Controlling turnover and adaptation speed

Every rebalance that changes the holdings or their weights generates trades, and trading costs money. The volume of trades is turnover. Three things set how much a portfolio trades: how often it rebalances, how much history it looks back over, and how strongly it resists change. The resistance is set by two inertia controls.

Ranking inertia governs how strongly the current assets resist being rotated out at the next rebalance. It is the score advantage an entrant must have over an incumbent before the incumbent is replaced. At 0 there is no inertia—any higher-scoring asset displaces an incumbent; at 1 there is full lock-in—incumbents are never rotated out. Higher inertia means lower turnover at the price of slower adaptation. It is set alongside each portfolio's goals.

Where ranking inertia governs which assets are held, allocation inertia governs how much their weights move. It is the same penalty the optimiser applies for transaction costs, but without the effect on returns: a higher value makes larger weight changes less attractive. The value adds to any transaction-cost rate already configured—the optimiser sees the sum of the two. At 1%, for example, a full rebalance (200% turnover) looks like a 2% cost to the optimiser, nudging it towards smaller changes; the realised return is unaffected. Allocation inertia applies to the Markowitz optimisation methods only—Equal Weight and Hierarchical Risk Parity ignore it.

Together, rebalance frequency, the lookback window, and the two inertia controls set how quickly a portfolio adapts. A fast portfolio—frequent rebalances, short lookback, low inertia—follows the market closely but trades heavily. A slow one—infrequent rebalances, long lookback, high inertia—is cheaper and steadier but lags. Most portfolios sit between the two.