Streaks · Weeks 1–3

A 9-of-10 streak is not an edge.

A 9-of-10 streak is not an edge. It is a coin flip you pay for. Through three weeks we posted every NFL prop where the player had cleared the line in at least 7 of his last 10 games — the same hot props sold everywhere — and graded all of them against the box score. They hit 51.2% (123–117). The 9-of-10 and 10-of-10 props hit 47.5%. Those props were priced at an average of -117.5, which means you needed 54% just to break even. The line already knows about the streak; that is why it is the line. We publish hit rates because they are useful context for asking questions, not because they predict the next game — and we grade every one so you can check us.

Key numbers

Claim being tested	n	W–L	Hit rate	95% CI	Needed at -110	Result
Hot prop (7 or more of last 10 over the line): Over hit	240	123–117	51.2%	45–57%	52.4%	Below 52.4%, CI spans breakeven
9 or 10 of last 10 (incl. 8 of 9, 7 of 8): Over hit	40	19–21	47.5%	33–62%	52.4%	Below 52.4%, CI spans breakeven
10 of 10: Over hit	6	3–3	50.0%	19–81%	52.4%	Below 52.4%, CI spans breakeven, n<30
Last 5 all over: Over hit	39	24–15	61.5%	46–75%	52.4%	Above 52.4%, CI spans breakeven
Last 3 all over: Over hit	90	52–38	57.8%	47–67%	52.4%	Above 52.4%, CI spans breakeven
Cold prop (0 to 3 of last 10): Under hit	254	140–114	55.1%	49–61%	52.4%	Above 52.4%, CI spans breakeven
Hot Receptions props: Over hit	72	42–30	58.3%	47–69%	52.4%	Above 52.4%, CI spans breakeven
Hot Rush Yds props: Over hit	51	22–29	43.1%	30–57%	52.4%	Below 52.4%, CI spans breakeven

Set B, over/under markets only (Pass Yds, Rush Yds, Rec Yds, Receptions, Pass TDs). Hot rows are read from the Over side, cold rows from the Under side. CI = 95% Wilson score interval.

Every interval in that table spans 52.4%. Three weeks of grading is enough to show that streaks are not an edge; it is not enough to show that they lose. This page re-runs every week after grading.

How to read this

One prop = one player × one market × one slate, on the main line at post time. Hit = actual ≥ posted threshold ('26+ Rush Yds' from a 25.5 line). Anytime TD hit = rushing + receiving TD ≥ 1. Every O/U line in the data is x.5, so there are no pushes.
L10 at post time = share of the player's last 10 regular-season games (2025 + 2026 played) that cleared the current line. Last-5 = the last five of those, oldest → newest. Both come straight from the pack that was published before kickoff.
Only HIT/MISS rows count. DNP, VOID and NOMATCH rows are excluded from every cell.
Two sets. The pipeline only posts (and so only officially grades) props with L10 ≥ 7/10 and games_used ≥ 8. That set says nothing about cold props, so: Set A, official graded set = the rows in GRADES_W0N.csv for every graded slate, joined to the feed (last-5, price) and the pack CSV (line move). Set B, extended set (superset of A) = Set A plus every other row in the pack CSVs (Thu/Sun/Mon) that was matched with games_used ≥ 8, graded here against the pack's latest line with the identical rule and the identical public box scores. This is the only place the 0–6/10 buckets exist. Caveat: the pack is curated toward extremes (L10 ≥ 0.70 or ≤ 0.30, big movers, injury context), so the 4–6/10 bucket is mostly line-movers, not a random slice of the board.
Breakeven. At -110 you need 52.4%. The 95% CI is a Wilson score interval; when it spans 52.4% the cell cannot tell an edge from a coin flip.

Row accounting

Set	Rows	Hit	Miss	DNP	Void	Graded (hit + miss)	Over/Under	Anytime TD
A · official graded set	257	123	118	16	0	241	232	9
B · extended set (A + every other matched pack row)	871	363	484	23	1	847	550	297

Full tables, by week, by market, by last-5 and by line move: /streaks/data.json. Grading rule and field notes for AI tools: /llms.txt.

Graded against the public box score at the posted line. Updated 2026-09-30 18:40 ET. 21+ · informational only.