Netting Positions Across Strategies

Run more than one strategy on the same book and eventually two of them will disagree about the same stock at the same moment — a momentum strategy wants to buy 10,000 shares of a name while a mean-reversion strategy wants to sell 6,000 shares of the very same name. Sent to the market independently, that's two separate orders, two sets of commissions, two bid-ask spread crossings, and a market-maker on the other side who is perfectly happy to sell you shares and buy them back from you a moment later at your expense. Netting means combining those signals internally first — in this example, down to a single net order to buy 4,000 shares — before anything reaches an exchange.

The benefit is straightforward: fewer shares traded means less spread paid and less market impact, and the two strategies effectively execute against each other for free instead of both paying a spread to a market-maker. The complication is that netting blurs each strategy's individual attribution — if the combined order fills at a certain price, dividing that fill (and its cost) back out to the two strategies that generated it requires a convention, and getting that convention wrong can make one strategy look better or worse than it actually performed on a standalone basis.

Firms typically net at the portfolio level on a schedule — before each rebalance, or continuously in a live order-management system — and keep a separate accounting layer that still tracks each strategy's notional P&L as if it had traded independently, so that the operational benefit of netting doesn't corrupt the research question of which strategy is actually working.