PMarket Arena: A Live Laboratory for Prediction Markets

I built PMarket Arena to make prediction market mechanics visible. A market price is not just a number. It is the trace of belief, liquidity, pressure, reflexivity, and institutional capacity moving under uncertainty.

Most prediction market demos show a clean probability chart. That is useful, but it hides the parts I care about most: who is trading, why they are trading, whether the price is tracking the latent truth, how quickly arbitrage responds, and what happens when the market becomes crowded with one kind of behavior.

PMarket Arena is a browser-based simulation workbench for that problem. It lets me run a synthetic market, change the trader population, inject shocks, watch price and truth move tick by tick, and inspect the pressure behind the movement.

How to Read This Project

PMarket Arena is best read as a research program with three public surfaces. The paper asks the formal question. The arena makes the mechanism visible. The Fieldworks notes explain the connection in plain technical language.

Surface	What it does	How to trust it
Paper and research briefs	State the claim, method, result, and boundary	Look for reproducible runs, metrics, and explicit limits
PMarket Arena	Lets the reader intervene in a live synthetic market	Treat it as a mechanism lab, not as live venue evidence
Fieldworks articles	Translate the research into inspectable public reasoning	Follow the series path and compare each post to the artifact it references

The Core Equation

The arena starts from a simple claim: the visible price is an output of a population system.

p_t = F(\text{beliefs}_t, \text{liquidity}_t, \text{composition}_t, \text{shocks}_t)

That equation is not a full exchange model. It is the organizing lens. To understand the price, I need to inspect the forces producing it.

What PMarket Arena Does

The arena treats a prediction market as an ecology of trader cohorts rather than one clean probability process. Rational traders, trend followers, panic traders, arbitrageurs, and noise traders all create different kinds of pressure. The market price is what happens after those pressures pass through liquidity, price impact, inventory, spread, and shock response.

During a run, I can inspect:

price versus latent truth
volume, spread, and order-flow intensity
activity by trader cohort
cohort risk, PnL, and capital stress
liquidity stress, funding stress, and slippage
interventions, shocks, and event history
rule-based market commentary grounded in the current tick

PMarket Arena setup controls

Interventions

The most important feature is intervention. A static simulation is useful, but an interactive arena lets me ask causal questions quickly.

For example, I can add trend followers, inject a bad-news shock, lower liquidity, or add arbitrage capacity. The interface records the intervention and shows how price, volume, spread, stress, and cohort pressure respond.

PMarket Arena flow and intervention view

This is the main reason the arena exists. It turns abstract market structure questions into something inspectable: when does price track truth, when does it drift, and when does liquidity or crowding make the signal fragile?

Architecture

The public version is intentionally stateless. Live simulations run in the browser. Serverless APIs return bounded simulation results and static replay/scenario data. Browser-created scenarios and replays are downloaded by the user instead of being written to the deployment filesystem.

interactive_simulation/
  public/
    index.html
    app.js
    simulation-engine.js
    styles.css
  server.js
  vercel-handler.js
  api/
  scenarios/
  replays/
  scripts/

That boundary keeps the app deployable on normal Vercel hosting and avoids long-running server-held sessions. If persistent collaborative sessions become useful later, they can be added as a separate layer.

Evidence Boundary

The arena is not the paper evidence engine. It is a fast, visual, cohort-level approximation built to demonstrate mechanisms: composition effects, trend pressure, panic flow, liquidity stress, arbitrage correction, and shock response.

For formal claims, I use a separate Python research workflow. The browser arena is an explanation and intuition layer. That split matters: the formal experiment layer protects rigor, and the arena makes the dynamics easier to inspect.

Layer	Role	Claim Status
Python research pipeline	Multi-seed experiments, logs, metrics, figures	Formal paper evidence
Node/browser arena	Interactive scenarios and visual interventions	Mechanism demonstration
Fieldworks articles	Public explanation and research narrative	Interpretation layer

How I Use It

Start with a balanced market and fixed seed.
Run the market long enough to see normal tracking behavior.
Add one intervention, such as trend followers, arbitrage capacity, a liquidity drop, or a news shock.
Watch price, truth, volume, spread, and commentary.
Export the replay or translate the observation into a more rigorous experiment.

Where It Goes Next

The next step is to make the arena a better bridge between research, product design, and public explanation: better scenario libraries, richer replay comparison, clearer calibration badges, and more experiments beyond prediction markets.

Open the arena here: https://arena.iolucidity.org/