Altus Alpha › Tools › Value Signal

Value Signal Classifier

Compares vig-adjusted Pinnacle fair values with Polymarket prices. Classifies the delta into No Signal / Watch / Trade / Strong Signal and names the direction.

Odds for „yes, the event happens". 1.85 = Pinnacle implies 54 %.

Odds for „no". Both odds together yield the vig.

Polymarket YES price in cents (= implied probability).

How it works

Pinnacle moves its lines first and accepts sharp action — vig-adjusted Pinnacle odds are considered the best available benchmark for fair probabilities. Polymarket prices reflect market-participant expectation; if the Polymarket price diverges from the Pinnacle fair value, that's a potential value signal.

Fair(YES) = (1 / odds_YES) / ((1 / odds_YES) + (1 / odds_NO))

Vig = (1/odds_YES + 1/odds_NO) − 1. Delta = Fair(YES) − Polymarket(YES) in percentage points. Positive sign → Polymarket too cheap → BUY YES. Negative → Polymarket too expensive → BUY NO.

The four tiers

No Signal

< 3 pp delta. Within odds noise. No action.

Watch

3–5 pp. Only worth it at low fees and high volume.

Trade

5–10 pp. Tradeable after spread check and edge verification.

Strong Signal

≥ 10 pp. Clear signal — but check whether there's a known reason.

When the signal goes blunt

On markets without a sharp Pinnacle line (niche sports, exotic events), the benchmark is weak. On large deltas (≥ 10 pp) the sanity check matters: does Polymarket have information Pinnacle doesn't? Has the Pinnacle line not yet updated? Is the Polymarket market illiquid?

What the classifier doesn't replace

The calculator classifies a single-price discrepancy. It says nothing about liquidity (can you actually fill the trade?), slippage (at what average price?), or resolution risk. For position sizing combine the signal with Kelly or EV.

Related

Kelly CriterionExpected Value← All toolsEuropean Edge (Pinnacle)

FAQ

Why Pinnacle as benchmark?
Pinnacle has the lowest vig in the industry (typically 2 %), takes sharp action instead of limiting it, and moves its lines first. Vig-adjusted Pinnacle odds are established in academic literature as the best available benchmark for fair probabilities.
What is the power method for vig adjustment?
Implied(YES) = 1/odds_YES, Implied(NO) = 1/odds_NO. Sum = 1 + vig (overround). Vig-adjusted: Fair(YES) = Implied(YES) / sum. This is proportional scaling — simple, robust, industry standard. Logarithmic and Shin methods exist but are barely better in practice.
Why the 3 / 5 / 10 pp tier thresholds?
Empirical rules of thumb from the Closelook track record. Below 3 pp the difference is within odds noise. 3–5 pp is only worth it at low fees and high volume (Watch). 5–10 pp is tradeable. Above 10 pp there's usually a reason (liquidity, information not reflected in Pinnacle) — still a strong signal.
Does it work without a two-way market?
No — the calculator needs both Pinnacle odds (YES and NO) to determine the vig. For one-sided markets you'd have to estimate the vig or use an industry assumption. Rule of thumb: 2 % for NFL/NBA, 4–5 % for soccer, 6–8 % for niche sports.
Are my inputs stored?
No. The classifier runs entirely in the browser — no server, no tracking. Reload the page and all values are gone.