Intuitively speaking, ETH seemed to have simply increased the average block generation rate from one every 10 min in BTC to 1 every 15 seconds. There doesn't seem to be any other protection against excessive branching.
So does ETH fundamentally have a lot more branches, and a lot more orphaned transactions than BTC does? Thanks!
Just to add to Rob's answer.
So does ETH fundamentally have a lot more branches,
Yes, as per Rob's answer.
and a lot more orphaned transactions than BTC does?
Yes, but, if most miners are using the same default transaction ordering* (see What is the default ordering of transactions during mining, in e.g. geth?), and have the same view of the tx pool, then the transactions in the orphaned blocks are probably the same transactions as in the winning block anyway. (If that's the case, then are these transactions actually orphans?)
(*There's probably no way of knowing this for sure.)
Yes, but they are routine and are less of a concern than they would be if the Bitcoin algorithm was arbitrarily speeded up, Ethereum's proof-of-work is not merely bitcoin's with a faster blocktime.
The concern is that faster blocks imply a predictable increase in branching because of network latency, and consequently faster blocks leads to loss of hash power - network inefficiency.
GHOST - Greedy Heaviest Observed Subtree is a solution to the loss of hash power, first introduced in 2013 by Yonatan Sompolinsky1 and Aviv Zohar. https://eprint.iacr.org/2013/881.pdf
Ethereum doesn't implement GHOST exactly as described in that paper, but the original work is a good place to start exploring Ethereum's implementation.
Have a look here for details about the Ethereum implementation: https://github.com/ethereum/wiki/wiki/Design-Rationale#uncle-incentivization
Hope it helps.
