We pricethe future.
Pantheon trades the two most inefficient auctions in modern finance — the sportsbook and the options chain. Same discipline. Same math. Different arenas.
Most firms look for an edge inside the market.
We build the market, then find where it is wrong.
A price is a claim about the future made under pressure. Sportsbooks publish thousands of them every hour, adjust them in seconds, and are answerable to a crowd that is loud, fast and frequently wrong. Options chains do the same thing in the language of variance.
Pantheon exists in the gap between those published claims and the distribution we can actually defend. We do not predict outcomes. We measure disagreement — and we size it.
§ 02 — MARKETS
Two arenas. One discipline.
Sportsbooks
The last market that still moves on opinion.
Thousands of correlated contracts repriced every minute by operators who must quote a two-sided market whether or not they want the risk. We model the underlying distribution independently, then trade the residual — line shopping, correlated parlays, in-play dislocation and closing-line convergence.
- Independent win-probability models per league
- Cross-book devigging and true-price synthesis
- Live in-play re-pricing on state change
- Correlation-aware exposure across a book of positions
Options
Variance is just a probability distribution with a bid.
The same machinery, pointed at listed equity and index options. We trade the difference between implied and defensible realised variance — surface dislocation, event-driven skew, term-structure kinks around known catalysts, and defined-risk structure rather than naked exposure.
- Implied vs. forecast realised variance
- Skew and term-structure relative value
- Catalyst-aware event pricing
- Defined-risk structures, hard portfolio limits
§ 03 — EDGE
Four rules. No exceptions.
Model, then market
Every position begins with a distribution we built ourselves. If we cannot state the fair price without looking at the screen, we do not have a trade.
Speed where it pays
Latency is not a virtue in itself. We spend it precisely where the half-life of an edge is measured in seconds, and nowhere else.
Risk is the product
Sizing is decided before entry by drawdown tolerance, not conviction. Correlation across sports and strikes is treated as one book, never two.
Adversarial by default
Every model faces a red team whose only job is to find the regime in which it breaks. Backtests are evidence, not proof.
§ 04 — Method
06 stages · one loop
From raw feed to filled order.
- 01
Ingest
Odds, order books, injury wires, weather, lineups, chains and prints — normalised into a single time-aligned event stream.
- 02
Price
Independent models produce a fair distribution for every tradeable contract, with an explicit confidence band.
- 03
Compare
Published prices are devigged and diffed against ours. Only disagreement beyond the confidence band becomes a candidate.
- 04
Size
Fractional-Kelly sizing under a portfolio-level correlation matrix and a hard daily drawdown gate.
- 05
Execute
Routed across books and venues to minimise footprint, slippage and limit attrition.
- 06
Post-mortem
Closing-line value, realised vs. modelled variance, and attribution by model — reviewed on every cycle.
Design targets for the desk currently in build — not a record of past trading.
§ 05 — CAREERS
We are hiring people who argue with the number.
We are at founding stage — the desk is being built now, so the first hires shape it rather than inherit it. You will own a model end to end within a month, and you will be asked to defend it in front of people who want it to be wrong.
- Quantitative researchers
- Sports modellers
- Options traders
- Low-latency engineers
- Data engineers
careers@pantheonofducks.com