So, I actually think dlss is making a very valuable point here: the argument that many people are making with relation to hiding point totals, is that if you can see the number of points a post has, it may affect whether you decide to vote at all: you may go "oh, that's already high enough, I won't bother voting".
In essence, people have a mental concept of how high or low they feel posts they are looking at are, and adjust, possibly very subtly, possibly without even realizing it in a way that they would explain if asked, their behavior to cause the post to hit that target, as opposed to showing their true interest with an up/down.
(For a reference on users discussing this effect, and some commentary from PG about having seen this effect on the high end of comment karma scores, see this example thread from another post: https://hackernews.hn/item?id=2465271 .)
How peoples' behaviors change is complex. On the one hand, you may see "your side" heavily lower than "the opposition", and decide "ugh, it's pointless". On the other, you may go "ok, that's just not fair", or "help out the team", and make extra certain to vote on that specific poll.
Due to these attitudes, which are pretty much the same things that are driving the changes in behavior with respect to comment voting, you don't get accurate information with public live polls. This really is the same effect, and should be thought of and treated the same.
Now, making the entire poll /permanently/ only visible to the poster would mean, as you say, that the poll becomes meaningless. However, I would say that's a strawman: there are numerous other ways to handle the situation that still make sense.
Example: polls could be private until they "time out", at which point the poll is no longer votable on, and the data becomes public. Frankly, this seems correct, and I'm glad dlss brought it up (although apparently people disagree with the way he worded it or something, which seems silly).
(I also think that this behavior would dovetail very well with "show the comment scores after people can no longer vote on them", which came up elsewhere in the comments for this post: https://hackernews.hn/item?id=3122362 .)
It's an analogy. Without publicly viewable karma, posts lose a lot of their information content.
Polls without results lose more, but some would argue[1] that there is no royal road to understanding what others think, and that the best way to use polls is to simply read the possible responses and ignore the results.
For example, [1] believes that if you saw a poll on "which database do you use?", you should ignore the poll results and researching each option on your own. The person he is replying to is advocating researching databases in HN popularity order.
In other words, the only content in a poll is the numbers.