Minor correction to reward vesting semantics

[theoreticalbts]

There are two reasonable behaviors for the Steem Power portion of post rewards: Either the reward for a post is vested when it is rewarded, or the reward for a post is vested when it is claimed.

The code does not implement either of those two reasonable behaviors. Instead, it implements rewarding a post with Steem, plus an option to vest that Steem at the VESTS / STEEM exchange rate at the time of the reward. Claiming rewards exercises the option, which generates profits since the actual exchange rate at the time of the claim is less favorable than the exchange rate at the time of the reward. Profits from the option exercise are funded by corresponding losses from the vesting pool.

Existing implementation

Here is an example of the bug at work:

• Suppose Alice is rewarded 10 Steem for a post.
• Suppose at the time of her reward, there are a total of 1000 VESTS and 1000 Steem.
• Therefore Alice's reward_steem_balance is 10 Steem, and Alice's reward_vesting_balance is 10 VESTS.
• Let's say Bob owns all 1000 VESTS, his value is 1000 Steem.
• Then some emissions occurs, now there are 1500 Steem and still 1000 VESTS. Bob's value is 1500 Steem.
• Alice executes claim_reward_balance_operation.

How many VESTS / Steem does everyone have after this sequence of events occurs?

• Bob has 1000 VESTS.
• Alice has 10 VESTS.
• These VESTS represent 1510 Steem.
• Alice's value is 10 * 1510 / 1010, or 14.95 Steem.
• Bob's value is 10 * 1510 / 1010, or 1495.05 Steem.

Carefully checking the numbers demonstrates that 4.95 Steem worth of value moved from Bob to Alice.

In this example, Alice represents 1% of the vesting pool, and the vesting pool was increased 50% from emissions. The error is approximately equal to the original vesting pool times the product of these two percentages, in this case, 0.5% of 1000 Steem. Fortunately for us, the error is hopefully small:

• The vesting pool is very large compared to the rewards
• This change was implemented after the hyperinflationary period; the vesting pool's increase in percentage terms was small
• If most users claim their reward balances quickly, each user only saw a small percentage increase between reward and claim

Vest when rewarded

What happens if post rewards start vesting when they are rewarded?

• Alice is rewarded 10 Steem for a post.
• At the time of her reward, there are a total of 1000 VESTS owned by Bob, the vesting pool is 1000 Steem.
• Alice's 10 Steem are converted to 10 VESTS
• Now there is 1010 VESTS and a vesting pool of 1010 Steem.
• Some emissions occurs, now there are 1010 VESTS and 1510 Steem.
• Alice's value is 10 * 1510 / 1010, or 14.95 Steem.
• Bob's value is 10 * 1510 / 1010, or 1495.05 Steem.

This is the exact same result as given by the existing implementation. Are we sure this is a bug?

Charlie scenario, existing implementation

Yes, it is a bug. We can explain why with a slightly different scenario:

• Alice is rewarded 10 Steem for a post.
• At the time of her reward, there are a total of 1000 VESTS owned by Bob, the vesting pool is 1000 Steem.
• Therefore Alice's reward_steem_balance is 10 Steem, and Alice's reward_vesting_balance is 10 VESTS.
• Some emissions occurs, now there are 1000 VESTS and 1500 Steem.
• Charlie deposits 15 Steem, he gets 10 VESTS.
• Bob withdraws 1000 VESTS, he gets 1500 Steem.
• There are now 10 VESTS and 15 Steem.
• Alice executes claim_reward_balance_operation.
• There are now 20 VESTS and 25 Steem.
• Charlie withdraws 10 VESTS and gets 12.5 Steem -- less than he put in!

Charlie scenario, vest when rewarded

Here's a scenario that shows correct behavior for vesting when rewarded:

• Alice is rewarded 10 Steem for a post.
• At the time of her reward, there are a total of 1000 VESTS owned by Bob, the vesting pool is 1000 Steem.
• Alice gets 10 Steem and 10 VESTS.
• Some emissions occurs, now there are 1010 VESTS and 1510 Steem.
• Charlie deposits 15 Steem, he gets 10.033 VESTS.
• There are now 1020.033 VESTS and 1525 Steem.
• Bob withdraws 1000 VESTS, he gets 1000 * 1525 / 1020.033 = 1495.05 Steem.
• There are now 20.033 VESTS and 29.95 Steem.
• Alice executes claim_reward_balance_operation.
• Alice redeems 10 VESTS for 10 * 29.95 / 20.033 = 14.95 STEEM.
• Charlie withdraws 10.033 VESTS and gets back his original 15 Steem (less rounding errors).

Charlie scenario, vest when claimed

Here's how the Charlie scenario works out, showing correct behavior for vesting when claimed:

• Alice is rewarded 10 Steem for a post.
• At the time of her reward, there are a total of 1000 VESTS owned by Bob, the vesting pool is 1000 Steem.
• Some emissions occurs, now there are 1000 VESTS and 1500 Steem.
• Charlie deposits 15 Steem, he gets 10 VESTS.
• There are now 1010 VESTS and 1515 Steem.
• Bob withdraws 1000 VESTS, he gets 1000 * 1515 / 1010 = 1500 Steem.
• There are now 10 VESTS and 15 Steem.
• Alice executes claim_reward_balance_operation, buying 10 Steem worth of VESTS.
• Alice gets 6.66 VESTS, there are now 16.66 VESTS and 25 Steem.
• Alice redeems 6.66 VESTS for 9.99 Steem (her original 10 Steem less rounding errors)
• Charlie redeems 10 Vests for 15.01 Steem (his original 15 Steem less rounding errors)

Which is correct?

Clearly the current implementation is incorrect.

The "vest when claimed" scenario fits better with the modular blockchain, which was the original motivation for this change, as detailed in #659.

[theoreticalbts]

@mvandeberg I would appreciate your input on this issue.

[mvandeberg]

The reason behind the implementation is that while the VESTS/STEEM price may decrease when rewards are claimed, the delay means an author always receives fewer VESTS than would have if vesting happens instantly. Vesting cannot happen instantly because it makes content a dependency of vesting balances, which are needed for the witness schedule. Thus making content required for the minimal module. That is a non-starter for scaling. I think within the context of STEEM, the errors are minimal enough that it is not a problem.

However, SMTs have the possibility of pushing corner cases, it may be warranted to re-evaluate this assumption. It really depends on what you define as stealing. For VESTS holders, a large reward being claimed does make their value decrease. However, only as much as you consider that their value is artificially higher than it should be because the rewards aren't claimed. The same argument can be made for the author that if they don't claim for a significant period of time, they end up with a smaller reward. (Opportunity costs from not vesting)