Polymarket Mid Price, Explained

• Mid = (best bid + best ask) / 2, on the 0.01 tick grid that quotes follow.
• It is a quote of quotes — the average of two resting prices, not a resting price itself.
• A symmetric book has a mid that is tradeable; an asymmetric or empty book has a mid that is fiction.
• Bid/ask ratios at top-of-book determine which side the mid misleads.

Two recorded examples define the trap: a BTC strike ladder printed a mid of 0.4995 with an empty 0.059–0.998 segment — half of the market mid over a 99.6-cent hole; and a 0.500 print rested on a 0.996-wide chasm after a one-second repricing.

In the settlement window the mid behaves differently: it is stable at 0.99 while depth at that price grows past 67,000 shares. The mid looks calm because the book is decided; the chart viewer sees no signal, the book reader sees the answer.

• Backtest fills against ask/bid, never the mid, unless you model spread-paid execution.
• Flag rows where spread exceeds N ticks; those rows are where mid-models fail.
• Use recorded crossed flags to mark books best-bid-more-than-best-ask.
• Report both mid and executable (ask) implied probability in signal papers.

• Identify the mid on a two-sided frame (the average of the executable bid/ask) and on a one-sided frame (undefined, or the last two-sided mid); the page teaches when the mid exists at all.
• Compute the mid-to-floor gap on a one-sided frame: the mid can sit several ticks from the executable side, and that gap is the honest immediacy cost.
• On a reprice window, compare the mid path to the floor path (the 0.815 → 0.065 mid sweep with a floor that moved at the same second); the mid lags the floor when the book is one-sided.
• Store the frame with its one-sided label whenever you cite a mid; a mid without a book label is half a price.

The mid is the average of the top bid and ask, and on Polymarket it is often a phantom: on one-sided books there is no executable mid, so the number displayed is a summary of a floor that is either far away or gone.

The measured reality — 90%+ one-sided frames and round-grid bunching — makes the mid a second-class citizen; the floor is the executable price and the mid is its smoothed echo.

The mid is the cheapest metric to misquote, so the archive stores the side for every frame; the discipline for any report is to label the mid with its book state or omit it, which is exactly what a one-sided venue demands.

The audited habit is to display the mid only next to its two anchors: the floor and the reference second. With those anchors the mid becomes a derived convenience instead of a primary truth, and on a one-sided venue that ordering is the difference between a waveform and a fact.

The mid exists only where both sides do; the book label decides.

Cite the floor when it matters; the mid is a display value, not an executable price.

On reprises the mid lags the floor by construction; the timestamped ladder is the record of both.

The mid is the display layer of the ladder; this page is the correction to every chart that treats it as executable.

Pair it with the spread and price-bounds explainers; both measure the distance between display and floor.

The one-sided-frame rule (no executable mid) is the exact discipline the execution pages assume.

The mid is a display value, real only where both sides rest; the floor is the executable price.

Label every mid with its book state and it stops being a phantom.

• Order book data
• Price history
• Crypto data
• Provider matrix