If I sliced the universe in half, would the slice go through a star?
There are about $10^{23}$ stars in the observable universe. Thanks to the expansion of the universe, those stars are currently spread over a sphere that is about $d=2.8\times 10^{10}$ parsecs across.
Of course some stars will have died whilst their light has been travelling towards us, but others will have been born, so I am going to ignore that complication.
If we imagine the stars uniformly spread through this volume$^{*}$, they have a number density of $n=3 \times 10^{-58}$ m$^{-3}$ (or $\sim 10^{-8}$ pc$^{-3}$). If we then define an average radius for a star $R$ we can ask how many stars lie within $R$ of a plane that goes through the Earth. The volume occupied by this slice is $2\pi d^2 R/4$ and the number of stars within that volume is $$N = \pi d^2 R n/2.$$
If $R \sim 1 R_{\odot}$ (many stars are much bigger, most stars are a bit smaller), then $N \sim 2\times 10^5$. So my surprising conclusion (to me anyway) is that many stars would be "sliced" by a plane going through the entire observable universe.
$*$ NB: Stars are not distributed uniformly - they are concentrated in galaxies and those galaxies are organised into groups, clusters and filamentary superstructures. However, on the largest scales the universe is rather homogeneous (see the cosmic microwave background) and so to first order the smaller-scale non-uniformity will not affect an estimate of the average total number of "sliced" stars across the observable universe, but may mean there is a larger variance in the answer than simple Poissonian statistics would suggest.
Could the clustering of stars affect the conclusion? It could if the clustering is strong enough that the median number of stars within $R$ of the plane becomes $<1$, but with the mean number unchanged. As an example consider an extreme bimodal model where all stars are found in galaxies of $N_*$ stars, where the average density is $n_*$. The "structure" of the universe could then be characterised by uniformly distributed galactic "cubes" of side $L = (N_*/n_*)^{1/3}$ and of voids with side $(n_*/n)^{1/3} L = (N_g/n)^{1/3}$. The number density of galaxies is the number of galaxies divided by the volume of the observable universe $n_g = (10^{23}/N_*)/(\pi d^3/6)$
The number of galaxies intersected by the plane will be $$ N_g \sim \left(\frac{6\times 10^{23}}{\pi d^3 N_*}\right)\left(\frac{\pi d^2}{4}\right) L = 1.5 \times 10^{23} \left(\frac{L}{N_* d}\right)$$ and in each of those galaxies there will be $\sim L^2 R n_* = R N_*/L$ intersections with a star.
If we let $n_*= 0.1$ pc$^{-3}$ (the local stellar density in our Galaxy) and $N_* =10^{11}$ (the size of our Galaxy), then $L= 10^4$ pc, $N_g = 5\times 10^{5}$ and the number of stellar intersections per galaxy will be about 0.25. thus the average number of intersections will be about the same (by design) but the variance won't be much different either.
I think the only way density contrasts could give an appreciable chance of no intersection is if $N_g<1$, and thus $L/N_* < 2 \times 10^{-13}$ - i.e. if galaxies/structures contain lots more stars and are very dense so that there is a good chance that the plane will not intersect a single "galaxy". For example if $N_* = 10^{21}$ and $n_* = 10^3$ pc$^{-3}$, then $L= 10^6$ pc and $N_g \sim 0.05$. In this circumstance (which looks nothing like our universe) there is a high chance that the plane would not intersect one of the 100 big "galaxies", but if it did there would be about $10^7$ stellar intersections.
As a rough approximation that is easy to try out with a child, you could try this:
Find or print out a big star chart or sky photo. Something like this.
Throw a long, narrow stick on top of it. See if it goes on top of any stars.
This will not be very accurate because not all stars are visible, and brightness hides the real sizes of the stars. But it should quite clearly demonstrate that the chances of a random plane hitting a star is pretty good.
--
Update: This seems to be rather popular answer. However I agree with the comments that the accuracy of this example is very poor, and it may actually be harmfully misleading. So it might be a good idea to follow it up with a discussion about its limitations, which, if nothing else, serves to illustrate the actual complexity in getting an accurate answer to the simple question.
Some points to consider:
How large a paper would you need to accurately represent the size of the stars? A map of the whole sky would have to be about 1000 kilometers in size to have most of the visible stars be 1 millimeter in diameter. More distant stars would be ever smaller in perspective projection.
How many stars are invisible? You can see about 5 000 stars with naked eye, but there are 10¹⁹ stars in the universe.
How thick is the stick? Even a hair would be wider than a distant star, so ideally you would need an infinitely thin edge for accurate results.
And then the biggest uncertainty, which was already mentioned in the question:
- How big is the universe really? Limiting oneself to the observable universe is one possibility, but that's probably not the way the question was originally phrased.
This is the Hubble Ultra-Deep Field, a long exposure photography taken by Hubble space telescope.
- It contains an estimated 10 000 galaxies. Each one of those contains an average of 100 billion stars.
- It shows a very small portion ($\frac{1}{13\ 000\ 000})$ of the whole sky. The diagonal is a tenth of a full moon's diameter.
- It was choosen because it has a low density of bright stars in the near-field. It looks completely black to the naked eye or to common telescopes.
- It looks very similar to other parts of the sky and the galaxies are very far away. The distribution would look the same anywhere else in the universe.
I couldn't manage to find a single line avoiding the galaxies. According to @RobJeffries' great answer, a line will cross an average of 100 galaxies on this picture alone, and will slice roughly 25 stars.