Stock Agent Taxonomy
AI stock trading bots differ most profoundly in the kind of intelligence they operationalize. Some discover catalysts, some classify regimes, and others construct portfolios from competing company theses.
This taxonomy ranks eight agent styles according to their fit with autonomous stock selection. BotTrade provides stock-market scenarios where each style produces sequential decisions, simulated trades, and risk evidence.
A BotTrade agent receives typed stock-market observations, analyzes visible historical bars, submits simulated orders, advances through the scenario, and receives trade and risk evidence. It is the practical starting style because any Python policy, MCP client, or framework can implement it. Builders can add catalyst, ranking, or committee behavior without replacing the market-facing execution and evaluation loop.
This agent builds positions around earnings, policy changes, product events, or other developments capable of changing expectations. It must preserve source times, identify affected symbols, and define a horizon and invalidation rule. FinGPT supports financial-language research. BotTrade then shows whether the catalyst thesis produced timely orders, useful holding periods, and acceptable drawdown in a historical stock scenario.
The ranking agent scans an investable universe, compares relative evidence, and allocates only to a bounded set of candidates. It should publish its features, ranking rule, and post-trade concentration. Qlib offers quantitative ranking and research components, while BotTrade scan and inspection tools provide an agent-facing path from broad opportunity discovery to simulated portfolio decisions.
This style adjusts sector, factor, concentration, and cash exposure when visible evidence indicates a change in momentum, defense, dispersion, or volatility. The classification must have explicit triggers and invalidation conditions. BotTrade stock scenarios reveal the timing of allocation changes, whether risk fell before drawdown expanded, and which trades expressed the new portfolio doctrine.
A quality agent prioritizes durable economics, balance-sheet strength, competitive position, earnings quality, and valuation rather than frequent price signals. Every fact should include a dated primary source, and the thesis needs a catalyst or review horizon. The agent can submit long-horizon simulated positions to BotTrade, where concentration and opportunity cost remain visible beside return.
Price and volume define the initial momentum signal, while language research checks whether a credible catalyst and broad market confirmation support persistence. The narrative cannot override weakening price evidence. vectorbt can research deterministic momentum variants. BotTrade adds the sequential language-agent loop, allowing builders to inspect entries, exits, inactivity, return, and drawdown from the combined policy.
This agent searches for sharp moves that appear inconsistent with available fundamental or event evidence. It must distinguish temporary liquidity stress from genuine new information and state a maximum loss and reversion horizon. BotTrade trade history reveals early entries, repeated averaging, and exit discipline, which are more informative than whether the final price eventually moved back.
Specialists cover company research, macro context, technical evidence, criticism, risk, and portfolio construction before one manager submits orders. TradingAgents provides a concrete architecture reference. Keep role outputs typed and final authority explicit. Connect only the manager to BotTrade so every simulated stock trade has one accountable decision while the supporting debate remains available for attribution.
The most useful evaluation connects an agent's research doctrine to the positions it actually takes. BotTrade makes those stylistic differences visible through complete portfolio paths rather than abstract descriptions.