At 15:14:05 UTC on 10 September 2026, three "Bitcoin above $X at 12PM ET" prediction markets went from a quoted mid of about 0.024 to about 0.500 in a single second. All three strikes moved in the same one-second snapshot: 78,400, 78,600 and 79,000.

The interesting part is not the size of the move. Mid prices do move half a dollar on Polymarket. The interesting part is that the new 0.500 mid was not tradable — the order book behind it asked 0.998 — and that the same change happened across independent strikes at the same instant. Prices alone cannot show you this. The book can.

The flip, second by second

The three contracts below resolve against the noon ET price snapshot at 16:00 UTC. Here are their mid prices around the flip, read from 1-second price history:

strike    15:14:04   15:14:05   15:14:06   15:14:08
78,400    0.0250     0.4995     0.5000     0.5000
78,600    0.0240     0.4995     0.4995     0.5000
79,000    0.0230     0.4990     0.4990     0.4990

Each strike's mid leapt 0.475 to 0.476 in one print. The 60-second scan we ran first had already flagged all of these strikes landing in the same minute bucket, including 78,800, which we did not verify second-by-second. At 1-second resolution, two of the strikes tick again a second or two later (0.4995 to 0.5000) as one more bid arrives. The 15:14:05 print itself is simultaneous across all three.

What the book was doing

This is the 78,600 contract's order book in the same second, top level only:

time        best bid   best ask   spread    mid
15:14:04    0.001      0.047      0.046     0.0240
15:14:05    0.001      0.998      0.997     0.4995
15:14:08    0.001      0.999      0.998     0.5000

At 15:14:04 the contract traded against a real, tight-ish quote: bid 0.001, ask 0.047, a 4.6-cent spread. That ask carried 280 shares. At 15:14:05 that ask was gone, and the shallow book that remained — 0.001 bid, 0.998 ask — produced a mid of 0.4995. The one executable price level was withdrawn, not filled. The 79,000 contract shows the same pattern: its 0.045 ask at 280 shares vanished in the same second, and its quoted spread went from 0.044 to 0.996 in the 1-minute metrics — the same 0.996 spread registered on both contracts. That one empty 99.6-cent chasm is the whole "0.50 repricing."

Two details support that reading. The book's liquidity field fell exactly 280 at the same minute bucket, matching the size of the withdrawn ask. And the 280-share level appeared again minutes later on the 77,600 contract, at different prices — a recurring size that looks like one quoting book sweeping the ladder, which is a theory, not something we can prove from snapshots alone.

Thirty minutes later: a flicker, not a repricing

The 77,600 contract did something messier at 15:44. Its mid alternated a full 0.47 apart on consecutive seconds:

time        mid
15:44:04    0.5000
15:44:05    0.0305
15:44:07    0.5000
15:44:08    0.0305
15:44:19    0.0295
15:45:14    0.0205

The alternation is a two-level book assembling and disassembling itself. A bid at 0.002 and an ask at 0.059 (280 shares each) appeared and vanished across seconds; the mid is 0.0305 while both are present, and 0.5000 — resting on a 0.998 ask — when the 0.059 ask has dropped out. The single-second mid series maps one-to-one onto the book state. In a 1-minute price series, 15:44:04 through 15:44:08 are one "volatile minute" with no mechanism attached.

What one second is worth here

Polymarket's own free API serves 1-minute price history. For both events above, a 1-minute reader sees a jumpy minute at 15:14 and another at 15:44. Two facts never appear:

  • That the strikes moved in the same second. Three independent contracts landing in one minute bucket is a different statement from landing in one second, and the simultaneity is the evidence that one quoting operation (not three independent traders) moved the ladder.
  • That the new price was not executable. A 0.500 mid over a 0.996 spread describes a market that has stopped quoting, not a market that has traded there. The mid tells you where the book used to be; the book tells you where it is.

That is the argument for second-level depth in plain form. We are not trying to convince you that Polymarket's prices are wrong — they are not, and for tracking probabilities the free API is the right tool. The claim is narrower: the mechanism behind the 0.02-to-0.50 print is only visible in 1-second order book snapshots.

What we are not claiming

We do not know why the quotes were withdrawn, and we did not attach a cause. We did not compare against a spot price or a news feed; nothing in this post depends on what Bitcoin actually did at 15:14:05.

The sample is also small by design: 402 resolved bitcoin markets scanned over the three days before 10 September, drilled down to a handful at 1-second resolution. Nine of the fifteen largest 60-second moves in the scan were up/down contracts (seven five-minute, two fifteen-minute) printing after expiry — the settlement pattern our earlier study of crypto-market repricing documented — and six were noon-ET strike contracts. We verified three strikes second-by-second. A wider survey of strike ladders would tell you whether this happens often; we have not measured that yet.

Checking it

Every figure above comes from contracts readable through the order book history endpoint at 1-second resolution. The slugs are bitcoin-above-78600-on-september-10-2026-12pm-et (and the same pattern for 78,400 and 79,000). The book snapshots and the price series come from the prices, books and metrics endpoints for the 15:13–15:16 and 15:42–15:46 UTC windows.

The point of this post is the difference between the two readings of the same event: the price history says "repriced," the book says "stopped quoting." If you are sizing orders off mid prices alone, that distinction is the whole trade.