Every pick, graded in public.
This is the whole ledger: every pick DiamondEdge has served since Jul 1, 2026, graded at the price and the line it was served at. Nothing here is ever edited or deleted, and nothing is restated — losses stay up. The losing nights are on this page at the same size as the winning ones, because a record you can only read the good half of is an advertisement — and so is a record that does not tell you which of its nights were called in advance. Ours does, below.
How to read this honestly
Most services show you a number like 259-248-19 and let you assume it means an edge. Here is what it actually supports, and what it does not.
526 picks is a small sample, and it is small in a specific, measurable way. A 95% Wilson interval around 51.1% on 507 decided picks runs from 46.7% to 55.4%. A rule that truly wins 46.7% of the time and a rule that truly wins 55.4% of the time would both produce a run like this one often enough that we could not tell them apart yet. The break-even hit rate at the standard −110 price is 52.4% — and it sits inside that interval. So this record is not yet proof of an edge. It is a start, kept in public so it can become one — or not.
The +4.73u did not come from the hit rate. 51.1% is below the 52.4% you need to break even at −110. Laid at a flat −110, this exact 259–248 would have returned −12.55u. It returned +4.73u instead — a +17.27u difference that came from the prices the picks were served at, not from being right more often. That is a real part of how the product works and it is a fragile one: it depends on getting the number you were quoted.
Both figures in this box are computed from the win and loss counts above — Wilson score interval at 95% over decided picks, and 100/110 per win against 1 per loss for the flat-price line. Everything else on this page is served exactly as it is graded.
Which of these nights were called in advance
A record only means something if the call came first. Not all of this one did. The data behind this page stamps every night with how it got here, and the honest thing to do with that stamp is print it.
Every night below was graded identically — one row per game, at the price and line that night’s ticket carried. What differs is when the rule that made those picks was fixed.
| How the night was fixed | Nights | Picks | Record | Hit | Units |
|---|---|---|---|---|---|
| Rule stamped before first pitch | 7 | 50 | 21-22-7 | 48.8% | −2.21u |
| Process re-run afterwards | 63 | 476 | 238-226-12 | 51.3% | +6.94u |
| The record above | 70 | 526 | 259-248-19 | 51.1% | +4.73u |
Read the two rows against each other: the profit in this record is in the re-run half. The 50 picks on nights whose rule was stamped before the first game are 21-22-7, −2.21u. The 476 picks on re-run nights are 238-226-12, +6.94u. Both are in the headline, because both were graded honestly — but only one of them is evidence that this desk calls games before they are played, and it is currently the losing one.
What “re-run afterwards” does and does not mean
It is not a rule picked because it won and then run backwards over the season. For each night, the selection is re-derived from a window that ends strictly before that night, that fence is checked per night rather than assumed, and the night’s games are graded against whatever rule the search produced. Nothing dated on or after a night can reach into its own selection.
What it cannot show is that the desk would have run that rule live. The selector doing the re-running was built later, with those nights’ results already in the world. That gap is the entire difference between the two rows, it is not something a fence can close, and it is why they are printed apart instead of averaged into one comfortable number.
41 of those nights could have been counted either way. Aug 10, Aug 11, Aug 12, Aug 13, Aug 14, Aug 15, Aug 16, Aug 18, Aug 19, Aug 20, Aug 21, Aug 22, Aug 23, Aug 24, Aug 25, Aug 26, Aug 27, Aug 28, Aug 29, Aug 30, Aug 31, Sep 1, Sep 4, Sep 5, Sep 6, Sep 8, Sep 10, Sep 11, Sep 12, Sep 13, Sep 14, Sep 15, Sep 16, Sep 17, Sep 18, Sep 19, Sep 20, Sep 21, Sep 22, Sep 23 and Sep 24 are a closer call: the selector’s own store says their rule was chosen hours before first pitch, but the as-served ledger carries no stamp for them. They are counted as re-run above, because the as-served ledger is the evidence that survives a change of selector version and the selector’s own store is not. Counting them the other way would make the called-in-advance side 394 picks at 191-188-15, −2.64u — worse, not better. Both numbers are here so that nobody has to take our word for which one we chose.
The two rows above are the same rows as the night-by-night table further down, summed the same way the headline sums them, and they add back to it exactly — 50 + 476 = 526 picks. No pick is counted twice and none is left out.
Month by month
Both months are here, in the order they happened, at the same weight.
| Month | Picks | Record | Hit | Units | Return |
|---|---|---|---|---|---|
| Sep 2026 | 166 | 74-88-4 | 45.7% | −16.52u | −10.2% |
| Aug 2026 | 246 | 126-108-12 | 53.8% | +14.72u | +6.3% |
| Jul 2026 | 114 | 59-52-3 | 53.1% | +6.53u | +5.9% |
The streaks, both directions
Over 507 decided picks. Pushes are neutral — they neither extend nor break a streak.
