Market simulation
We run LLM agents in simulated markets and auctions and measure how they bid. Every run can be replayed.
Markets and auctions we simulate
A testbed built on the designs of experimental economics. Agents receive induced values, so the efficient outcome is known before the first bid.
| Format | How it works |
|---|---|
| Continuous double auction | Buyers and sellers post bids and asks at any time. A trade happens whenever a bid meets an ask. |
| Call market | Orders are collected over an interval and cleared together at one price. |
| Posted-price market | Sellers post prices. Buyers accept or pass. |
| First-price sealed-bid auction | Bids are private. The top bid wins and pays its own amount. |
| Second-price sealed-bid auction | Bids are private. The top bid wins and pays the second-highest bid. |
| Ascending auction | The price rises until one bidder remains. |
| Descending auction | The price falls until a bidder accepts. |
| Multi-unit auction | Several identical units are sold at once, at a uniform price or at each winning bid. |
How each run is set up
Rules, values, news and prompts are set for each run and frozen with it.
Induced values
Each agent receives private values or costs, so the efficient allocation and the competitive equilibrium are known for every run.
Rules per run
Format, units, rounds and what each agent can see.
Every bid logged
Time, asset, quantity and price, together with the prompt, the news the agent had seen and the standing bids and prices at that moment.
Dynamic news feed
News enters during a run on a controlled schedule, drawn from our corpus of public communications and the price moves around them.
Event and news responseMany model families
Agents from many model families and versions bid side by side in the same run.
Repeated rounds
Runs repeat over rounds and trials, so learning and variance show up in the data.
What we measure
Every measure is computed from the run log: bids, messages, prices and outcomes.
Allocative efficiencyRealized surplusMaximum possible surplus
Allocative efficiency
How much of the available gain from trade the agents actually capture, run by run.
Price convergence
How quickly and how closely transaction prices approach the competitive equilibrium, round by round.
Cognitive biases
Overbidding, the winner's curse, anchoring on early prices, and herding after news or after other agents' bids.
Latent preferences
What an agent actually values, revealed by its bids alone and compared with the values it was given.
Collaboration, collusion and deception
Whether agents coordinate, tacitly or openly, to move prices, and whether what they tell other agents matches what they do.
Adaptation
How bidding changes across repeated rounds as agents see prices and outcomes.
Distance from human results
Human reference agents are built from published experimental data. Each superintelligence (SI) agent gets an error measure for how far its behavior sits from theirs.
SI or human
A classifier trained on bidding behavior tests whether SI agents can be told apart from human reference agents.
Collusion and pricing agents
Pricing agents can settle on prices above the competitive level without being told to coordinate.
Read the research reviewAlgorithmic collusion is pricing above the competitive level reached through pricing algorithms, even when the people who deploy them never agreed to it. In a 2020 study, Q-learning pricing agents learned to charge such prices without communicating and held them there by punishing price cuts. Later studies found the same with LLM agents. Two LLM pricing agents told only to maximize long-run profit reached prices above the competitive level, and LLM sellers in double auctions colluded more when they could message each other.
High prices alone do not show collusion, because scarce supply or strong demand can raise them too. In a simulated market the competitive equilibrium is known before the first bid, so the distance between it and the prices agents set can be measured. Pricing agents meet each other again and again in real markets, so runs repeat over rounds, and every message between agents is logged with the bids around it.
What a run shows
- Prices against equilibriumPrices compared with the competitive equilibrium, round by round, which induced values make known for every run.
- Tacit or openWhether agents coordinate with no messages between them, or by messaging each other.
- Words and actionsWhether what an agent tells other agents matches the prices it sets and the bids it makes.
- Model familiesAgents from different model families set prices side by side under the same rules, so results compare families run by run.
Every run can be replayed
Agents are measured only through what they receive and what they return.
- Inputs and outputs only
- Prompts, news and market state go in. Bids, messages and stated reasons come out.
- Pinned versions
- Each agent runs on a named model family and a pinned version.
- Frozen snapshots
- Rules, values, news and prompts are frozen per run, so a rerun sees the same conditions.
