Three stocks. Same risk. Watch the weights.
| Return forecasts fed in | Stock A wt | Stock B wt | Stock C wt |
|---|
| 12 . 11 . 10 pct | 91 | 9 | 0 |
| 12 . 12 . 10 pct . one cell +1 | 50 | 50 | 0 |
| 11 . 12 . 10 pct . two cells swap | 9 | 91 | 0 |
Ten years of history pins a 20-vol stock's true return no tighter than plus minus 6.3 points. The optimizer reads 12.00 as gospel and bets 91 pct on a number that honest math writes as 12 plus minus 6.
The settling test: 14 optimization models, 7 datasets, none beat plain equal weight out of sample consistently. A 25-asset book needs ~3000 months of data, about 250 years, before the math reliably wins.