Every night, in order
One row per night the board posted picks, newest first. The last column is the rule that was in force for that night — it changed 13 times across these 70 nights, and the record counts every one of them straight through. A product that quietly restarts its record when the model changes is not keeping a record.
The mark beside each night is the one from the section above: stamped means that night’s rule was written down before its first game; re-run means the process was replayed for it afterwards behind its own fence. 7 nights and 63 nights respectively.
| Night | Called | Picks | Record | Units | Rule in force that night |
|---|---|---|---|---|---|
| Thu, Sep 24 | re-run | 8 | 5-3-0 | +1.59u | The rule that won the search |
| Wed, Sep 23 | re-run | 10 | 3-7-0 | −4.06u | The rule that won the search |
| Tue, Sep 22 | re-run | 9 | 3-5-1 | −2.08u | The rule that won the search |
| Mon, Sep 21 | re-run | 3 | 1-2-0 | −1.04u | The rule that won the search |
| Sun, Sep 20 | re-run | 13 | 5-8-0 | −2.99u | The rule that won the search |
| Sat, Sep 19 | re-run | 10 | 3-7-0 | −3.98u | The rule that won the search |
| Fri, Sep 18 | re-run | 11 | 4-7-0 | −3.10u | The rule that won the search |
| Thu, Sep 17 | re-run | 6 | 0-5-1 | −5.00u | The rule that won the search |
| Wed, Sep 16 | re-run | 12 | 7-5-0 | +1.57u | The rule that won the search |
| Tue, Sep 15 | re-run | 9 | 6-3-0 | +2.98u | The rule that won the search |
| Mon, Sep 14 | re-run | 3 | 0-3-0 | −3.00u | The rule that won the search |
| Sun, Sep 13 | re-run | 3 | 2-1-0 | +0.88u | The rule that won the search |
| Sat, Sep 12 | re-run | 10 | 5-5-0 | +0.02u | The rule that won the search |
| Fri, Sep 11 | re-run | 5 | 2-3-0 | −1.00u | The rule that won the search |
| Thu, Sep 10 | re-run | 1 | 1-0-0 | +0.98u | The rule that won the search |
| Tue, Sep 8 | re-run | 12 | 6-5-1 | +0.49u | The rule that won the search |
| Sun, Sep 6 | re-run | 5 | 3-2-0 | +0.72u | The rule that won the search |
| Sat, Sep 5 | re-run | 11 | 5-6-0 | −1.17u | The rule that won the search |
| Fri, Sep 4 | re-run | 12 | 6-6-0 | −0.09u | The rule that won the search |
| Thu, Sep 3 | re-run | 7 | 3-4-0 | −1.08u | The Full Committee — 31 strategies vote |
| Tue, Sep 1 | re-run | 6 | 4-1-1 | +2.85u | The rule that won the search |
| Mon, Aug 31 | re-run | 6 | 4-2-0 | +1.88u | The rule that won the search |
| Sun, Aug 30 | re-run | 14 | 9-4-1 | +5.07u | The rule that won the search |
| Sat, Aug 29 | re-run | 9 | 2-6-1 | −4.16u | The rule that won the search |
| Fri, Aug 28 | re-run | 9 | 5-3-1 | +1.86u | The rule that won the search |
| Thu, Aug 27 | re-run | 6 | 2-4-0 | −1.95u | The rule that won the search |
| Wed, Aug 26 | re-run | 13 | 8-5-0 | +2.70u | The rule that won the search |
| Tue, Aug 25 | re-run | 11 | 7-4-0 | +2.87u | The rule that won the search |
| Mon, Aug 24 | re-run | 8 | 6-2-0 | +3.72u | The rule that won the search |
| Sun, Aug 23 | re-run | 12 | 6-5-1 | +0.95u | The rule that won the search |
| Sat, Aug 22 | re-run | 9 | 5-4-0 | +0.84u | The rule that won the search |
| Fri, Aug 21 | re-run | 7 | 4-3-0 | +0.76u | The rule that won the search |
| Thu, Aug 20 | re-run | 8 | 5-3-0 | +1.95u | The rule that won the search |
| Wed, Aug 19 | re-run | 10 | 6-4-0 | +1.98u | The rule that won the search |
| Tue, Aug 18 | re-run | 14 | 6-8-0 | −2.17u | The rule that won the search |
| Sun, Aug 16 | re-run | 12 | 8-4-0 | +4.33u | The rule that won the search |
| Sat, Aug 15 | re-run | 14 | 6-8-0 | −2.45u | The rule that won the search |
| Fri, Aug 14 | re-run | 3 | 2-1-0 | +0.96u | The rule that won the search |
| Thu, Aug 13 | re-run | 2 | 0-2-0 | −2.00u | The rule that won the search |
| Wed, Aug 12 | re-run | 5 | 4-1-0 | +3.04u | The rule that won the search |
| Tue, Aug 11 | re-run | 6 | 2-4-0 | −2.12u | The rule that won the search |
| Mon, Aug 10 | re-run | 7 | 2-5-0 | −3.03u | The rule that won the search |
| Sun, Aug 9 | stamped | 13 | 6-5-2 | +0.49u | The rule that won the search |
