If I had 51% of the networking power a year ago...

[Riley]

...couldn't I still be working on a false chain that would be as long as the widely-accepted chain?

Say there are 2 miners on the network right now, and I control one of them.  Why can't I...
a) wait 'til my miner creates the longest chain
b) at that moment, disconnect from the network and just keep processing my chain, making it longer
c) at an arbitrary time, say, "hey guys, guess what, I've got this chain and it's 4 blocks longer than yours and I have all the bitcoins in it!"

[1685395039]

1685395039

[rezin777]

The longest chain is the most difficult chain.

[MysteryMiner]

Overall chain difficulty, not quantity of blocks makes longest chain. Read the wiki!

[brybot]

Difficulty and length are somewhat intertwined. If you keep making coins at a constant rate, your difficulty stays constant. If you increase your hashing power, you making coins faster, but the difficulty goes up. So with a rising difficulty, you are creating a longer chain than you would otherwise.

[CydeWeys]

Just to answer the same question in a slightly different way that makes the most intuitive sense to me -- In order to have a blockchain longer than the present blockchain, you would have had to have had MORE average hashing power than the network as a whole.  Here's why -- In the same amount of time (a year or whatever), you would have found more blocks than the current chain; thus, your average difficulty level would've had to have been higher on those found blocks.  You can't find lots of low difficulty blocks really quickly; the Bitcoin client would reject them as invalid because the difficulty level should've been a lot higher.

[bcearl]

To get four blocks ahead you really need a lot advantage in computing power and still some luck.