Skip to main content

Weekly Intelligence · Edition 2

Five wins out of six, and the record barely moved

Our best week on stake so far moved the all-time number by three tenths of a percent. That is the whole argument for reading this page instead of a highlight reel.

What is counted here

Every published pick from September 21 to September 27 that cleared our own bar on the day it went out — the same rows anyone could have read on the free board. Nothing is added afterwards and nothing is dropped for losing.

published = true AND data_source = 'Live' AND excluded_from_record = false AND edge_pct >= 0.04 AND confidence >= 3 AND sport NOT IN ('WNBA','NBA','NCAAB') AND game_date BETWEEN 2026-09-21 AND 2026-09-27

Start with the only number that has had time to mean anything. Since June 16 we have published and graded 500 picks on the selection below. They came back 287205–8. Against 451.90 units staked that is -0.29u — a return on stake of -0.1%.

We have won 58.3% of the 492 picks that were decided, and we are still behind on stake. Those two facts belong in the same sentence and almost nobody puts them there. Winning most of your bets and making money are different achievements: a good share of ours are short-priced favourites, where being right four times in five still does not pay for the fifth. A record that reports winners without reporting stake is showing you the flattering half.

This week is the evidence, and it cuts our way

The week made money. 6 of the 8 picks published in this window have been graded, and they came back 51. Against 3.80 units staked that is +2.74u, a return on stake of +72.2%. 2 picks have not graded yet and are in none of these numbers.

That is a better week than we have had all season on stake — and it moved the all-time figure to -0.1%. That is what a good week does to a real record: almost nothing. The arithmetic is not mysterious — 3.80 units of stake cannot move 451.90 very far — but it is the part that gets left out when a week like this is the whole advertisement.

And the week is not finished

These are the picks that have been played so far in the window, by sport: MLB 41 (+1.88u) and NFL 10 (+0.86u). Last week we published 68 picks and 49 of them fell on a single Saturday college-football card, which went 15–25–1 and cost 11.29 units on its own. That Saturday is what made last week a losing one, and this week’s has not been played. Anyone quoting this week’s return at you today is quoting a handful of quiet days.

What we could not show you

We capture the line when a pick goes out and again near the close. Not every pick can carry a closing line, and we would rather say which than quietly show fewer rows. This covers all 8 picks in the window, including the 2 still ungraded.

  • 2 had a closing line we could check against the published side.
  • 1 were moneylines or totals — a moneyline has no line to capture, so there is nothing to compare.
  • 1 carry a stored benchmark on the opposite side to the published pick. We do not trust it, so we show no close rather than a number that might be measuring the wrong team.
  • 1 had no second capture before kick-off — the market was not read again in time.
  • 3 carry no readable published line to compare a close against.

On the 2 we could check, the number we published was better than the close on 2 and worse on 0. That is a measurement of where our price sat, not a claim about what it is worth, and 2 is far too few to read anything into.

What changed on the board

The free board now carries every fixture we track for seven days, not only the ones that became picks — 148 rows this week, and not one of them blank. A fixture with a healthy market states the line, the number of books quoting it and the minute it was captured; a thin or missing market says so in words rather than showing a zero. Picks no longer disappear when their game starts: they stay on their day, marked started, then graded, win or lose. This week’s 1 loss is sitting there with the rest.

Every figure above comes from a stored row you can check. Each pick on the board states the books behind its price and the minute it was captured. The full graded record is public and always has the losses in it, and the board shows what we are watching before anything is posted.

For entertainment and information only. Not betting advice. 21+. Odds move — verify before placing. Past results do not predict future outcomes.