Testing Technical Analysis, Kelly Sizing, and Price Predictability
Summary
The author recounts using Kelly-style position sizing, mathematical optimization, historical backtests, and a neural network to investigate trading systems. The discussion first explains why a positive win rate alone does not prevent ruin: bet size and losing streaks matter. It then presents Kelly sizing as a way to allocate capital when the odds and win probability are known, while arguing that apparent strategy advantages still depend on assumptions about whether historical indicators predict future prices.
The author reports that an optimized system using many technical indicators performed well in backtests but later experienced a drawdown that its historical record had not suggested. A neural network analysis of past prices and future prices then led the author to question whether price history had predictive value. These are personal claims, not a reproducible study: the document gives no dataset, model specification, validation procedure, or detailed performance results. Its useful lesson is to make trading hypotheses falsifiable and treat backtest success as evidence requiring further scrutiny.
Key ideas
- A favorable win rate does not prevent ruin when position sizes are too large for the available capital.
- Kelly sizing links bet size to estimated odds and win probability, so its usefulness depends on reliable estimates.
- Historical backtests can look strong while concealing assumptions that fail in live trading.
- Trading ideas should make predictions that can be tested and potentially disproved.
- The author’s neural network conclusion about price history is not supported by enough methodological detail to generalize.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.