MMath Warwick (First Class, 83%). Software Developer at Singletrack, four-plus years of maths tutoring, and a quant-finance creator on Instagram. Currently building a cross-sectional momentum strategy on crypto as the capstone for the WallStreetQuants bootcamp. Speedcubing and lifting in between.
A Python implementation combining three signals — recent price momentum, tweet-attention divergence (driven by tweet counts, not sentiment) and order-flow taker imbalance — applied cross-sectionally to a basket of eight Binance assets. Weekly rebalance with a 20bps per-side cost model, and a BTC regime filter that flips short exposure to long in bull markets.
Net Sharpe 2.25 across the backtest, against BTC buy-and-hold at 0.68 and an equal-weight basket at 0.51. Near-zero beta suggests the alpha is not crypto market exposure.
Tweet-attention divergence is the dominant alpha source (information coefficient +0.096 in 2022), with measurable decay over time — consistent with increasing market efficiency.
The BTC regime filter (short → long in bull markets) improves risk-adjusted returns across all in-sample years.
Turnover settles at 0.174 at 20bps/side; net Sharpe stays within 0.6 of gross.
A 2D parameter sweep across price and tweet lookback windows shows the result is a robust neighbourhood, not a lucky local optimum.
A few things I spend time on.