Reading the v4 Enrichment
What it is
Every edge in our system is enriched with v4 metadata: how we computed the fair value, which books contributed, and how confident we are. If you understand v4, you understand whether to trust the edge.
When to use it
- Whenever you are about to size a bet up.
- Whenever an edge looks larger than usual.
- Whenever Pinnacle is missing from the books_used list.
How to read it
| Field | Meaning |
|---|---|
| v4_blended_prob | The final consensus fair probability for that selection. The number we use to compute EV%. |
| v4_method | How the blend was computed. Codes below. |
| v4_books_used | List of sharp books that contributed (e.g. [pinnacle, lowvig, circa]). | tr>
| v4_consensus | The numeric blended fair value displayed alongside individual book quotes. |
| v4_pinnacle_fair | Pinnacle's solo devigged fair value. Useful as a sanity check independent of the blend. |
| v4_weight | The total weight successfully applied (e.g. 1.0 = full method, 0.45 = Pinnacle only). |
| v4_anchor_count | How many sharp books contributed. 3+ is ideal. |
v4_method codes
| Code | Meaning | Confidence |
|---|---|---|
| v4_full | All three primary sharp anchors contributed (Pinnacle 45%, LowVig+Circa 30%, BetOnlineAG 25%). | Highest |
| v4_pin_lowvig | Pinnacle plus LowVig or Circa contributed; BetOnlineAG missing. | High |
| v4_pin | Only Pinnacle had a quote, weight renormalized to 100%. | Medium |
| v4_consensus_no_pin | LowVig/Circa/BetOnlineAG contributed but Pinnacle is missing. Less anchored. | Medium |
| v4_single | One sharp book only, not Pinnacle. Use with care. | Low |
| v4_fallback | No sharp books available, fell back to wider market median. | Lowest |
Worked example
Example field readout
selection: Yankees -1.5 +145 v4_blended_prob: 0.421 v4_method: v4_full v4_books_used: [pinnacle, lowvig, circa, betonlineag] v4_consensus: +138 (fair odds) v4_pinnacle_fair: +141 v4_anchor_count: 4 EV% computed: +3.4% Tier: A
This is the cleanest possible enrichment: all four sharp anchors quoting, Pinnacle agrees with the blend within 3 cents, multiple confirmations. You can bet this with confidence at full size.
Contrast with:
selection: Some +700 longshot on Polymarket v4_blended_prob: 0.085 v4_method: v4_single v4_books_used: [polymarket] v4_anchor_count: 1 EV% computed: +18.5%
Massive EV but only one source, and that source is a prediction market not a sharp book. Treat as speculative or skip. See When to Fade.
Common mistakes
- Sizing v4_single the same as v4_full. Fewer anchors = wider error bars. Reduce size proportionally.
- Trusting v4_fallback at face value. Fallback means no sharp anchor. Treat as untrusted unless backed by Game Lab / Player Lab agreement.
- Ignoring the Pinnacle-vs-blend gap. If Pinnacle says one thing and the blend says another by more than ~2 cents, dig deeper.
- Confusing books_used with all books quoting. books_used is the sharp anchors used to compute fair value. Other books may be quoting too but they don't anchor anything.
- Reading EV% without reading method. A 5% EV under v4_full is real. A 5% EV under v4_single might be noise.