If you only promote the standard formula winners, you risk failing to find the losers that can be eventually better when given the chance. In other words, if t

If you only promote the standard formula winners, you risk failing to find the losers that can be eventually better when given the chance. In other words, if the formula isn't perfect, it isn't good to apply it strictly. Some randomization of the low end threshold / lottery for the losers seems to be a good idea.