Two Wins, One Loss, And A Win This Site Argued Against

Does this site's record prove anything yet?

No, and the reason is narrower and more awkward than a small sample. Ten calls sit on Called Shots. Three have been settled: two right, one wrong. Neither of the two wins is attached to an article. The loss is. And on one of those wins, this publication did eventually put an argument in print, two weeks later, taking the opposite side.

That is the finding. The rest of this is the arithmetic.

The three settled calls#

The build prints a warning on every run naming ledger entries with no article behind them. It has been printing these two since they were entered. Nobody acted on it.

The Stenberg problem#

On May 19 the ledger recorded a call to take Ivar Stenberg over Chase Reid at No. 2 overall.

On June 2 this site published a piece headlined "What the Sharks should do at No. 2: take Reid, and the one move that's even better." Its bottom line reads, in bold: "Draft Chase Reid." Directly beneath that, also in bold: "Do not draft Ivar Stenberg." The piece is careful and it is not a throwaway. It concedes Stenberg is the better pure prospect and argues he is the wrong pick anyway, which is a real distinction honestly made, and it works the problem from roster need rather than from the board.

On June 26 San Jose took Stenberg. Reid went seventh, to Seattle.

The ledger scores the May call a win. The article is still on the site, still reachable, and still says take Reid.

Both things are true at once, and that is permitted. A writer is allowed to hold a position in May, talk himself out of it in June, and be right the first time. What is not permitted is a public record that keeps the correct version and leaves the published version out of the accounting, because the published version is the one readers actually saw, and it is the one that was wrong.

The order matters here and makes it worse rather than better. The unpublished call came first. The published argument came second and moved away from it. So this site's most recent stated position before the draft was Reid, and the entry on the scoreboard is Stenberg.

To the site's credit, the piece was flagged internally as superseded rather than quietly pulled, and the house style file already warns that this specific pair of items must never be linked to each other by matching dates. Somebody saw the collision. Nobody drew the conclusion from it.

The remedy is not to delete anything. It is to make the ledger reflect what happened: the entry should say the call was made privately on May 19, that a piece taking the opposite position ran on June 2, and that both are part of the record. A ledger that records only the version that aged well is not a ledger. It is a highlight reel with dates on it.

As of this column, that note is on the entry. The verdict is unchanged, because the May call was right and softening a verdict to settle an argument about provenance would be its own kind of dishonesty. What changed is that a reader clicking the row now learns the site published the other side.

The Klingberg entry has the milder version of the same disease. It was correct and it is unattached, so a reader has no way to confirm the thought existed in public before the outcome did. That is not an accusation of backdating. It is the observation that the ledger as built cannot distinguish a call made in advance from a call remembered fondly, and a ledger that cannot make that distinction is decoration.

The loss, which was graded properly#

The Trouba call is the one piece of this that works, and it works because of how it was scored rather than how it turned out.

The site predicted a rising cap would bid defensemen up and priced Trouba near three years and $17 million. He signed for four years and $33 million. The consensus board had him at two years and $3.5 million, so the call was directionally right and closer than the field by a wide margin.

It is filed as a miss. The note attached says being less wrong than everyone else is not being right. That is the correct standard, and the first time it costs something is exactly when it is worth keeping.

What the miss is not charged with is bad process. The reasoning about cap inflation was sound and the market confirmed it; a single magnitude was badly wrong. Those are different failures with different repairs, and a column that treats every miss as proof of sloppy thinking gets ignored, which is the outcome this one is trying to avoid.

The correction file#

One correction has been issued, on July 25, and it was a double.

A piece arguing the Sharks overcommitted last offseason had held up an alternative plan as one that was available. Vincent Desharnais was listed near two years and $2 million; he had signed in Washington on July 1 for four years and $16.8 million. John Carlson was presented as a gettable premium addition; he signed in Tampa Bay, and he had said throughout that he wanted to be back east near family.

The correction sits in the open at the foot of the piece, states both errors plainly, and calls them errors. The policy worked.

The failure underneath is repeatable and worth naming. The Desharnais figure came from a planning document written before free agency and reached print as a fact about the world rather than an assumption about the future. A counterfactual asserts a better path was purchasable, which makes every leg a claim about the market, and claims about the market have to be checked against it.

Checking the inconvenient numbers also improved the piece. The real prices, $4.0 million, $4.2 million and $5.0 million for comparable defensemen against San Jose's $8.25 million, indict the contract more effectively than the invented alternative did.

Silva, and the temptation to look busy#

The numbers desk opened this week and filed three of the ten entries on the ledger in one afternoon, all resolving on or around August 3. Thirty percent of the site's record now rests on one week of one sport and gets graded in nine days. That is the job and it is fine.

One of the three should not stand in its present shape. The call is that Luis Arraez is traded and that the headline return is not a consensus top-100 prospect. The attached note concedes ESPN put his trade probability at 90 percent and that the first half is "close to a freebie." That concession is honest, and it is also an argument against the entry. A two-part call whose first part is nearly certain reads as one call, scores as one call, and carries materially less risk than the format implies.

Split it, or drop the freebie and file only the half that can hurt. Silva's own method section requires a claim be scoreable by a stranger holding a box score. The stricter form of his rule is that a claim should also be able to cost something, and since he wrote it, he can be held to it.

Thompson's debut carries no call, which is correct. Features are not obliged to predict. The build's warning about long pieces without takes does not distinguish a feature from an analysis piece, and until it does, that warning wants a human reading it rather than being obeyed.

It also only looks at hockey. The check that flags long pieces carrying no call is written to examine Sharks articles and nothing else, so a lengthy and opinionated piece anywhere else on the site is invisible to it. This column qualifies on length and predicts nothing, and the reason it does not turn up on the list is not that somebody decided audits are exempt. It is that the list has never been asked about anything but one team.

Two standing problems#

Five long pieces carry no falsifiable call at all: the Chase Reid scouting report, the Graf player review, the Trouba evaluation, the Kesselring deep dive and the Toffoli line review. Each argues a position at length. A site whose stated rule is that every call gets graded should not run that many thousands of words of judgment the ledger never sees.

And this is a Sharks ledger with a Giants annex. Six entries on hockey, four on baseball, none on the 49ers, the Warriors, Michigan football, Michigan basketball or the college basketball model, several of which have published pages and one of which has a roster tracker due to be graded on cut day. The accountability apparatus covers roughly the sections it was born in.

This column does not review this column. This publication has been made aware of the arrangement and has declined to correct it.