Paper: arXiv 2309.10729

Authors: Masanori Hirano, Ryosuke Takata, Kiyoshi Izumi

Abstract

This paper presents a new artificial market simulation platform, PAMS: Platform for Artificial Market Simulations. PAMS is developed as a Python-based simulator that is easily integrated with deep learning and enabling various simulation that requires easy users’ modification. In this paper, we demonstrate PAMS effectiveness through a study using agents predicting future prices by deep learning.

Complexity vs Empirical Score

  • Math Complexity: 2.0/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Philosophers — conceptual discussion, lower math and empirical scores

Why this score: The paper presents a simulation platform (PAMS) with minimal advanced mathematical derivations, focusing instead on software architecture and integration with deep learning. Empirical rigor is low as the work centers on platform introduction and a demonstration study rather than extensive backtesting with real-world data or statistical performance metrics.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop & validate PAMS platform"] --> B["Methodology: Construct Python-based simulator<br>with deep learning integration"]
  B --> C["Input: Historical market data &<br>agent behavioral parameters"]
  C --> D{"Computational Process: Artificial Market Simulation"}
  D --> E["Agent training: Predict future prices<br>using deep learning models"]
  E --> F["Simulation runs &<br>market dynamics analysis"]
  F --> G["Key Findings: PAMS enables effective<br>simulations with user modifications"]
  G --> H["Outcome: Validated platform for<br>financial market research"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes PAMS, a new Python-based artificial market simulation platform that integrates easily with deep learning and allows user-customizable simulations.

Source sentence
This paper presents a new artificial market simulation platform, PAMS: Platform for Artificial Market Simulations.

Assumes 5

  • The total number of agents is fixed at 100 (nsa + nda = 100). model
    Source sentence
    The total number of agents constant: nsa + nda = 100.
  • Deep Agents use an LSTM to predict the up/down direction 100 ticks ahead based on the past 100 price series. model
    Source sentence
    2) Deep Agent: The deep learning-based agent predicts the up/down of the price 100 ticks in the future when its turn to act comes, using an LSTM-based neural network based on the past 100 price time series (the price and mid-price series).
  • A Deep Agent's model is only used for trading if its evaluation accuracy exceeds 51%. model
    Source sentence
    Only when the prediction accuracy in the evaluation data exceeds 51%, it is assumed that a good model has been created and is actually used for trading.
  • The market mechanism is a continuous double auction. market
    Source sentence
    The simulation consists of a continuous double auction market, nsa Stylized Agents, and nda Deep Agents.
Show 1 more
  • Simulation data are generated synthetically within the PAMS platform. data
    Source sentence
    n": [0.0]}, 38 "noiseWeight": {"expon": [1.0]}, 39 "meanReversionTime":{"uniform" :[50,100]}, 40 "noiseScale": 0.001, 41 "timeWindowSize": [100, 200], 42 "orderMargin": [0.0, 0.1] 43 } 44 } 45 46 saver = MarketStepSaver() 47 48 runner = SequentialRunner( 49 settings=config, 50 prng=random.Random(42), 51 logger=saver, 52 ) 53 runner.main() Fig. 2.

Data

  • Synthetic simulation data generated by PAMS synthetic unknown
    Source sentence
    n": [0.0]}, 38 "noiseWeight": {"expon": [1.0]}, 39 "meanReversionTime":{"uniform" :[50,100]}, 40 "noiseScale": 0.001, 41 "timeWindowSize": [100, 200], 42 "orderMargin": [0.0, 0.1] 43 } 44 } 45 46 saver = MarketStepSaver() 47 48 runner = SequentialRunner( 49 settings=config, 50 prng=random.Random(42), 51 logger=saver, 52 ) 53 runner.main() Fig. 2.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • Deep Agents can make a profit in the simulated market. performance
    Source sentence
    E. Discussion First of all, one thing that can be observed from the experimental results is that Deep Agents can make a profit.
  • As the number of Deep Agents increases, the average profits of both Stylized Agents and Deep Agents decrease. performance
    Source sentence
    In addition, the experimental results show that as the number of Deep Agents increases, the average profits of both Stylized Agents and Deep Agents decrease.
  • The emergence of advanced strategies like Deep Agents reduces profits of existing strategies and cancels out profits between advanced agents. theoretical
    Source sentence
    According to these results and discussions, the emergence of agents with advanced strategies like Deep Agents not only reduces the profits of existing trading strategies but also cancels out the profits between agents with advanced strategies.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 3 papers in the archive

Code

https://github.com/masanorihirano/pams — the authors' own repository

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.