A prediction market is a simulation environment that happens to use real money as its scoring function. That is the core insight of this article, and it is the reason the S8 series treats Polymarket not as a trading venue but as a laboratory. Every market on the Gamma API is a resolved or resolvable experiment: a question was posed, participants committed capital to opposing answers, and the market settled at a price that reflected collective probability. The price is the simulation’s output. The edge is the gap between that output and what you believe the true probability should be.
The market as simulation engine
When you scan Polymarket through the edge-detection CLI, you are not looking at a stock exchange. You are looking at a distributed simulation engine where thousands of participants have already done the work of estimating probabilities. The market price for a YES contract is a probability estimate — the crowd’s guess that the event resolves YES. If the market price is 0.62 and your model says the true probability is 0.73, the edge is 11 percentage points. That edge is the same quantity the simulation loop in S8.1 measures: the gap between your estimate and the market’s estimate, expressed as a tradable signal.
The difference between a prediction market and a traditional simulation is that the prediction market settles against reality. A Monte Carlo simulation generates synthetic outcomes. A prediction market generates a price that converges toward the actual outcome as resolution approaches. This convergence is what makes it useful as a simulation environment — you can test whether your edge-detection model consistently identifies mispriced contracts before you commit real capital.
The three formulas
The edge-detection CLI implements three formulas that convert market data into trading decisions. Understanding these formulas is understanding the simulation’s logic.
The edge formula is simple: edge = p_true - market_price. Your model estimates p_true using spot momentum data from Coinbase and Binance — if Bitcoin is trending up on spot exchanges, the probability that a crypto prediction market resolves YES increases. The market price is what Polymarket is currently charging for that contract. The edge is the discrepancy. The threshold is 7% — below that, noise dominates signal.
The expected value formula adds the payoff structure: EV = (p × profit) - (q × loss), where q = 1 - p. A positive EV means the trade is profitable in expectation, even accounting for the possibility of losing. The EV formula prevents you from taking edges that look large but carry disproportionate downside.
The Kelly Criterion sizes the position: f* = (p × b - q) / b, where b is the odds received. Full Kelly is too aggressive for prediction markets — the default is quarter Kelly (25% of the optimal fraction), because the model’s probability estimates are uncertain and overconfidence in sizing destroys bankrolls faster than bad signals do.
Together, these three formulas form a simulation pipeline: detect the edge, evaluate whether it is profitable in expectation, size the position conservatively. The simulation loop in S8.1 replays this pipeline against resolved markets to test whether the formulas actually work. The edge-detection CLI runs the same pipeline in real time against live markets.
What simulation reveals about the market
Running the edge scanner against Polymarket data produces a specific kind of knowledge: you learn where the market is inefficient. Crypto markets with high trading volume tend to be efficiently priced — edges are small and transient. Political markets with low liquidity can have large, persistent edges because fewer participants are doing the same analysis. Sports markets fall somewhere in between, with edges appearing around lineup announcements and injury reports that the crowd has not yet incorporated.
This is simulation in its purest form. You are not predicting whether Bitcoin will go up. You are measuring the efficiency of a market’s probability estimation. The market itself is the simulation, and your edge-detection model is the probe that measures its accuracy. When the probe detects a large edge, it means the market’s simulation has diverged from reality — the crowd’s probability estimate is wrong, and you can profit from the correction.
The KOT scoring system grades these edges on a scale from SPECULATIVE to PROVEN_EDGE. A SPECULATIVE edge has appeared in a single scan with no confirmation. A PROVEN_EDGE has survived multiple scans, maintained its size above the 7% threshold, and shown positive expected value across different timeframes. The scoring is the simulation’s quality gate — it prevents you from acting on noise.
The dashboard feedback loop
The edge scanner does not run in isolation. The Hermes dashboard — the observability layer running on port 9119 — tracks which scans were run, which edges were detected, and which trades would have been placed. The task tracker records the verdict: PROVEN_EDGE, SPECULATIVE, or NO_EDGE. Over time, the dashboard builds a history of the simulation’s performance.
This feedback loop is critical because prediction markets are non-stationary. The set of available markets changes daily. Liquidity shifts. New participants enter with better models. An edge that was profitable last week may be arbitraged away this week. The dashboard shows you this degradation in real time — if the number of PROVEN_EDGE detections drops, the market is becoming more efficient and your model needs updating.
The simulation loop (S8.1) handles this by replaying the edge detection against resolved markets on a rolling basis. If the model’s edge predictions no longer correlate with actual outcomes, the simulation flags it. The promotion gate (S8.3) then decides whether the model should be adjusted or retired. This is the Simulation → Live pipeline working as designed: simulate first, measure continuously, promote only when the evidence is strong.
Why this matters for the S8 series
Prediction markets are the ideal simulation environment for the Kingdom of Truth’s architecture because they satisfy every constraint: they are real (prices reflect actual capital commitments), they are measurable (outcomes are binary and resolved on a schedule), and they are safe to simulate (paper-trading costs nothing). The edge-detection CLI is the tool that converts this environment into actionable knowledge. The KOT scoring system is the quality gate. The dashboard is the observability layer. The task tracker is the memory.
S8.9 completes the picture that S8.1 started: the simulation loop is not just a backtesting engine — it is a live probe that measures market efficiency in real time. The prediction market is the simulation environment. The edge scanner is the instrument. The score is the verdict. And the promotion gate decides when the simulation has produced enough evidence to justify real execution.
The market is the simulation. Your edge is the measurement. The score is the gate. That is the architecture, and it works because prediction markets were designed to do exactly what the S8 series needs: convert collective belief into a price, settle against reality, and let you test your models against both.
Grounded in wiki concepts `simulation-mode`, `edge-detection-cli`, `dashboard`, `task-tracker`, entity `polymarket`, and source `simulation-mode-20260803`. Skills: `edge-detection-cli`, `kanban-orchestrator`. Ninth article in the S8 series on Simulation → Live. Design notes on a running system, not a sales pitch.
