In the last piece we said the only short-interest percentage worth reading is the one measured against the Float, not against all shares outstanding.
Fair enough. So where do you look the float up?
Mostly, you cannot. And the reason is not that we have not got round to it. It is what float actually is.
Here is our own database, all of it, with nothing rounded off in our favour.
Read that bottom line again, because it is the uncomfortable one. On the liquid names people actually trade, we can establish a genuine float 4.5% of the time.
Now ask yourself where the screener you use gets its number.
Float is derived, not published
No company files a document that says "our float is 41,200,000 shares." There is no such filing. Float is something you calculate, and the calculation is:
Shares outstanding, minus the stock that is not really available. Insider holdings, restricted stock, lockups, strategic stakes that are not going anywhere.
The first half is easy. Companies report shares outstanding on the cover of every 10-Q and 10-K, and we hold that figure for 6,941 tickers.
The second half is where it falls apart. Insider ownership is spread across Forms 3, 4 and 5, proxy tables and footnotes, filed by different people at different times in formats that were designed for lawyers rather than parsers. Sometimes it is unambiguous. Most of the time it is not.
When we cannot establish it confidently, we do not guess. The row gets shares outstanding and a note saying so, explicitly described as an upper bound on the float rather than the float itself. That is 6,293 of our 9,005 tickers.
That middle bar is worth a sentence of its own. 2,066 of those tickers are not common stock. They are warrants, units, rights, fund lines and preferred or depositary series. The issuer's share count describes the common stock and simply does not apply to them. Printing a float for a warrant is not a missing number, it is the wrong number, and a screener that shows you one is telling you it does not know what the ticker is.
The part that will annoy people: low float is rare
Of the 646 floats we can actually establish, here is how much of the company is genuinely available to trade.
474 of 646, roughly 73%, have a float of 80% or more of shares outstanding. For those, float and shares outstanding are nearly the same number and the distinction you were warned about barely moves the answer.
Genuinely low float, under 25% of the company, is 23 tickers out of 646. About 3.6%.
That is the thing worth taking away. "Low float" gets attached to almost any small stock that moves, and most of the time it is simply not true. The float is 85% of the company and the move was something else.
The number that is missing from the float entirely
This one catches experienced people, so it is worth being slow about.
Float counts stock that exists today. It does not count stock a company has already promised to somebody: convertible notes, warrants, options. Those shares are contracted and not yet issued, so they are correctly excluded from float, and they are also completely real. They arrive later, and when they do, they arrive into your float.
453 rows in our table carry a note flagging exactly that, with the contracted share count and the date it was reported.
So a name can be genuinely low float today and still have its supply doubled on a schedule that is already public. If you are trading a tight float on the theory that there is no stock available, the Overhang is the thing that will settle the argument against you, and it is sitting in the filings right now.
A float number with no overhang figure beside it is half an answer.
What good looks like, and how to test any source
We publish "we do not know" on 93% of our tickers. That reads badly for about ten seconds, until you ask the alternative: what is a source doing when it shows a confident float for all 9,005?
It is doing one of two things. Either it has a vendor estimate it did not tell you was an estimate, or it quietly used shares outstanding and called it float. That second one is not hypothetical and we are not above it. On 2026-08-15 two surfaces of this site printed short interest for $SIDU as 25% in one place and 143% in another. Same underlying row. One divided by shares outstanding, the other by a float our data could not actually establish. Neither said which. We fixed that by naming the denominator everywhere and refusing to print the number when the float is not credible, which is why you now see blanks where we used to have confident-looking figures.
Four questions for any screener, ours included:
- Is this float or shares outstanding? If it will not say, it is shares outstanding.
- Where did the insider figure come from, and when was it filed? Float built on a two-year-old proxy is a two-year-old float.
- Does it show anything for warrants, units and fund lines? If it does, it is not checking the security type.
- Is there an overhang figure anywhere? If not, you have the float and not the supply.
Our low float stocks board only ranks names where we hold a real float, which is why it is shorter than you might expect. The low float movers board applies the same rule. The most shorted stocks board names its denominator on every row, and the short squeeze board filters it further.
If you want the companion piece, the four numbers that decide a short squeeze is where the float number actually gets used, and where getting the denominator wrong costs you the most.
A shorter list you can trust beats a complete one you cannot check. That is the whole argument.