Just something I've noticed, though I don't have the test results here right now. I was testing a cross-setional momentum strategy on an asset list of the S&P 500 and looping over the csv file. Version was Zorro 1.96.

The backtest weight for A was ~10% (I think) while others were fractions. When I removed A from the list, the same thing happened to B. That got me thinking, so I tried it with the next single-letter stock and got the same result. Removing all single letter stocks, the weights then all made sense; changing these asset names and renaming the history files and putting them back in, the phenomenon disappeared.