This article is part of the SaskPoly story. Read how it all started.
The Total Review: Every Call the Desk Has Made, Tallied
The full ledger, counted: 518 resolved model calls at 56.0% (+62 flat), 409 bet-grade calls at 55.3%, the home-run tracker's color gate nearly doubling green over red, and two survivor tickets standing 3-0 — with the caveats printed in the same size type as the wins.
Once a season it is worth stopping the daily machinery and asking the only question that matters: what does the ledger say? Below is every call this desk has published or generated, counted from the sources themselves — the SBFG outcome export, the Home Run Tracker v2 backtests, the survivor decision log, and the market boards. Wins first, caveats in the same size type, nothing rounded in our favor.
The SBFG moneyline model: +62 on 518 resolved calls
The Smart Bet Field Guide export covers 61 report days, July 6 through September 7 — 825 baseball games graded against the market (the ledger).
| Grade | Calls | Record | Win rate |
|---|---|---|---|
| Bet (home/away) | 409 | 220–178, 11 pending | 55.3% |
| Lean | 121 | 70–50, 1 pending | 58.3% |
| Combined | 530 | 290–228 | 56.0% |
| Pass — no edge | 295 | — | 35.8% of the board walked away |
Read it the way the field guide teaches: the edge is not any single call, it is the ledger at scale. Fifty-six percent across five hundred-plus calls is a real, bankable number if the prices hold — and the pass column matters as much as the bet column. More than a third of all games got no call at all, which is the model declining to invent an edge that isn't there.
The caveats, printed honestly: the export records wins and losses, not closing prices, so this is a flat-stake tally, not an ROI figure — a 56% record at short prices and a 56% record at even money are not the same business. It is self-graded, from the same machinery that makes the calls. And the window ends September 7, three weeks short of yesterday's finale; the full-season ledger belongs to the archive and can be appended when it lands.
The Home Run Tracker v2: the gate is the product
The backtest files grade themselves now — that was the whole point of the rebuild — and the latest window (97 games, 269 picks) reads: #1-pick hit rate 23.3%, overall 26.4%, and the color tiers that replaced the old flat number doing exactly what they promised: green 35.6%, yellow 40.9%, red 20.6%. Green nearly doubles red. Over thirteen unique backtest dates in the local archive the daily tally runs rougher — 486 picks, 18.3% on the top selection across the counted days — but the trend since the August rebuild runs the right direction, and the honest summary of a home-run model is that it will never look like a moneyline model: the events are rarer, the variance fatter, and the only claim worth making is the one the tiers support. The tiers support it.
The survivor portfolio: two tickets, 3-0 and 3-0
Six picks, five wins, one expensive education:
| Ticket | Weeks 1–3 | Standing | Burned |
|---|---|---|---|
| Silver Bullet | DET (OT), SF, KC | 3-0 | DET, SF, KC |
| ESPN entry | JAX, SF, BUF | 3-0 | JAX, SF, BUF |
| Classic III | TB | eliminated, Week 2 | — |
Around the picks, the published reads: Tampa Bay named the pick to avoid in Week 2 — it lost. Carolina named the tourist trap in Week 3 — it lost. The Week 3 board read the chalk correctly everywhere it mattered. Against that, one genuine error — the prediction that our pool would pile onto Kansas City — published, corrected within a day, and graded in yesterday's post. And the swap: Green Bay to Buffalo before Thursday's lock, a deviation logged as a deviation, which saved the ESPN entry when the Falcons routed the Packers. That one was luck, and the column has said so three times now. It keeps needing saying because it is the easiest thing in this business to forget.
The market boards: 2-for-2 on Game 162
The final-day MLB board priced Houston at 66.9% and Philadelphia at 87.4%, with the tiebreaker asymmetries doing the analytical work. Both favorites paid. One owned error in the same post — the Phillies' opponent mislabeled — corrected on the page and in the wrap. The board's record is two markets, two winners, one correction; we'll take that ratio.
The bottom line
Five hundred thirty model calls at 56%, a home-run gate that sorts green from red at nearly two-to-one, a survivor portfolio at six wins against one death, and a market board that closed the baseball season on a sweep. No month of that is a guarantee about the next one — three weeks of baseball sit outside the export, flat-stake tallying understates what good prices would earn and overstates what bad ones did, and the survivor tickets are three good Sundays old. What the tally is: proof that the process, run honestly and graded in public, holds up when somebody finally counts. Somebody counted. Bring your own pencil.
Sources: SBFG calls export (July 6–Sept 7), HRT v2 backtest reports, the desk survivor decision log, and published market boards. Tallies computed from the files; errors in counting are the desk's own.
Correction, Sept. 28 (evening) — read the whole page, not one column. The pool's full pick history is now visible: the middle column is Week 2, the first column is Week 3. The pool's Week 3 pile was Kansas City — roughly 55–60 of the alive entries, the desk included — with San Francisco held by only about 5–7. The "54% SF" figure this post relied on was the pool's Week 2 history, and this desk's Saturday correction corrected an error into a different error. The pick, the results, and the ticket grades are unaffected; the field-position framing was wrong. Third lesson of the month on this pool page: never read one column.