you are right the miners don't really notice the change, but what ever shares where left behind in the other coin will not payout since we will no longer find a block there
Which doesn't matter, does it?
You can either try to find a block on the old coin or the new one. If the new one pays better, you should try the new one.
Say the pool has accumulated 10,000 shares on coin A, and not found a block. Coin A is predicted to payout 1 BTC's worth of reward. Coin B rises in profitability, and is predicted to payout 1.1 BTCs in profitability.
Scenario 1: The pool stays on coin A, and after another 5,000 shares it finds a block. Reward per total shares of work is 1/(10000+5000) = 0.0000666 BTC
Scenario 2: The pool switches to coin B, and after 5,000 shares it finds a block. Reward per total shares of work is 1.1/(10000+5000) = 0.0000733 BTC.
The key thing is that you are not working towards finding a block. Each share has an independent chance of solving the block, and the fact that you've already put in 10000 shares of work makes absolutely no difference to the chance that the next share found will solve a block.
so in your scenario 2, we still get paid for the 10,000 shares left in the other coin?? not so sure it works that way, but eleuthria can clear it up....
No, but you are paid a lot more for the 5000 shares in the new coin.
The end result is that you get paid more, so what does it matter how?
In scenario 1 you actually receive 0.0000666 per share submitted for coin A. Lets say you are 1% of the pool, so you submitted 150 shares, and get 0.01 BTC for coin A.
In scenario 2 you actually receive 0.00022 per share submitted for coin B, and 0 per share submitted for coin A. You submitted 50 shares, and get 0.011 BTC for coin B, and 0 for coin A.
In total, you get more money from scenario 2.
I see, but that is assuming we don't spend more time/shares trying to find the new coin right?
The pool switches to the coin that is expected to provide the greatest value, calculated (simplistically) as block reward/expected number of shares to find a block.
So if the pool switched to coin B, it means that the expected reward
per share submitted is greater than coin A.
So on average it doesn't matter if it takes more or less time to find a block for coin B than coin A, that has already been factored into the equations.
Say instead we had a scenario 3: Coin 3 is expected to payout 30 BTC, but is 20 times more difficult to solve than coin A.
So the pool switches to coin C, and after 100,000 shares it finds a block for coin C. Reward per share for coin C is therefore 0.0003.
In scenario 3 you actually receive 0.0003 per share submitted for coin C, and 0 per share submitted for coin A. You submitted 1,000 shares, and get 0.3 BTC for coin C, and 0 for coin A.
Your reward per share submitted in scenario A is 0.0000666. Your reward per share submitted in scenario C is 0.0002727.
There are two variables:
a) How much is the pool going to make overall
b) How much is my share of the pool's profits going to be.
b) Is simply your accepted hashrate divided by the total pool accepted hashrate, so what matters is a), which is maximised by switching to the most profitable coin. No matter how often the pool switches coin, it doesn't (in the long term) affect b).