Strategy Hallucination
The model emits plausible trading logic that looks quantitative but has no market rationale, no statistical grounding, or invalid assumptions.
It creates code that passes syntax checks while smuggling in false causality.Analise Estruturada
Uma comparacao estruturada de quatro geracoes de software de trading: plataformas da era de indicadores, frameworks de estrategias, pipelines de sistemas e fluxos de trabalho LLM prompt-para-codigo.
A mudanca principal nao e apenas linguagem ou UX. Cada geracao alterou a abstracao central que os traders usavam para expressar, testar e operacionalizar ideias de estrategia.
| Dimensao | Gen 1: Indicator Era~2005-2012 | Gen 2: Strategy Era~2012-2018 | Gen 3: System Era~2018-2023 | Gen 4: LLM Era2023-present |
|---|---|---|---|---|
| Time period | ~2005-2012 | ~2012-2018 | ~2018-2023 | 2023-present |
| Representative products | MT4 / MQL4 | Backtrader, freqtrade, vnpy | QuantConnect, WorldQuant BRAIN | ChatGPT + brokerage API |
| Core abstraction | Modular indicators | Packaged strategy logic: entry, exit, sizing | Feature engineering pipeline | Natural language to code |
| Typical workflow | Drag indicators, set conditions, backtest | Write strategy code, optimize parameters, backtest | Build feature pipeline, train, validate, execute | Prompt, generate strategy, backtest |
| Main progress | Indicators moved from books and forums into reusable components | Complete strategies became packageable, shareable, and reproducible | The pipeline became the product, with standardized validation | Generation is fast and the entry barrier is low |
| Typical trap | Indicator worship and holy-grail thinking | Parameter overfitting: genetic search finds coincidence, not robustness | Data leakage and survivorship bias | Strategy hallucination and backtest overfitting blindness |
| Failure root cause | Rules are manual and lack systematic validation | The optimization target is backtest return, not robustness | Model complexity hides data-quality problems | The model skips infrastructure accumulated across the first three generations |
Os LLMs reduzem o custo de produzir codigo de estrategia, mas nao eliminam a necessidade de infraestrutura de validacao.
The model emits plausible trading logic that looks quantitative but has no market rationale, no statistical grounding, or invalid assumptions.
It creates code that passes syntax checks while smuggling in false causality.The model treats a profitable backtest as validation and misses leakage, parameter mining, unstable regimes, or survivorship bias.
It accelerates curve fitting because generation speed multiplies untested variants.spread = asset_a.close - hedge_ratio * asset_b.close
z_score = (spread - spread.mean()) / spread.std()
if z_score > 2:
short(asset_a)
long(asset_b)
elif z_score < -2:
long(asset_a)
short(asset_b)Modelo + Ferramentas
Os LLMs sao uteis quando integrados em ferramentas. Uma stack de trading seria inclui validacao de dados, motores de precificacao, logica de execucao, controlos de risco, monitorizacao e disciplina de implementacao.
Sistemas ao estilo Jane Street mostram o padrao: o modelo e uma camada dentro de uma cadeia de ferramentas maior, nao o produto inteiro.
Posicionamento StratCraft
StratCraft nao e uma reivindicacao Gen 5. Oferece a programadores de frameworks de estrategias e pipelines de sistemas desempenho de backtesting local de nivel C++, isolamento de plugins e fluxos de trabalho de validacao repetiveis.
Utilize o motor de backtest local e o ecossistema de plugins para passar de ideias geradas a sistemas validados.