Well, they didn't really.
You may have noticed in the past few weeks when Solana’s blocks got faster, staking yields appeared to fall at the same time.
On the surface it seemed like staking yields got worse, but actually it was the APY formula underneath adjusting to the new and faster Solana blocks.
So what actually happened?
You may have seen the news in the past few weeks that Solana block times were reduced from 400 milliseconds to 350 milliseconds (August 21) and then one week later (August 28), they came down again from 350 milliseconds to 300 milliseconds.
Quick reminder on how Solana staking works: it runs on epochs, and an epoch is exactly 432,000 slots. A slot is the network's own unit of time, a fixed window in which one validator gets to produce a block, and it ticks whether or not a block actually turns up. Rewards for those 432,000 slots are calculated and paid out at the start of the following epoch.
So with block times getting faster on Solana but the same total number of slots per epoch, the amount of human time for 1 epoch decreased by about 25%! This means instead of an epoch taking two days, it now takes around 1.5 days.
| Slot time | Epoch length | Epochs per year |
|---|---|---|
| 400ms (before epoch 1020) | ~48 hours | ~183 |
| 350ms (epoch 1020) | ~42 hours | ~209 |
| 300ms (epoch 1024+) | ~36 hours | ~243 |
So what does this have to do with APY?
APY itself isn’t an objective or factual datapoint sitting on-chain. Instead, it’s a calculation with some assumptions baked into it.
The two most important inputs to the staking APY calculation are:
- How much SOL was earned in a given epoch; and
- Roughly how many epochs are there in a year (the 'A' in 'APY').
Number 2 is where the recent block time changes tripped up APY display calculations.
Why is there no such thing as a 'true' APY?
Because input 2 can never be known exactly.
There's no guarantee exactly how long an epoch will run in our clock time. Some blocks are slower and some are faster due to the behaviour of the validators adding blocks to Solana. This means even though an epoch is always 432,000 slots, it may not always be 36 hours.
So APY is always a projected estimate with assumptions built into it, the number of epochs in a year being the biggest.
Conclusion
Solana is on a mission to IBRL (Increase Bandwidth and Reduce Latency) and every time it makes blocks faster, staking display APY's can temporarily fall. The good news is, the underlying earnings are untouched and it usually just takes a few days for the display formulas to smooth out.