Staged sleeves from one equity forecast

Three ideas kept apart

The archive separated three portfolio ideas. Diversification meant holding many similar names so that risk unique to any one security was reduced. Strategic-asset-allocation meant a low-turnover mix across asset classes, chosen so that a decline in one class would not dominate the whole book. Tactical-asset-allocation meant changing the weight of two or more asset classes when perceived value or price trend changed.

The mechanical allocator was specified as a two-asset problem: cash versus a U.S. equity book assumed to move with a broad large-cap index.

A forecast turned into buy and sell levels

The forecast was described as an intermediate-term multiple regression of market outcomes on a large library of fundamental series. That score was then turned into paired buy and sell levels inside a strategy simulator.

A buy-and-hold reconstruction set the buy forecast-threshold so the book was invested at the start of the sample window and the sell forecast-threshold so that no exit fired, because the most negative forecast in the sample never reached it.

The baseline rule pair entered long on a positive forecast and exited on a negative forecast. A later table listed many long-only, unlevered, no-stop variants, with more demanding buy levels and more tolerant sell levels. Across that grid, completed buy-and-sell round turns were few, and wider gaps between the entry and exit levels produced fewer trades.

Staged sleeves as the general case

All-in cash-or-stock switching was treated as a special case of tactical-asset-allocation. That special case is market-timing: a jump between a fully invested equity book and fully invested cash.

The staged alternative committed a staged-sleeve at successively higher forecast-threshold levels and unwound those sleeves at successively more negative levels. Each staged-sleeve was a fixed fraction of the book, often one-quarter, with its own forecast pair rather than one all-in switch. The book became fully invested only after the last buy sleeve was filled, and fully in cash only after the last sell sleeve was withdrawn.

The archive advised against deepening the cash-out forecast-threshold. A deeper exit would have remained invested through a major crash in the sample, whereas the shallower exits would have left before that peak.

A paper replica as quality control

A paper-only pair with a milder buy forecast-threshold and a tighter sell forecast-threshold was proposed as a quality-control-overlay. It was a tight-threshold replica of the same mechanical rules, used only on paper. It completed more trades over the same long window, with holds that ranged from very short to well over a year. A significant real loss on that tight pair was to be read as evidence the equation had failed.

The archive warned that inventing more complex rules from the tabulated grids was back-fitting unlikely to travel to other periods. The staged-versus-baseline comparison assumed the forecast was run daily with trades placed the following day. Forecast level, current exposure, and that execution lag together defined the mechanical-allocation-rule, including the choice to stay in cash.