Hyperthesis LogoHyperthesis

The Conditional Betting Algorithm

The Hyperthesis conditional betting algorithm routes user capital only to outcomes where market prices are better than their own belief. At settlement, exposure is always binary (all-or-nothing), but the maximum matched amount per outcome is strictly limited by conviction.

Try It Yourself

Interactive Conviction Calculator
Side A75%
25%Side B
100% Side BNeutral (50/50)100% Side A

Formula:

Smax(X)=1000×Belief(X)75%S_{max}(X) = 1000 \times \frac{\text{Belief}(X)}{75\%}

Side A

Belief: 75%

Conviction: 100%

Max Exposure: $1,000

Side B

Belief: 25%

Conviction: 33%

Max Exposure: $333

How It Works

1
Placing a Conditional Bet

User submits:

  • A bet size SS
  • A belief vector assigning probabilities to each outcome, summing to 100%

System calculates for each outcome XX:

MaxExposure(X)=S×Belief(X)max⁡(Belief vector)\text{MaxExposure}(X) = S \times \frac{\text{Belief}(X)}{\max(\text{Belief vector})}

The user is only ever matched up to this amount for outcome XX.

2
Real-Time Routing

The system monitors market prices for all outcomes.

No user is ever matched at odds worse than their stated belief.

3
Dynamic Adaptation

While the market is open and the bet is not locked, the user's position remains fluid. If market prices shift, the bet "follows the value":

  • The system continually seeks to match capital against outcomes with edge
  • Positions adapt automatically to market movements
4
Locking

User may lock their position (manually or via time/condition trigger):

  • Exposure is fixed at the current market price
  • Position becomes a standard tradable asset

If not locked, the bet is automatically settled per final market state at event resolution.

5
Settlement

When the outcome is determined:

  • User's full exposure is allocated to the resolved outcome only
  • Exposure is limited by conviction-limited maximum for that side
  • No splitting: all-or-nothing per bet, never partial fills

Numeric Walkthrough Example

Let's see how the algorithm works with three users placing bets on a binary outcome:

Numeric Walkthrough Example

Three users with different belief distributions

UserBet SizeBelief VectorExposure if A winsExposure if B wins
User 1$300A: 75%, B: 25%$300$100
User 2$100A: 20%, B: 80%$25$100
User 3$150A: 50%, B: 50%$150$150
Settlement if A wins
  • User 1: $300 exposure (full, since 75% is max)
  • User 2: $25 exposure ($100 × 20/80)
  • User 3: $150 exposure (full)
Settlement if B wins
  • User 1: $100 exposure ($300 × 25/75)
  • User 2: $100 exposure (full)
  • User 3: $150 exposure (full)

Multi-User and Multi-Outcome Handling

  • For each market, all user bets are pooled
  • Payouts for winners are covered by the capital of losing bets
  • Standard prediction market mechanism applies
  • No sequencing or queuing: all eligible bets are filled proportionally