Criteria, we give trophies to the top 3 and send top 2 to council races.As an alternative to Darin's suggestions on ranking, which depended on absence of lane bias, consider using "Who occupied rank A (maybe ranks A and B) during the last N rounds" to select finalists. If N > 1, then A-C oscillations are compensated for. This can probably reduce the effects of a single "bad run" late in the racing.
We raced 6 rounds, for each group. For our larger ranks (16+), we race 4 lanes, and we raced 3 lanes for the groups smaller than 15.
For the 4 lane configuration, from A group we take the top 3 finishers, and the winners of B&C. Note that the winners of B&C would have moved to the next round anyways. The 3rd place car of group A would have dropped down to B, but gets to stay in the trophy round.
For the 3 lane configuration, the 3 cars from Group A, made into the final 4, and then the winner of Group B (who would have moved up to the A group) make it into the final 4. I might add the top 2 from B next year, because I dislike racing 4 and having only one boy not get a trophy.
I was satisfied that this worked well. There is a percieved fairness if when we announce that the top 3 of A go or the winner B goes to the finals round, if you come in 4th then you don't get a trophy.