Practice
Game Theory & Decision Science
The expected value is not your expected value. Most decision math quietly assumes the decider gets to keep playing — and owners don’t.
All capabilitiesOur approach
Our perspective.
Game theory, probability, and behavioral analysis applied as working tools, not garnish: payoff maps, negotiation trees, and forecasting discipline an owner can run long after the engagement ends.
Two teams face the same negotiation. The one that priced its walk-away before the first meeting wins it, and neither is smarter. Decision quality is a property of structure: a firm that maps its payoffs, prices its walk-aways, and scores its own forecasts will out-decide a smarter firm that does none of it. For an owner the stakes are unusually final — the sale, the succession, the partner dispute each happen once, and the structure is either in place that day or it is not.
It follows that the firm does not sell forecasts. In fat-tailed domains, the seller of a precise prediction is bluffing; the honest work is payoff design: cap what a loss can cost, leave what a win can pay uncapped, and stop paying for precision the future will not honor. Positioned that way, a client no longer needs to be right more often than the other side. Their errors simply cost less and pay more.
The craft is native here. The firm’s founder competed at the Pokémon World Championships and holds multiple regional top-eight finishes — a game that is nothing but payoff structure, sequence, and variance.
Common challenges
The challenges we help address.
Deciding on the wrong average
The average outcome of a hundred owners taking a risk once says nothing about one owner taking risks a hundred times — because ruin stops the sequence. Every expected-value calculation quietly assumes you get to keep playing. Sizing that survives the sequence comes before any optimization, and no projected return compensates for the point where you must stop.
The confident interval
Ask a leadership team for ranges they are nearly certain of, then score those ranges against what happened: the intervals come back too narrow, and credentials make them narrower, not better. Budgets, timelines, and diligence models inherit that arrogance. The fix is structural: widen every human range, score every estimate against the outcome, and let history simulate what people were asked to guess.
The risk outside the model
The diligence model prices the earnings, the multiple, and the debt service — and misses that the landlord is the seller’s brother-in-law, or that half the revenue rides on one relationship no key-man policy was ever written for. Models built from a game’s own odds are blind to the rule-breaking event outside the game. We build the models anyway, then spend the serious effort on robustness to what they cannot see.
How we work
How the engagement runs.
- Step 1
Diagnose
The decision audit: the recurring decisions that move the most money, each classified by reversibility, by who actually bears the downside, and by payoff shape — linear, convex, or ruin-exposed.
- Step 2
Architect
The games named: counterparties and their incentives mapped, negotiation trees built with the walk-away priced before the first meeting, auction and pricing formats chosen for how the other side must respond — not how the spreadsheet hopes they will.
- Step 3
Build
Decision infrastructure installed: ranged estimates replace point guesses, forecasts are scored against outcomes, simulations run on historical actuals, and sizing rules are pre-committed so no single loss can end the sequence.
- Step 4
Operate
A standing decision review scored on process, not outcome — sound calls that lost are kept, lucky calls that won are flagged. Reading cadence is rationed on purpose: reviewed weekly, noise cancels; watched hourly, it compounds.
Deliverables
What the work produces.
- Decision audit & payoff map
- Decide which recurring decisions deserve engineering, and which shape of error each can afford.
- Negotiation tree & counterparty model
- Decide the sequence, the concessions, and the walk-away before the other side sets them.
- Forecast calibration system
- Decide with ranges that mean something: every estimate scored against what actually happened.
- Sizing & survival rules
- Size the stake so that no single loss ends the sequence.
Evidence
From the case studies.
Related insights
The thinking behind the practice.
Get started