| Sat, Aug 8 | stamped | 13 | 3-7-3 | −4.15u | The rule that won the search |
| Fri, Aug 7 | stamped | 6 | 3-3-0 | −0.18u | The Full Committee — 21 strategies vote |
| Thu, Aug 6 | stamped | 2 | 1-1-0 | −0.09u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Wed, Aug 5 | stamped | 4 | 1-3-0 | −2.00u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Tue, Aug 4 | stamped | 5 | 3-1-1 | +1.83u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Mon, Aug 3 | stamped | 7 | 4-2-1 | +1.89u | Shallow neural desk ensemble (0.54 margin) |
| Sun, Aug 2 | re-run | 8 | 5-3-0 | +1.96u | Shallow neural desk ensemble (0.54 margin) |
| Sat, Aug 1 | re-run | 3 | 1-1-1 | −0.04u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Fri, Jul 31 | re-run | 3 | 3-0-0 | +2.91u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Thu, Jul 30 | re-run | 3 | 3-0-0 | +2.92u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Wed, Jul 29 | re-run | 7 | 4-3-0 | +1.01u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Tue, Jul 28 | re-run | 7 | 5-2-0 | +3.35u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Mon, Jul 27 | re-run | 4 | 2-2-0 | −0.04u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Sun, Jul 26 | re-run | 6 | 4-2-0 | +2.03u | Adaptive room-shape rule: 2 of 4 OVER → UNDER |
| Sat, Jul 25 | re-run | 9 | 3-6-0 | −2.84u | Regularized desk meta-model (0.46 margin) |
| Fri, Jul 24 | re-run | 7 | 5-2-0 | +2.81u | Stacked analyst-rule model (top 5, 2+ signals) |
| Thu, Jul 23 | re-run | 1 | 0-1-0 | −1.00u | Fade the conditions analyst + the hitting analyst + the pitching analyst when that block agrees |
| Wed, Jul 22 | re-run | 9 | 1-7-1 | −6.05u | Regularized desk meta-model (0.46 margin) |
| Tue, Jul 21 | re-run | 7 | 2-5-0 | −3.00u | Regularized desk meta-model (0.46 margin) |
| Mon, Jul 20 | re-run | 10 | 8-2-0 | +5.79u | Probabilistic analyst blend (0.4% edge) |
| Sun, Jul 19 | re-run | 12 | 7-5-0 | +1.96u | Probabilistic analyst blend (0.4% edge) |
| Sat, Jul 18 | re-run | 5 | 1-3-1 | −2.02u | Fade the conditions analyst + the hitting analyst + the pitching analyst when that block agrees |
| Fri, Jul 17 | re-run | 10 | 4-5-1 | −1.15u | Weighted four-analyst form blend (6% threshold) |
| Thu, Jul 16 | re-run | 1 | 1-0-0 | +0.96u | Fade the conditions analyst + the pitching analyst when they agree |
| Sun, Jul 12 | re-run | 11 | 6-5-0 | +0.89u | Probabilistic analyst blend (0.4% edge) |
| Fri, Jul 3 | re-run | 1 | 0-1-0 | −1.00u | Probabilistic analyst blend (1.8% edge) |
| Thu, Jul 2 | re-run | 1 | 0-1-0 | −1.00u | Probabilistic analyst blend (1.2% edge) |
What counts, and what does not
every game DiamondEdge served as an official pick, graded at the price and line it was served at, one row per game, across every engine change. Nothing re-graded, nothing restated, and no part of it is a strategy replayed over days it did not actually play.
One clause there is worth pinning down. “No strategy replayed over days it did not play” means the record is never one favourite rule run backwards over the season — each night is graded under a rule derived for that night and nothing else. It is not a claim that every night’s rule was fixed before the games; on 63 of these 70 nights it was not, and the split is above.
A pick is frozen before the game — side, line and price — and it is graded against that frozen ticket whatever the number moves to afterwards. Picks are counted one row per game. Units are flat: one unit a pick, every pick, no staking plan doing quiet work in the background.
This page shows the MLB product record, which is the only record DiamondEdge has run long enough to publish. NFL, WNBA and MLS boards keep their own separate ledgers from the day each one armed, and they are never blended into the figure above. Records are updated as nights finish grading; this page is current through Sep 24, 2026.
Every pick, every write-up and the full record are free, to every reader — one honest call per game, frozen before first pitch, graded here win or lose.
See today’s board