Honestly, I would say use the best you can, or best you can afford. I competed in a non-standard 2lb (~907g) competition with a single pair of Vex servo motors. Most of the competing robots were Lego NXT based, and there were a couple other Vex bots. I didn’t have the best pushing power there. My winning design won with a very low wedge, and tracked onto the other bots very quickly so that it was never being pushed from the side.
The Vex (discontinued 3 wire) motors are 100rpm, 6.5lb-in (104oz-in). Two on each side with 2.5 inch wheels would give 83.2oz of pushing force, if wheel slippage is zero, and a top speed of ~13in/sec. A proportionally equal 500g robot would have 45.9oz of pushing power, and a top speed of 7.2in/sec. I’m going to calculate off of 1.5" wheels. With two 200rpm, 45oz-in motors, you would get 60oz (1700g (force grams, not mass grams)) of pushing force, and a top speed of 15.7in/sec. Assuming good programming and a wedge bot, that should be very competitive. You wont beat the Korean or Japanese bots with it, but it should do well in most other competitions. If you wanted to raise the bar a little, you might want slightly faster motors. ExSpurt, which did very well competitively, has a top speed of 47in/sec, and pushing power of 1080g after wheel slippage
Wheel slippage is a problem. The best wheels only just barely pass the paper stickiness test. Casting your own tires is the best solution, and Brookbots has a guide on how they did it. Not the easiest or cheapest task in the world. Without good tires though, 60oz of spinning wheels doesn’t do very much. You want the weight to be as close to the maximum as possible, as more weight will give you more traction. Make the robot weigh just under 500g, and have some small attachable weight with you in competition day so that you max out on their scale (which might not be exactly the same as yours).
You should try to have at least a couple weeks between finishing the hardware of your robot and the competition day. You want to have plenty of time to make your programming perfect. If your robot fails 1 out of 10 runs, it will fail during a bout. If you can test against other robots, make sure to do so. Even test against larger bot, remote control cars, heavy weights and anything else you can think of. Programming you bot to avoid a head-to-head is a very good strategy, and a wedge or scoop will help a lot.
Just a random fact, this type of sensor does not work well when spinning quickly.