2 added 399 characters in body
source | link

Python 2, 369369 368 bytes

print"It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash[id(id0)%20]/7%20]

Python 3, 371 bytes

print("It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20])

TheI was previously using the hash builtin to index (hash(id)%20), which returns a random value per-start of the Python interpreter ever since https://bugs.python.org/issue13703. It's not random for the empty-string though (always 0), so need to use something else, the id builtin!

On second look, I could use id directly, but it seems to always produce even numbers. IIRC, id(object) in CPython just returns the memory location of object, so this makes sense. Maybe if I used Jython or IronPython, I could skip the divide-by-7. Anyways, hash(id) vs id(0)//7 is equal in Python 3, but can use the / operator for truncating integer division in Python 2, saving a byte.

Python 2, 369 bytes

print"It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20]

Python 3, 371 bytes

print("It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20])

The hash builtin returns a random value per-start of the Python interpreter ever since https://bugs.python.org/issue13703. It's not random for the empty-string though (always 0), so need to use something else, the id builtin!

Python 2, 369 368 bytes

print"It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[id(0)/7%20]

Python 3, 371 bytes

print("It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20])

I was previously using the hash builtin to index (hash(id)%20), which returns a random value per-start of the Python interpreter ever since https://bugs.python.org/issue13703. It's not random for the empty-string though (always 0), so need to use something else, the id builtin!

On second look, I could use id directly, but it seems to always produce even numbers. IIRC, id(object) in CPython just returns the memory location of object, so this makes sense. Maybe if I used Jython or IronPython, I could skip the divide-by-7. Anyways, hash(id) vs id(0)//7 is equal in Python 3, but can use the / operator for truncating integer division in Python 2, saving a byte.

1
source | link

Python 2, 369 bytes

print"It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20]

Python 3, 371 bytes

print("It is certain.It is decidedly so.Without a doubt.Yes definitely.You may rely on it.As I see it, yes.Most likely.Outlook good.Yep.Signs point to yes.Reply hazy try again.Ask again later.Better not tell you now.Cannot predict now.Concentrate and ask again.Don't count on it.My reply is no.My sources say no.Outlook not so good.Very doubtful".split('.')[hash(id)%20])

The hash builtin returns a random value per-start of the Python interpreter ever since https://bugs.python.org/issue13703. It's not random for the empty-string though (always 0), so need to use something else, the id builtin!