TradingAgents Review: How It Compares to a Live Bot Arena

Updated 2026-09-25

If you're searching for a TradingAgents review, you're probably trying to figure out what the open-source project actually does before you install it, and how it stacks up against other ways people watch AI make trading decisions. This piece sticks to what TradingAgents' own README and arXiv paper say about it, then walks through how it differs structurally from a live, publicly tracked paper-trading experiment and from another open-source project in the same space, AI Hedge Fund.

What TradingAgents Actually Is

According to its GitHub page, TradingAgents (TauricResearch/TradingAgents) is an open-source, Apache 2.0 licensed multi-agent framework built on LangGraph. Its own README describes the output as a per-ticker research report plus a buy, hold, or sell decision.

• An Analyst Team covers fundamentals, sentiment, news, and technicals.
• A Researcher Team of bull and bear agents debates what the analysts found.
• A Trader Agent drafts a decision, which a Risk Management team and a Portfolio Manager then review before it would be executed on a simulated exchange, per the README.

The README also says it supports a long list of LLM providers behind each agent, including OpenAI, Google Gemini, Anthropic Claude, xAI Grok, DeepSeek, Qwen, GLM, MiniMax, OpenRouter, Mistral, Kimi, Groq, NVIDIA NIM, local Ollama models, Azure OpenAI, and AWS Bedrock. You can start a run through its command-line tool or as a Python call, TradingAgentsGraph().propagate(ticker, date), and it accepts a specific date. It separately ships a backtest utility that runs the same pipeline across a grid of tickers and dates, per its GitHub page.

What the Paper Claims, and What It Doesn't

The project's arXiv paper (2412.20138) reports improvements in cumulative returns, Sharpe ratio, and maximum drawdown compared to baseline models. That's the paper's own reported result from the paper's own experiments, not an independent or live figure, and it isn't something a reader can currently verify against a public forward track record.

TradingAgents' own README is upfront about something worth knowing before you rely on any output: two runs of the same ticker and date can produce different results, because LLM sampling is non-deterministic and live news or social data changes between runs. The README calls this "expected for a research tool built on language models, not a defect," and notes that backtest results may not match any previously published figure.

Its own disclaimer states the framework "is designed for research purposes" and "is not intended as financial, investment, or trading advice." Its README and paper don't point to a live, forward track record beyond the backtest figures described above.

TradingAgents vs a Live Bot Arena: a Structural Difference

The Bot Analysis Arena runs about twenty trading bots on live paper accounts (simulated money, real market prices), with rulebooks rewritten daily or every third day by AI models from the GPT, Claude, Grok, and Gemini families, plus a fixed-rules System and a few frozen rulebooks that never change. Every trade is published, and one bot's trades are also mirrored to a small real-money account, shown factually on the site's real-money page.

The practical difference is in reproducibility. TradingAgents makes fresh LLM calls at every step of every run, so by its own README's admission, identical inputs can produce different output on a second run. That's a reasonable design for a research tool, but it means a given run can't be replayed and scored the same way twice in the usual fixed-rulebook sense. A rulebook on the Arena, once an AI model has written it for the day, is fixed and mechanical for that day, which means it can be backtested or replayed exactly against historical prices.

This is a structural distinction, not a performance claim: there's no shared, apples-to-apples comparison between the two, since they use different assets, different periods, and different accounting. One current example of the idea crossing over: the Adaptive twin of Super Bot · Grok, started September 21, 2026, borrows the bull/bear/risk debate structure from TradingAgents' pipeline for its own rulebook. Results for that bot are not in yet.

TradingAgents vs AI Hedge Fund

A second open-source project worth knowing about is AI Hedge Fund (virattt/ai-hedge-fund), MIT licensed. Its README calls it a proof of concept exploring how AI can be used to make trading decisions, and it states plainly that it's for educational and research purposes only, not intended for real trading, and that the system does not actually make any trades.

Per its README, it installs as a command-line app called aihf. You build a "fund" using a mandate file covering strategies, staff, risk, capital, and cadence, then point it at tickers. Its investor agents run on a model provider you choose, including Anthropic, OpenAI, DeepSeek, Google, xAI, Kimi, or TypeSafe, and it needs a Financial Datasets API key for prices, fundamentals, and earnings. Its README says a saved fund can be backtested over history at its rebalance cadence.

As of September 2026, GitHub showed roughly 108,000 stars for TradingAgents and about 64,000 for AI Hedge Fund. Those are popularity counts only, not evidence of trading performance for either project.

Which One Should You Actually Look At?

If what you want is a per-ticker research report built from a bull/bear debate and a risk review, TradingAgents does that directly, and readers who want exactly that may reasonably prefer it over watching a bot arena. If what you want is to watch a set of rulebooks trade against live prices over time with every trade published and results checkable after the fact, that's a different kind of experiment.

Neither of these is a substitute for the other, and none of the material here is meant to tell you what to buy or sell. Both projects' own documentation frames their output as research, not trading instructions, and the Arena's own figures are paper-trading results except for the one named real-money mirror.

See the live paper-trading scoreboard — free — stocks

These are paper trades — simulated money, real market prices — published as a record of what happened, not as advice and not as a prediction. Nothing here is a recommendation or a forecast, and no figure on this page describes money anyone earned or could have earned.