I'll take a stab at this. As mentioned, you would need to find a way to keep all variables constant, EXCEPT for weight. The only way to limit all the myriad variables re: track conditions, you'd need to devise some sort of treadmill that you could run the car on, while keeping the car stationary. The rear wheels would push the treadmill belt when accelerating, while the weight of the car would rest on all four tires. Then, calculate the run-time until a specific point, such as when the car speed drops down to 5 mph.
Then, add or remove a set amount of weight; i.e. 10 grams, and re-charge the same battery, and run the test again. Record the change in run-time.
Repeat the test in 10 gram increments.
Graph the run-time results by weight and you'll begin to see a linear (most likely) correlation; this will give you a ratio - x/y.
You could then estimate run-time as a function of weight by multiplying by that ratio.
Easy-peasy...
