Let me step this back a bit …
OK. It’s a little more complicated than that, but we can work thru the simple example.
So 2000 steps is very improbable, if the steps are independent, even if each step is 99% certain.
Get a 100-sided die or percentile dice, each roll is independent. a roll of “1” is a failure, “2-100” is a success. Start rolling …
By the time you have rolled 2000 times, the probability you have rolled 2000 successes and zero failure is indeed very small, ~210^-9. On average after 2000 rolls you should have about ~1980 successes and ~20 failures. Roll 22 more times and you have about 50% chance of getting the 2000th success. Roll 22 more, for a total of 2044 rolls, and it’s nearly certain you will have at least 2000 successes. *
Your are calculating the probability of exactly 2000 successes in 2000 trials with p=0.99. Equivalently you could be calculating the probability of zero failure in a single roll of 2000 dice. Your math is correct but it means nothing; no one thinks evolution works this way.
There is some related math for calculating in the thread below. The topic comes up fairly often. 
** Operating in parallel, rolling more than one die in each trial with multiple trials, will obviously speed things up even more, requiring fewer than 2000 trial for 2000 successes. Exactly how much faster requires requires some sort of simulation of sexual mixing, which seems beyond the scope of current discussion.