STEEZANOMICS.
← all games

Philadelphia Phillies @ Arizona Diamondbacks

Wed, Sep 2 3:41 PM·Chase Field·186 lines

odds pulled 2026-09-02 · built 2026-09-02 23:00:48 UTC

Prices last pulled 1:00 PM ET.

StartersAndrew Painter vs Taylor Clarke▾
Philadelphia Phillies starterAndrew PainterRHP5.55 ERA · 7.78 K/9 · 1.37 WHIP · 105.1 IP
HR/PA allowed3.9%barrel allowed8.2%hard-hit allowed36.1%fly-ball rate26.9%HR/PA vs RHB3.4%HR/PA vs LHB4.2%PA faced433recent HR/PA2.7%
Arizona Diamondbacks starterTaylor ClarkeRHP2.53 ERA · 5.34 K/9 · 0.88 WHIP · 64.0 IP
HR/PA allowed2.5%barrel allowed9.1%hard-hit allowed33.2%fly-ball rate23.5%HR/PA vs RHB3.0%HR/PA vs LHB2.2%PA faced238recent HR/PA6.9%
1 start sample, season line is noise
to

one row per line. Both prices sit on the same row: O is the over, U the under.

every rate and every bar is the OVER's result. A 20% over rate means the under landed 80%, so read the under off the same number rather than looking for a second figure.

An edge is only ever shown when at least one no-vig exchange priced the market. Roughly a third of the board is quoted by retail books alone, and a fair built only from those is a retail consensus rather than an independent read - those rows keep their price and their fair and show no edge, which also keeps them out of the price-edge ranking. Where exactly one exchange is behind it the number is dimmed: real, but thin.price edge = the price vs a de-vigged fair built from the books quoting that side. It is market-derived, not a predictive model. It says the price looks off, not that the play wins. Almost every prop prices negative because retail books carry roughly 7% vig, so a small negative is normal.

Three sorts, three different questions, and they are not versions of each other.recent form and lean strength both read the game log and ignore the price: form ranks by 0.6 × L10 + 0.4 × season from the over's side, lean strength by whichever side sits furthest from a coin flip, so a 90% under and a 90% over score the same. price edge reads only the price and ignores the game log entirely. A row can top one and sit near the bottom of another, and that is the two of them disagreeing rather than either being broken. The history liking a side is not evidence the price is wrong, and a soft price is not evidence the player is in form.

bars = the games in the selected timeframe, oldest → newest: L5 draws 5, L10 draws 10, L20 and SZN draw up to 20. cleared the line, missed, no game data. The dashed rule is the posted number; labels are that game's stat value. Bar heights use one fixed scale so switching timeframe doesn't rescale the picture. Season history runs deeper than 20 games: the chart shows the last 20, the SZN percentage uses all of them.

my books re-prices every row from that book's own quote. Rows the book doesn't post are marked, never hidden and never left showing another book's price. With a book selected each row also statesvs best price: how far that book's over and under sit below the best price posted anywhere on the board, measured in payout rather than in American points so a move from −110 to −105 and one from +200 to +210 are not called the same size.

The % beside a price and the green rule down the left of a side belong to the price edge sort and appear only there. On a form or hit-rate sort the row has been ranked on the game log, and a price mark sitting in it answers a question you did not ask. It can point at the under while the form points at the over. Switch to price edge and both come back. The rule marks the side whose edge is positive, and only then. Most props price negative on both sides because retail books carry roughly 7% vig, so no mark is the normal state and is not a fault. On WNBA in particular many books quote one side only, which leaves a real price with no independently judged fair: those rows show the price and leave the edge blank rather than inventing one.

N sharp quotes = how many sharp/exchange books were in the fair price. It proves quotes exist, not that they agree. thin = under 15 games of history; rates show a dash when the sample can't support them.