Comment by latchkey
PT's concept of velocity isn't just tracking a number. It is tracking what work will get done within a period of time based on the average amount of work done previously. It allowed PM's to actually estimate when a feature would be completed and how adding or removing stories (or people!) would impact future feature deadlines. Initial velocity was meaningless and only developed over time as an average value of a team of people working together, taking into account things like vacations and sick days. None of this could be done realtime with a rack of index cards as it was something that happened over many weeks.
That concept of velocity is in the original Extreme Programming practice, which predates Pivotal, and it was indeed done with racks of index cards.
It's explained pretty clearly in the first edition of Extreme Programming Explained, but i can't find a copy of that online right now. The second edition was absolutely ruined for some reason, but still contains a rough description of it:
> Whichever units you use, hours or points, you will need to deal with the situation where actual results don't match the plan. [...] If you are estimating in points, modify the budget for subsequent cycles. A simple way to do this, dubbed "yesterday's weather" by Martin Fowler, is to plan in any given week for exactly as much work as you actually accomplished in the previous week.
Tracker uses some kind of rolling average rather than just last week's number, but it's the same idea.