Yesterday's picks: how the model did (July 25)

Saturday graded as a split at the top of Verdexed's MLB card. The model published two reads with a stated direction, and it went one for two: its firmest side, Los Angeles at the Mets, won, and its early-window lean on Detroit lost. The three other games it published were flagged in advance as unreadable or edgeless, and none of them grade.

One for two is neither a good day nor a bad one. It is what a thin-edge card looks like.

The scorecard

Verdexed published five model reads on Saturday's full slate. Two carried a clear direction.

The Dodgers at the Mets was written as the clearest confidence side on the board, built on Los Angeles carrying the best record in the majors and Yoshinobu Yamamoto owning the day's strongest listed starter number at a 2.78 ERA. That side won. The lean was called firm rather than extreme because Nolan McLean at 3.34 was a real counter and the Mets were at home, and the final margin justified that hedge.

Royals at Tigers was the other direction. Detroit was the model's side in the 1:10 PM window with Casey Mize opposite Michael Wacha, but the column stated plainly that the market had already priced the Tigers near -180 and that the number was not generous. It lost.

The remaining three published reads do not grade. Cubs at Pirates was written as a near-even game, because team strength and starter quality cancelled each other out. Padres at Marlins was described as priced about right, with the model tracking the market rather than fading it. Yankees at Phillies was published as explicitly provisional, because New York's starting assignment had not settled and public listings conflicted between Ryan Weathers and Cam Schlittler. Declining to state a side on three of five is the model's read of a thin board, not an evasion, though it does make the graded sample small.

A precise full-slate graded record for Saturday, including the day's Brier score, was not retrievable from Verdexed's public accuracy surface at the time of writing, so no figure is claimed here. What can be verified is the published output, and on published output Saturday was a split.

Best call / worst miss

The best call was the Dodgers. Los Angeles won 4-3. Yamamoto went six innings without allowing an earned run, struck out five and walked three, and moved to 11-6. McLean fell to 7-7, and Brock Stewart closed it out. The firm side held, but by one run against a live opponent, which is roughly what a firm-rather-than-extreme lean should look like when it works.

The worst miss was Detroit. Kansas City won 3-2. Mize did his job, allowing one run on three hits over six innings, and the bullpen did not hold it. Salvador Perez tied the game with a solo home run off Kyle Finnegan in the seventh, a shot that also broke the Royals' all-time home run record, and Nick Loftin homered off Jacob Waguespack in the ninth for the winner. The model's stated concern here was the price rather than the side, and the price turned out to be the smaller problem. Flagging a bad number in advance does not convert a losing pick into anything else.

Two ungraded games broke notably. In Pittsburgh, the reason the model saw no edge was Paul Skenes, and Skenes was the input that failed: the Cubs beat him for five earned runs in 5 1/3 innings and won 11-0, with Pete Crow-Armstrong going 4 for 5 with a home run and three RBI while Shota Imanaga held Pittsburgh to four hits over six. Passing on that game cost the model a win it never claimed. In Miami, the model declined to back the home favorite near -134, and San Diego won 7-2 behind three RBI from Ty France, extending Miami's losing streak to 11. Not picking was the better outcome there. The Yankees beat the Phillies 3-1 in the game the model would not read, with Cody Bellinger's RBI single putting New York ahead.

What it means

One split Saturday is one split Saturday. Friday graded as a losing day, with the model's firmest side, Milwaukee at -243, buried by Colorado. Saturday's firmest side held. Two days, two opposite results at the top of the ladder, which is the ordinary texture of a market-anchored model rather than a signal about either day.

The more durable item is not the day's record but the abstention rate. The model stated a direction on two of five published games and said in advance, on the other three, that it had no side worth stating. Cubs at Pirates shows the cost of that discipline, a rout it did not call. Padres at Marlins shows the benefit, a favorite it did not back. Across a large sample those roughly cancel, which is the argument for passing rather than manufacturing a lean.

Verdexed's month-to-date accuracy figure and Brier trend were not retrievable at the time of writing, so no claim is made here about the longer-run record. On Saturday alone: one firm side held, one badly priced side lost, and three games the model declined to read broke in three different directions. Logged, not overreacted to.

Responsible play

**This is a record of past model output, not betting advice.** Verdexed grades its model in public for transparency and entertainment. Past results do not predict future outcomes, nothing here is a recommendation to place a wager, and no pick is ever a guarantee, even a strong model loses regularly.

**You must be 21+** (or the legal age in your jurisdiction) and located where sports wagering is legal to bet. Sports betting is not available everywhere; know and follow the laws where you are. Never bet more than you can afford to lose, and never chase losses.

If gambling is affecting you or someone you know, help is available. In the US, call or text **1-800-GAMBLER** (1-800-426-2537), or visit ncpgambling.org. Please play responsibly.