I think your code is fine - it gets the point across. Why use a list comprehension (and then convert that back to a string)?
hidden_word = ''.join([c if c in guesses else '-' for c in word])
Is that really better? You could change it to a generator expression, but still...
hidden_word = ''.join(c if c in guesses else '-' for c in word)
EDIT: Testing this with a 1000 character "word" using:
import timeit
setup = """import random
chars = "abcdefghijklmnopqrstuvwxyz"
s = "".join(random.choice(chars) for _ in range(1000))
guesses = "agjoxwz"
"""
t1 = "hidden_word = ''.join([c if c in guesses else '-' for c in s])"
t2 = "hidden_word = ''.join(c if c in guesses else '-' for c in s)"
t3 = """hidden_word = ""
for c in s:
hidden_word += c if c in guesses else '-'"""
Results:
In [24]: timeit.timeit(setup=setup, stmt=t1)
Out[24]: 100.88796829901968
In [25]: timeit.timeit(setup=setup, stmt=t2)
Out[25]: 147.86355467070305
In [26]: timeit.timeit(setup=setup, stmt=t3)
Out[26]: 247.9441536138757
Whoa. So the list comprehension is in fact must faster (and better than a generator expression).
With only 50 letters per "word", the differences are less pronounced, but the list comprehension still wins:
In [28]: timeit.timeit(setup=setup, stmt=t1)
Out[28]: 5.416419290962722
In [29]: timeit.timeit(setup=setup, stmt=t2)
Out[29]: 7.828715333297168
In [30]: timeit.timeit(setup=setup, stmt=t3)
Out[30]: 7.984714775332918
hidden_word = ''.join(c if c in guesses else '-' for c in word).