And the data points are not terribly reliable, such as the flight from MSP to MSP.

The problem with this sort of data is that it doesn't address how the flight ended up oversold/overbooked (equip downgrade, bumps from previous flights, etc.) which is potentially more valuable to know and which most passengers probably have no interest. There's even unique cases, like UA allowing 1Ks to book Y on oversold flights, even if it means bumping another passenger. You could not possibly forecast such a thing.