A good signal is not yet a portfolio

A ranking leaves important decisions open

A signal can rank securities without deciding how much to hold, when to rebalance or what to do when observations are missing. A portfolio process must define an eligible universe, weighting rule, concentration limits, cash treatment and implementation timing. Different choices can produce very different risks from the same underlying signal.

Translate turnover into a consistent cost estimate

Define turnover before using it. Suppose absolute traded notional over a period totals 40% of portfolio value, counting both buys and sells. At an assumed cost of 0.25% of traded notional, the estimated cost is 0.40 × 0.25% = 0.10% of portfolio value. A different turnover convention needs a corresponding adjustment; the word alone is not a complete calculation.

Test the whole process

A research test should include the data available at the time, realistic delays, costs and the actual selection and weighting rules. Separate development observations from later evaluation observations. Inspect drawdowns and parameter sensitivity as well as average returns. Trying many variants and keeping only the strongest result can exaggerate the evidence for a repeatable effect.

Review risk against the mandate

Compare exposures and results with an appropriate benchmark and the original objective. Inspect whether apparent diversification is several positions exposed to the same driver, whether liquidity worsens under stress and whether losses can exceed a planned stop during gaps. The final research pack should explain the method, its failures and the evidence required before changing or suspending it.

A learning example, not a market recommendation. Work through the assumptions before applying any model.