On development data our US score had a mean rank IC of +0.0085: very slightly right, on average. Yet the long-short portfolio built from it (top 10% against bottom 10%, rebalanced weekly, after costs) had a Sharpe ratio of −0.21. How can a ranking be right and the portfolio wrong?
1. IC uses everyone, the portfolio uses the tails
Rank IC measures the whole ranking, all ~100 stocks. The portfolio only owns the top and bottom 10. A score can order the middle of the pack slightly well and the extremes badly, and the portfolio only feels the extremes.
2. Averages hide size
IC treats each day equally. A portfolio's result is dominated by a few big days. Many small correct days can be wiped out by a handful of large wrong ones. In our sealed US year, three stocks (QCOM, DOW, TMO) accounted for 43% of the portfolio's result.
3. Costs
Weekly rebalancing means turnover, and every trade costs something. We charge 5 basis points per unit of turnover, and also report the result at double costs. An IC near zero rarely survives costs.
The India example goes the other way
India's sealed year had IC +0.009 (t 0.21, meaningless) but a long-short Sharpe of +0.81. Same lesson in reverse: one portfolio number from a tiny edge is mostly luck about which stocks landed in the tails.
Takeaway: report both, plus the version without the top three names. If they disagree, believe neither yet.
From Quantooo Lab, the makers of quantooo.com, a free, open and educational quant lab: every formula, every test and every failure is public. See it here: https://quantooo.com/research
Educational, not investment advice.
