The official track

Official picks currently come from our ranking model — an experimental ranking system on licensed sports-market data, forward-tested since Oct 2. It ranks a daily slate, applies the 70¢ cap, and freezes the remaining selections, entry prices, and timestamp before play.

The official record contains 25 settled qualifying picks with five pending on Oct 6. Oct 5 settled 3W–2L, +1.26u.

The 70¢ cap

Entries above 70¢ remain in the complete frozen history but are excluded from the official record. Heavy favorites can pad a win rate while bleeding units; the edge we are testing is in plus-money and lower-priced entries.

Through Oct 3 the model also played favorites above 70¢; on Oct 4 we capped entries at 70¢ and restated the record ex-chalk.

The research track

Our in-house research model is the stricter build. It evaluates public prediction-market activity and only trusts what it can reproduce.

The grader runs daily as research. Its picks, passes, wins, and losses stay on a separate ledger. It takes over the official slot only when its own forward record earns it — not because one day looks good.

What the research model checks

The in-house grader applies a stricter qualification bar. A research pick only clears when its filters agree:

• Independent agreement: more than one qualifying signal on the same side, not one loud account.
• Real liquidity: enough market activity that the read is not one tiny order.
• Buying that builds: the position grows over time instead of appearing in one spike.
• No clean opposition: conflicting qualified signals mean no clean read. The model stands down.

That is why research can log a pass while the ranking model still posts an official slate. Different rules, separate ledgers.

Why disagreements stay visible

On Oct 4, the official ranking model took Commanders while the in-house grader took Colts −4.5. Same game, opposite sides. We do not smooth that over. Each result belongs to the model that made it, so the long record can show which approach holds up.

The public record

Every pick is posted with a timestamp before play, at the entry available when the slate freezes. After the event settles, the result goes on its model’s ledger. Flat 1u staking keeps every result comparable.

The previous in-house regime is not erased: France vs Italy, France at 68.5¢, lost −1.00u on Oct 1. It now sits in the research history, where it belongs.

What we don’t do

• ×No cherry-picked streaks. We’ll never advertise “7 of our last 10” while the full record says otherwise. The full record is the record.
• ×No edited picks. Once a row is on a ledger, it stays, graded as it settled, at the price it posted.
• ×We don’t mix the ledgers. Official picks and in-house research are scored separately, always.
• ×No certainty theater. Anyone promising certainty is selling something other than honesty.
• ×No hype. No hype plays, no bombs, no “biggest play of the year” three times a week. Every pick gets its price and its result.

Limitations

• The track record is short. Twenty-one settled official picks cannot prove a durable edge.
• The model is new. The current official model has been live since Oct 2.
• The record was restated. The 70¢ no-chalk cap was disclosed and applied to the record on Oct 4.
• Past performance does not predict future results. A winning start can still fail.

Judge the long record

Twenty-one settled official picks under the 70¢ cap is a start, not a verdict. Q4 2026 is the proving quarter. The ranking model stays official for now; the in-house grader keeps building in research.

See the official record →

Back tomorrow. Same time.