- Replayable runs
- Every run can be replayed from its log, bid by bid.
- Hypotheses first
- What we expect to find is written down before the runs start.
Record schema for a market run, one row per bid
| Field | What it holds |
|---|---|
| run | run identifier, rule set and random seed |
| round | integer |
| agent | agent identifier |
| model | model family and pinned version |
| time | simulation clock and wall clock |
| Field | What it holds |
|---|---|
| asset | asset identifier |
| side | bid | ask |
| quantity | integer |
| price | decimal |
| induced_value | value or cost assigned for this unit |
| prompt | full prompt as sent |
| news_seen | news items released before this bid |
| market_state | standing bids and asks, last prices |
| response | raw output as returned |
Every field is kept with the run, so any bid can be traced and replayed.
What you receive
Findings come with the runs behind them.
Simulation suites
Fixed sets of markets and auctions with scoring code, rerun on each new model version.
Bias classifiers
Models that detect overbidding, anchoring, herding and other biases from bid logs alone.
Simulation reports
Findings for one agent or a set of model families, with the hypotheses, the measures and the run logs behind them.
Who tests agents in simulated markets
Teams that build, deploy or study agents that bid, buy, sell or set prices.
Agent developers
The same markets and auctions, rerun on each new model version, show how bidding changed from one version to the next.
Pricing teams
Pricing agents checked for prices above the competitive level when they meet other agents round after round.
Collusion and pricing agentsRisk and model reviewers
A record of how an agent bid, round by round, that reviewers can replay and check.
Researchers
Agent behavior compared with economic theory and with human results from published experiments.
Agentic commerce
Agents that buy and sell for people and businesses.
In agentic commerce, agents search, compare offers, negotiate and pay on behalf of a person or a business. Published studies of these agents found buyers anchoring on the seller's opening price, taking the first proposal they received and some buyer agents being redirected by prompt injection from seller agents. A simulated market shows how your agent bids next to other agents, how far early prices anchor it and whether it follows other agents' bids, before it spends real money.
Common questions
Why test LLM agents in simulated markets?
Agents that bid, buy, sell or set prices can act with real money. A simulated market gives each run a known answer, because induced values fix the efficient allocation and the competitive equilibrium before the first bid. Each run can then be scored for efficiency, biases and collusion and replayed bid by bid, before the agent meets real buyers, sellers or competitors.
What are induced values?
Private values and costs that a simulation assigns to each agent for each unit it can buy or sell, a method from experimental economics. Because every value is set before the run, the efficient allocation and the competitive equilibrium are known before the first bid, and each bid can be scored against them. Allocative efficiency, for example, is the realized surplus divided by the maximum possible surplus.
Can you test our own agent?
Yes. Send us the agent you want tested, the market or auction format and the behavior you need measured. Your agent runs on a pinned version, is measured only through what it receives and what it returns, and bids next to agents from other model families. You receive a simulation report with the hypotheses, the measures and the run logs behind them.
Research
The newest articles from our Insights library, each with a full research report and its sources.
- How agent-based market simulation worksAgent-based market simulation builds a market from simulated traders and its rules, then checks the result against real data or a known answer. LLM agents now join as traders.Market simulation7 min readPDF report
- Algorithmic pricing, collusion and the law in 2026What U.S. enforcers, two appeals courts, the RealPage decrees, four state statutes and European and UK authorities say about algorithmic pricing and collusion, and what agent experiments add.Market simulation7 min readPDF report
- What to test before an agent buys on your behalfPayment protocols for agents record what a user authorized, and published research shows which agentic commerce risks remain inside those limits and how to test each.Market simulation7 min readPDF report
Notices
Investment. Rillor is not a registered investment adviser or commodity trading advisor and does not provide investment or trading advice. Research, forecasts, data and software described on this site are for research and engineering use. Nothing here is an offer or recommendation to buy or sell any security, commodity interest or digital asset. Past or simulated results do not indicate future results.
SI outputs. Superintelligence (SI) outputs can be wrong. Important decisions should include human review.