A stop-loss order does not sell at your stop price. It sells
when the market reaches your stop
price, and what it sells into is whatever bid happens to be standing there at that moment. The stop
price is the trigger. The bid is the outcome.
Most explanations of this stop at "so you might get a worse price". That is true and useless. The
question a trader can act on is: how much slippage, and does it depend on anything you can see in
advance. We recorded the answer.
1. What a stop order actually does
A stop-market order sits with your broker as an instruction. When the stock trades at or through
your level, the instruction converts into a market sell. A market sell takes the best bid.
So the cost of a stop has two parts. The first is how far the tape jumped between the last price
above your stop and the first price at or below it. The second is the gap from that print down to
the bid you actually sell into. The second part is a spread you pay, and unlike the first part it
is visible in the quote at the very instant the stop fires.
2. What we recorded
Our engine stores a snapshot every second for each stock on the day's scan list: the last price,
the full quote, and sizes. For this measurement we took 38 sessions sampled evenly out of 264
recorded sessions, which gave 745 symbol-days and 4,400,640 quoted seconds inside the regular
session.
Of those seconds, 387,948 had a last price below the previous second. That is 8.8% of the
recorded time. Those are the seconds in which a stop sitting somewhere just below the market would
have fired, and they are the population everything below is measured on.
3. The popular claim, and what our data says
The claim you see repeated is that stops get filled badly because they trigger exactly when
spreads blow out. On average, in our recording, that is not what happens.
| Population | Mean quoted spread |
| All recorded seconds | 0.91% |
| Seconds where the price fell | 0.82% |
The spread on falling seconds was slightly
narrower than the all-day average, not wider. The
reason is mundane: a second in which the price moved is a second in which somebody traded, and
seconds with trading in them carry tighter quotes than the dead stretches in between. The all-day
average is dragged up by quiet periods where a wide quote sits untouched. The median spread across
all recorded seconds was 0.53%.
So the blanket claim does not survive measurement. What replaces it is more specific, and worse.
4. The spread scales with the size of the drop
Split the same down-seconds by how far the price actually fell in that second.
| Fall in one second | Down-seconds | Mean quoted spread |
| Up to 0.25% | 199,344 | 0.50% |
| 0.25% to 0.50% | 95,144 | 0.85% |
| 0.50% to 1.00% | 63,643 | 1.17% |
| 1.00% or more | 29,817 | 2.11% |
The relationship is monotonic across all four buckets, and the ends differ by 4.2 times. The thing
that triggers your stop is the same thing that widens the spread you have to cross to obey it.
This is the part that matters for position sizing. A stop is not insurance that pays a fixed
premium. It is insurance whose premium is set at the moment you claim, by the size of the event you
are claiming for. Small drops are cheap to exit. The drops you actually put a stop in for are not.
5. How far the bid sat below the last print
The spread is the round trip. What a seller pays is the half of it below the print, so we measured
that directly: at each down-second, the distance from the last price down to the best bid.
| Percentile of down-seconds | Bid below the last print |
| Median | 0.12% |
| 90th | 0.79% |
| 99th | 2.90% |
In 28.2% of down-seconds the bid was at or above the last print, so a market sell would have
received the print price or better. The typical case is mild. The tail is not, and the tail is
concentrated in exactly the seconds measured in section 4.
6. The same measurement, by price
Cheap stocks quote wider, and a stop on them costs more to obey.
| Stock price | Down-seconds | Mean spread | Bid below print, median | 90th |
| Under $2 | 71,448 | 1.22% | 0.34% | 1.14% |
| $2 to $5 | 88,940 | 0.67% | 0.19% | 0.61% |
| $5 to $20 | 155,810 | 0.72% | 0.13% | 0.76% |
| $20 to $100 | 67,765 | 0.72% | 0.12% | 0.66% |
| $100 and up | 3,985 | 2.82% | 0.10% | 1.98% |
Read the last row carefully rather than as a trend. It holds 1.0% of the down-seconds, and its mean
is pulled by a small number of thin, high-priced names in our scan lists. Its median is the
tightest in the table at 0.10%. When the mean and the median disagree that violently, the mean is
describing a handful of rows, not the bucket.
The sub-$2 row is the one to take seriously. It has 71,448 down-seconds behind it, and both its
mean and its median are the worst in the table.
7. What a one-second grid cannot see
Our snapshots are one second apart. Anything that happened between two snapshots is invisible to
us, and that limits one of the two cost components.
The jump from the previous second's price to this second's price had a median of 0.23%, a 90th
percentile of 0.88% and a 99th of 2.65%. Those are
upper bounds on what a stop would have
suffered on the trigger, not estimates of it. If the price walked down through intermediate prints
inside that second, a stop sitting in between would have fired on one of those prints, above where
we see the second close.
This is why the article is built on section 4 and section 5 rather than on that number. The spread
and the bid distance are read from the same snapshot as the fall itself, so they are true no matter
how the second is sliced. We are stating the limitation rather than quietly reporting the bigger
number, because the bigger number is the one that would have made a better headline.
8. What to do with this
Three things follow, none of which require predicting anything.
First, a stop distance is not a loss size. If you place a stop 2% below your entry on a sub-$2
stock, the exit is not at 2%. Add the bid distance for the size of the move that would reach your
stop, and on that price bucket the 90th percentile alone is 1.14%.
Second, a stop-limit is a different trade, not a safer one. Limiting the price is exactly a refusal
to cross a wide spread, and section 4 says the spread is widest in the moves you most wanted out
of. The order will sit unfilled while the stock keeps going.
Third, this cost is measurable in advance for the stock you are about to trade. The quote is on the
screen before you enter. A name that quotes 1.22% while calm is telling you what a forced exit will
cost, and no amount of stop-placement care changes it.
9. Sample accounting
The sample is 38 sessions drawn evenly from 264 recorded sessions between 2025-04-02 and
2026-07-27, giving 745 symbol-days and 4,400,640 seconds, of which 387,948 were down-seconds.
The universe is the stocks our own scan surfaced on each day, which are the day's active movers
rather than a cross-section of the market. That is the population a day trader places stops in, and
it is also why the spreads here are wider than an index name would quote. Only the regular session
is included, since a stop placed for the open behaves differently before 09:30 ET.
Seconds were used only where the quote was complete and the two snapshots were exactly one second
apart, so gaps in the recording cannot masquerade as price jumps. The bid distance is floored at
zero, meaning the 28.2% of down-seconds where the bid stood at or above the last print enter the
percentiles as no gap rather than as a negative cost.
We did not measure returns, entries, or whether stops help. This is a cost measurement.
