918|08:00:08|S|15|296.250000||
Q|TSLA|20180918|08:00:08|S|200|296.999900||
Q|TSLA|20180918|08:00:08|S|100|296.999900||
Q|TSLA|20180918|08:00:08|S|50|296.550000||
Q|TSLA|20180918|08:00:08|S|100|296.980000||
Q|TSLA|20180918|08:00:08|S|100|296.609900||
I filtered these out and added up the volume fields (6th field) - which gave the 3,359,275 shares sold short and 20.3% of daily volume figure. (Note that this includes pre-market and after-market trades as well.)
Also note that these are subject to the usual caveats, but should still be considered an
overestimate of short selling activities: it's possible for some of those shares to not be related to short selling due to transaction aggregation - but I think it's not possible to do short selling that impacts the NASDAQ market which wouldn't be present in these transaction logs.
Note that it's possible to have more shorting activity in secondary markets, if a buyer and seller in a secondary pool is matched and crosses out each other without reaching the NASDAQ pool - but my impression is that this is statistically rare, most short sales do reach NASDAQ, as that's the intent of short sellers when they want to visibly mark down the price, plus higher buying liquidity would want to be present in the NASDAQ book as well and not give the shorts a discount via dark pools. Finally, there's also a requirement for dark pools to offer at least as good of a deal as the NASDAQ price - which too pushes most of the liquidity into this transaction log. In short: I think well over 90% of all shorting activities (maybe more than 99%), are present in this log.
To give another example of an 'ordinary' trading day with low volatility: September 17 last month, when the daily range was between $240-$246 and there was only a daily trading volume of 3.87 million TSLA shares, the FINRA transaction log shows short sales of 522,360 shares - or 13.5% of the daily trading volume.
I believe these percentages to be both more accurate and more believable than the daily short selling percentages reported by sites like volumebot.com. There is no way on Earth that shorts are trading around 50% of the daily volume - they don't have the position sizes for that, even on a hugely shorted equity like TSLA.
Just another example: on September 17 I also counted the number of short sales for Apple (AAPL), and it was 3,163,079 shares for the day, of 18.32 million shares traded in total - or 17.2% of the daily volume.
Also note that only a fraction of the transactions marked as short sales are held over the end of the day: I believe a lot of it is intraday trading.
Do you have a different interpretation of the FINRA data?