• Built several strategies including trend-following (MA crossover, time-series momentum), mean reversion (RSI), and breakout (Donchian channels) plus signal filters and hard/soft voting ensembles

• Wrote a vectorised backtesting engine over a real cash ledger, applying per-trade fees and basis-point slippage, and enforces a one-bar execution delay to eliminate lookahead bias

• Validated out-of-sample via walk-forward analysis, with parameter sweeps favouring robust plateaus over fragile single-point optima. Achieved an out-of-sample Sharpe of 1.19 against a buy-and-hold benchmark of 0.79

• Added Monte Carlo block-bootstrap resampling to produce distributions of final return and max drawdown rather than the single historical draw, plus a noise-perturbation check measuring how much a strategy's edge depends on the exact price path