The current design of Ouroboros allows for anyone who holds stake to technically have a chance to be a slot leader and therefore advance the network


[5:45]
in practice, most will want to delegate their stake to a stakepool that is reliable


[5:45]
so there are two eras


[5:46]
One is when we are still in the replicated world


[5:46]
meaning that more delegates add no additional resources to the system


[5:46]
and one is in the Sharded world


[5:46]
where more delegates means more resources


[5:46]
during the first era, it makes sense to try to keep the number around 25-100


[5:47]
Assuming the majority show up reliably for their slots


[5:47]
During the latter era, it would be nice to have thousands


[5:47]
This requires slightly different incentives and designs


[5:47]
So for Shelley, we are going to focus on the following


[5:48]
1) Make it really easy to register and run a stakepool


[5:48]
2) Make it really easy for an end user to delegate their stake to a stakepool


[5:48]
3) try to stabilize the network topology around 25-100 stakepools maintaining the network


[5:49]
4) Make sure there are lots of relays and full p2p nodes to provide network resilience


[5:49]
There are a lot of subtle things we have to update, change or improve to make this fully work out


[5:50]
and we are also examining different game theoretical models


[5:50]
the game theory is basically breaking the world into a set [ _  _ _ ... _ _ _ ]


[5:50]
with two players


[5:50]
H


[5:50]
and L


[5:50]
H has a high probability of showing up for their slot when the time comes (>0.9)


[5:51]
L has a low probability of showing up for their slot when the time comes (> 0.1)

[5:51]
we assume a U distribution of players so most will cluster around these two roles

[5:51]
So there is a target network efficiency

[5:51]
NE is defined as the amount of slots filled over the total amount of slots

[5:51]
so 1 means 100 percent of all slots have been filled for the epoch

[5:52]
so you set a target and have to calculate the ideal balance of H to L in the system


[5:52]
and then you have two levers you can pull to shift players from L to H or H to L

[5:52]
rewards

[5:52]
which equal inflation + transaction fees


[5:52]
and punishments

[5:53]
which are things like slowing down quality of service to even making coins inaccessible for some time period if a person doesn't delegate or is in the L set (edited)

[5:54]
the goal is optimize the game so that the least rewards are given to ensure NE is above target


[5:54]
therefore, we minimize the operational cost of the network


[5:54]
so there are lots of interesting questions


[5:54]
such as P(H | high percentage of stake)


[5:54]
P(L | high percentage of stake)


[5:54]
and so forth


[5:55]
and how to estimate these without necessarily having empirical data


[5:55]
The other challenge is dealing with expectation of ada value


[5:55]
this greatly impacts whether someone thinks the reward is sufficient


[5:56]
but this isn't a well understood notion


[5:56]
So we are starting with some very simple models right now with the Oxford team


[5:56]
and building complexity slolwly


[5:56]
and our hope is to deduce an ideal model


[5:57]
We can always make an educated guess and that's actually what Satoshi and others have done


[5:57]
but it's not optimal


[5:57]
punishments are also interesting, they are understudied as well