This paper investigates price formation and cross-market dynamics in Bitcoin markets, which are highly fragmented across multiple exchanges and instruments. Using granular limit orderbook and trade data from the 14 most liquid Bitcoin markets (including perpetual swaps, futures, and spot), the authors construct microstructural features such as orderbook imbalance and order flow imbalance at 50ms resolution. These features are used to train linear models that predict 500ms future returns, achieving R-squared values between 10% and 37% depending on the target market. The models reveal a clear leader-lagger network: markets with lower taker fees (e.g., FTX, Binance) tend to lead price discovery, while higher-fee markets (e.g., Bitmex, Bybit) lag. This asymmetry is attributed to the fee structure influencing liquidity and order aggressiveness. The authors validate their models through trading strategies. Taker strategies (using market orders) are profitable only on low-fee markets after accounting for transaction costs. More importantly, they derive a maker strategy (using passive limit orders) and test it in a live trading experiment with $1.5 million notional volume, showing significant outperformance over a naive benchmark. The study underscores the importance of fee regimes in determining market leadership and the potential for cross-market predictive signals to generate alpha, even after costs. The findings contribute to understanding price formation in fragmented markets and offer practical insights for algorithmic trading in cryptocurrencies.