I've talked a lot about probability and how important it is when building a list and making decisions in game. However, potential is something that is often more difficult to interpret, because it relies on you going against the expected value.
Potential has less of a significance in list building but there are certain cases in which it can be utilised. Generally units that rely on potential are referred to as fear factor units, and it is the basis of deathstars. Deathstars in itself don't go against the expected value, they have the ability to dish out a lot of damage and soak up some in return. However, if you consider the entire game as one big probability tree, relying on single units can start to turn slightly risky. That's why nobody advocates pure deathstar lists, you always need supporting units. This leads to an interesting playstyle. We can take the American ETC 2012 Ork list as an example (I believe the guy also won the ESC). It's a green tide with a big squad of nob bikers and warbosses. He forces his opponent to deal with the deathstar (and pretty fast) whilst the 120 or so boys push up and are still a problem to your opponent. Your opponent has to destroy that unit because it has the potential to rip his entire army apart.
This can also be utilised on a smaller scale with units such as leman russ executioners or shokk attack guns. They aren't so crazy, but still something your opponent has to play against (e.g. spread out his units to minize the amount of hits) because they can potentially harm him pretty bad. It's a fine line between having a good potential unit and a shit army. Still, it's good not to be too fixated on pure probability.
In game decision making is even more linked with potential. There are always the "easier" options, and I see people often advocating destroying the easy targets first with target priority. However, in my opinion the probability of success should always be contrasted with the benefit of a given outcome. I played a game against Jabbdo a couple of days ago where I forced a morale test on his destroyer lord and immortals. He failed ld10 twice in a row and ran off the table. Then I was able to use his scarabs as a springboard for troops and coupled with deepstriking reserves I was suddenly all over his objectives. The destroyer lord and 10 extra immortals might have made all the difference. Now I was able to win with max points.
It wasn't very likely that he would fail ld 10 twice in a row, but what did I lose by trying? A couple of random shots that probably would have killed the same amount of models from other units. It's good to know your probability but you shouldn't read it like the bible. There are other factors that should be weighing in on your decisions.