When to Fade an Edge
What it is
Not every flagged edge is real. Some are traps: stale lines, prediction-market-only quotes, single-book outliers, or model leans the market has correctly priced past. Knowing when to walk away is as important as knowing when to bet.
When to use it
- Before placing any bet that looks unusually large in EV.
- When the edge depends on a single book or a single market mechanism.
- When you have a gut sense that the price is wrong for a reason you don't understand.
How to read it
The trap checklist
| Warning sign | Why it's a fade |
|---|---|
| Polymarket-only | Prediction markets have low liquidity, late settlement, and structural mispricing that does not graduate to real sportsbook value. Polymarket edges are rarely actionable. |
| Single-book width-1 | Only one book quoting, and the market is tiny (e.g. an obscure prop). Likely a stale or error line that will be pulled before you can lock it in. |
| No Pinnacle quote | Pinnacle is missing from books_used. Without the market's truest anchor, the fair value estimate is shakier. |
| v4_single or v4_fallback method | Only one sharp anchor (or none). Treat as untrusted regardless of headline EV%. |
| Conference dog with KenPom veto | In NCAAB, KenPom-driven home favorites in conference play historically beat the underdog edges our model flags. The KenPom veto exists because we lost money on these. |
| Stale time-to-game | Edge persisting unchanged for hours typically means the line was simply forgotten by the book, and the book will pull it (or void) once they notice. Volume of edges that hit while time-to-game decays is suspect. |
| Books_used count = 1 | One anchor, no cross-check. Even if Pinnacle is the one, treat as v4_pin not v4_full and size smaller. |
| Sport you do not follow | If you cannot tell a friend why this edge exists in one sentence, you do not understand the bet. Skip. |
The two-condition rule
If two or more of the above are true for a single edge, fade unconditionally. If one is true, downsize to 0.5u or skip; do not bet at full tier sizing.
Worked example
A trap that looked great
Selection: Random small-conference dog +320 Book: Single soft book v4 method: v4_single (Pinnacle missing) EV% claimed: +9.4% KenPom: Strongly disagrees (favorite home efficiency edge huge) Time to tip: 18 hours, line unchanged Trap signals: 3 (single book, no Pinnacle, KenPom veto) Verdict: skip.
This edge would have shown +9.4% EV in All Edges. It was a trap on three dimensions. We have logged dozens of these losing as expected. The +9.4% is mostly variance from a stale or misposted price.
A real edge for contrast
Selection: NFL favorite -3.5 -110 Book: FanDuel v4 method: v4_full Books_used: [pinnacle, lowvig, circa, betonlineag] EV% claimed: +3.2% Game Lab: Agrees (model has favorite -4.8) Time to game: 4 hours Trap signals: 0. Bet at full tier-A size.
Common mistakes
- Chasing the EV column. The biggest EV is often the biggest red flag. Highest EV with v4_full + multiple anchors is what you want, not just highest EV.
- Trusting prediction markets as fair value. Polymarket is a market, not a sharp anchor. Different beast.
- Overriding the KenPom veto in CBB. It is there because the data said so. Trust the data.
- Betting a sport you do not understand because the number looks great. EV is computed from prices, not from common sense. Some prices are wrong for reasons you cannot see.
- Ignoring stale-line patterns. An edge that has been live for 6 hours unchanged is rarely actionable. The book will fix it or void it.