Draw a line from a subshell to the one to the bottom left of it. If you're at the edge of the table, move one space to the right. Hard to explain with words, so 1s 2s 2p 3s 3p 3d 4s 4p 4d 4f 5s 5p 5d 5f 5g So bottom left of 1s is nothing, next then is 2s. Bottom left of 2s is nothing, so next is 2p. Bottom left is 3s, then nothing, so 3p. Bottom Left of 3p is 4s, (a break in the pattern), so it's 4s. Then go to the next of 3p, 3d. Bottom right of 3d is 4p, another break. then after 4p is 5s, yet another break in the pattern. Then 4d, 5p, etc. Not only is there this muddle, but you'll find that, in the transition metals, its common to have a more stable arrangement with a full s subshell and a partially full p subshell. For example, Vanadium goes 3d3, 4s2, while Chromium goes 3d5, 4s1! The d sub shell took an electron from the s!

Jeztyr explained a mnemonic for when the pattern doesn't work, but if you want to know why the pattern doesn't always work, it comes down to solving the schroedinger equation: [-hbar^2/2m del^2 + V(r)] phi = E phi The bits mean: hbar = Planck's constant. V(r) = spherically symmetrical potential E = Energy. phi = probability "amplitude" for an electron in terms of a point in space. When the equation is solved, you get certain allowed energy levels, which correspond to different solutions for phi. There are components for the distance of the electron from the center (the shells) and components for the spherical harmonics: the various ways in which the electron can be oriented wrt the nucleus in space. But the energy contributions from the distance part and the orientation part overlap, which sometimes breaks the neat pattern - sometimes it's less energy to go to the next shell, instead of the next "letter". The math to get the solutions is REALLY hard - I had to do it once - but the solution is relatively simple. This is a good web page to explain the math: http://spiff.rit.edu/classes/phys315/lectures/lect_5/lect_5.html