引導式測驗 · 中等 等級
AI 能預測股市嗎?測驗
透過分離預測目標、市場效率理論、樣本外測試、交易成本和詐欺警告來評估股市預測主張。
問題 1 的 8
Why define a forecast target and horizon before evaluating a stock model?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- Why define a forecast target and horizon before evaluating a stock model?
- So one score can compare different target variables without separate validation
- So the test measures one stated task, such as next-week volatility or next-day return
- So the model can be tuned on the final evaluation period
- So transaction costs can be excluded from an investable forecast
- What does market-efficiency theory provide in this guide?
- A framework for asking how information is reflected in prices
- Prices respond only after scheduled company disclosures
- Market efficiency means every investor has identical information
- An efficient price is guaranteed to equal the security’s fundamental value
- What term describes a useful predictive signal losing value as other traders exploit it?
- Look-ahead bias: using information unavailable at forecast time
- Survivorship bias: excluding failed strategies from a sample
- Data snooping: selecting a winner after many tests
- Alpha decay: competition erodes the opportunity
- What risk grows when researchers try many features and settings?
- Many trials guarantee at least one genuinely predictive feature
- The best result may fit chance in the historical sample
- Selecting a winner after many searches eliminates selection bias
- The number of tests affects computation time but not false discovery risk
- What should a point-in-time backtest ensure?
- Later-revised values can replace the values actually available then
- Signals released after the simulated decision can be treated as contemporaneous
- A timestamp is unnecessary if the sample uses daily bars
- Inputs were available by each simulated decision time
- How should an apparent directional edge be evaluated for trading use?
- Compare only directional accuracy without reporting costs
- Ignore turnover if the hit rate is high
- Include a baseline and realistic fees, spread, and slippage
- Use the training sample as the final performance estimate
- Which claim is a warning sign in an AI trading advertisement?
- A promise of high or guaranteed returns with little risk
- An identified benchmark and evaluation period
- A warning that model returns vary
- A transparent description of fees and uncertainty
- What does one good backtest establish?
- Reliable returns across all market regimes
- Performance under its data and assumptions, not guaranteed future returns
- Proof no alternative model could ever perform better
- Independent verification of the strategy