Yesterday's picks: how the model did (September 7)

Verdexed stated no MLB sides for Labor Day, so the graded line for September 7 is zero picks, zero correct, zero incorrect. By this desk's running count that is thirty-four consecutive slates with nothing belonging to the model to score. The model neither won nor lost on Monday, because it did not say anything.

Baseball was again the only place a Verdexed number could have been graded. The NBA and NHL are in their offseasons and the NFL regular season has not opened, so one league played a holiday board and no stated side went out on any game in it.

The scorecard

Zero stated directions, zero correct, zero incorrect for September 7.

The stated-side ledger does not move: 12-7 across 19 decisions, unchanged since August 4.

There is no high-confidence tier to report, because no confidence tier was attached to anything a reader could see. The 65 percent and 70 percent bands this column exists to audit went unpopulated again. No Brier value and no trailing accuracy figure are claimed here, because none was retrievable. The verdexed.com prediction, edge, and accuracy surfaces remain unreachable from the environment these columns are written in. Search returned only the published methodology, which advertises a seven-day accuracy record on the free tier and a full tracked history for Pro subscribers. Neither was readable here, so any accuracy figure quoted elsewhere is unconfirmed by this desk.

There is a small market ledger to fall back on, and it lost. Monday's picks column published no model board but did quote two prices before first pitch, labeled explicitly as market reads rather than model output. Miami was quoted near minus 113 against the Mets, and New York won **9-4**. Toronto was quoted around minus 201 at the Athletics with a total near 8.5, and Sacramento produced eleven runs and a Toronto loss. That is 0-2 on quoted favorites and over on the one quoted total. It is not the model's record, and it is still a losing day for the only numbers this operation staked in advance.

Best call / worst miss

There is no best call and no worst miss belonging to the model, because no side was stated. Naming one after the fact would be inventing a record, and this column will not do that.

What the picks column did read correctly, in shape rather than as a graded pick, was the top of the board. It flagged Braves at Phillies as the clearest separation on starting pitching. Jesus Luzardo then threw the first complete-game shutout of his career in his 150th start, allowing two hits and one walk while striking out 12. Kyle Schwarber's eighth-inning solo home run was the only run scored. Philadelphia won **1-0** and cut Atlanta's National League East lead to four games. A matchup read is not a pick, and a pick is what would have counted.

The worst miss belongs to the quoted prices, and specifically to Toronto. Minus 201 was the widest number on the board, and the column said outright that it was where a model probability would have mattered most. Dylan Cease was scored on in the first inning after a walk, a stolen base, and a ground-rule double. Toronto clawed back late, and Jeff McNeil's bases-loaded infield single walked the Athletics off **6-5**.

One process error deserves its own line. The picks column listed Bryce Elder as Atlanta's probable starter. Grant Holmes started instead, with Elder moved to the bullpen. Naming the wrong starter in a preview whose entire argument was starting pitching is the kind of stale input that would have corrupted a model probability had one been published.

Elsewhere on the confirmed board: Baltimore beat Cleveland **6-4** behind Trevor Rogers and two Coby Mayo home runs, San Francisco beat St. Louis **5-4** in eleven innings, the Dodgers beat Cincinnati **6-3**, and Boston beat the Angels **5-2**.

What it means

Nothing about the model's skill can be read from Monday, or from the five weeks before it. Grading requires stated output, and there has not been any. A blank is not evidence of accuracy and it is not evidence of inaccuracy either, so readers should draw no inference in either direction from a run of zeros.

The only longer-run number this column can honestly anchor to is small and stale: 12-7 across 19 stated decisions, last updated August 4. Nineteen decisions is a sample in which a good process and a lucky one are indistinguishable, and a record that stopped growing five weeks ago is a paused record rather than a good one.

The 0-2 market day means little on its own and should not be dressed up as a signal. Two prices is not a sample, and a minus 201 favorite is expected to lose roughly one time in three, so one such loss is an ordinary Monday rather than a market failure. What has actually gone wrong is upstream of all of it. The pipeline that is supposed to publish, log, and score model probabilities has not delivered one in over a month. When it does, the zeros stop and the grading resumes, wins and losses printed the same way.

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 of legal betting 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.