Part 1Marc de Batlle
0%
Part 1Marc de Batlle
All work08 / 11
Nine

Polymarket Arbitrage BotA volatility straddle on prediction markets.

Chapter concepts
  1. 1A passive-arbitrage strategy that trades the volatility around 50/50 Valorant markets instead of trying to pick winners.
  2. 2Role: Solo build. Status: Backtested.
  3. 3Stack: Python, pandas, statsmodels.
  4. 4Key figure: +11% estimated EV per trade.
08
Period2025 · Dec
RoleSolo build
StatusBacktested

The bet

Most bots try to predict who wins. This one does not. Around a coin-flip match (~50/50), in-game events make the odds overreact: a pistol round, an economy swing, a retake. I buy both sides, sell the collapsing side into the overreaction, and hold the favorite to resolution.

Review of
Formulas

1.q > 1 − pc + c
(9-1)
q
is probability the favorite wins
pc
is exit price of the cheap side
c
is fees + slippage buffer

The trade is positive expected value whenever this holds.

Exhibit 9-3
Backtested exit thresholds
ExitHitsWin-rateEst. EV
17%13994.2%+11.2%
18%13993.5%+11.5%
19%13992.1%+11.1%
*Graduated exit: sell the cheap side at 19% / 18% / 17% (≈ a third each), holding the favorite throughout.

Status

Is it trading?

No. It is a validated analysis and backtesting framework across 139 occurrences. The live execution layer is next.

Why not automate it already?

Because the edge has to survive fees, slippage and a graduated exit before it is worth a bot. Confirming the edge first is the whole point.

What would change the answer?

A live sample where the cheap side fails to collapse as often as the backtest says. The condition q > 1 minus the exit price plus costs is the line.

StrategyVolatility straddle
StackPython · pandas · statsmodels
Validation139 backtested events
StatusPre-automation
Continue
Previous chapterWRDB