Yesterday's picks: how the model did (July 28)

Tuesday was a clean day for Verdexed's MLB model, and the cleanest part of it was the part the model was least sure about. The column published five reads on the July 28 slate. Three of them carried a stated direction and all three landed: Miami won 1-0, Seattle won 7-6 at Dodger Stadium, and Detroit won 14-0. A fourth read, the day's firmest side, never got played. The value end of the card carried the day and the chalk barely got a chance to.

The scorecard

Three graded reads, three correct. That is the whole record, and the shape of it matters more than the number.

The firmest side on the board, Atlanta behind Chris Sale at Citi Field, was postponed by heavy rain in New York and rescheduled as a split doubleheader Wednesday, with Sale moved to the nightcap. That is the second consecutive day a stated high-conviction side has been washed out before it could grade. For the second straight column, the high-confidence tier contributes nothing to the day's record. Readers should hold that against the result in both directions: Tuesday's 3-0 is not evidence the model's chalk is working, because its chalk did not play.

The fifth read was Astros at Angels, written as a game the model did not separate at a near pick'em board. No side was endorsed, so nothing grades. Houston won 3-2, a one-run outcome that is at least consistent with a game the model declined to call.

A precise full-slate graded record for Tuesday, including the day's accuracy percentage and Brier score, was not retrievable from Verdexed's public accuracy surface at the time of writing, so no figure is claimed here. The verifiable material is the published output and the confirmed final scores, and on that basis Tuesday was three for three.

Best call / worst miss

The best call was Miami, and it worked on precisely the mechanism the column named. The argument was that Philadelphia's listed price, reported between -175 and -185, was charging a heavy favorite rate while handing the ball to Aaron Nola and his 5.82 ERA against Sandy Alcantara. The column said the price ran backward on the pitching matchup alone and then stated the honest caveat: the market had a reason, because Miami's offense had gone cold. Both halves came true. Alcantara worked seven innings on four-hit ball, Kyle Stowers homered, the Miami offense produced exactly one run, and one run was enough. Marlins 1, Phillies 0. The model was right about the arms and right that a cold bat rack could still get there.

There was no graded loss, so there is no miss to report, and this column is not going to invent one. What is worth reporting instead is how close the second win came to being one. Seattle, the second-widest gap on the board at roughly +154, needed Dominic Canzone's tiebreaking two-run homer in the eighth, his second of the game and Seattle's fifth of the night, to get past Los Angeles 7-6. Shohei Ohtani homered, tied the game with an RBI single, and doubled in a run in the ninth before Andres Munoz closed it out. A correct call decided by one swing in the eighth is worth less than the box score suggests, and pretending otherwise is how a scorecard stops being useful.

The Detroit read had no such ambiguity. Troy Melton allowed three hits and two walks over seven innings, Dean Kremer gave up eight runs in four, and the Tigers scored in every inning they batted in a 14-0 win. The column's stated worry about that read was regression against Melton's small sample. It did not arrive Tuesday.

What it means

One 3-0 Tuesday on three graded reads is a good day and almost nothing more. Sweeps of a three-read card happen on ordinary variance, and the confidence tier that would test the model's core sat out in the rain.

The item worth tracking is the one this column has now written three days running from the other side. Sunday's widest listed edge lost on its own logic. Monday's widest listed edge lost on its own logic. Tuesday's widest listed edge won on its own logic, with Alcantara outpitching Nola exactly as the argument required, and the second-widest won too. Two losing days on value reads followed by one winning day is noise, not a correction of a pattern, because there was no established pattern to correct. Three days is not a sample, and the model's calibration is settled on the accuracy page over weeks.

The open question is unchanged and now overdue: the high-confidence tier has gone ungraded two days in a row. Wednesday's doubleheader in New York puts Sale back on the board and gives that tier its first live test since Sunday. Until then, this week's record is a record of the model's edges only, and it should be read as exactly that narrow.

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.