Not understanding gas calculation on an EVM network [Fantom]

Question

This is a question about Gas fees for transactions. The bottom line is I can't understand the price (in gas units) of transaction - And I will walk you through a specific transaction to show you my calculations, so you can point out what I'm missing.

So, the example I used is for this transaction which is on fantom, but fantom is EVM compatible (from the ISTANBUL hardfork). Also, I'd like to point that I've read this great article about gas calculation and I will use it throughout this question.

So lets start: The gas provided for this transaction was 1,000,000 units.

Initial fee is always 21,000

Now for the Input data fee - we can take the input data from the ftmscan, and with a short python script calculate the fee:

s = "0xcac637c8000000000000000000000000000000000000000000000000000000000000004000000000000000000000000000000000000000000000000000000000000000a0000000000000000000000000000000000000000000000000000000000000000200000000000000000000000021be370d5312f44cb42ce377bc9b8a0cef1a4c83000000000000000000000000aa621d2002b5a6275ef62d7a065a865167914801000000000000000000000000000000000000000000000000000000000000000673706f6f6b790000000000000000000000000000000000000000000000000000"
s = s[2:]
chunks = [s[i:i+2] for i in range(0, len(s), 2)]
sum = 0
for c in chunks:
   if c == '00':
       sum += 4
   else:
       sum += 16
print(sum)

the result of this is 1,560. So the gas we paid so far, before any opcode, is 1,560+21,00=22,560

Ok, Now, I've looked at the traces for this transaction (by doing what it says here

curl -H "Content-Type: application/json" -X POST --data '{"method":"debug_traceTransaction","params":["0x0bef4e97cc39b98431c81c4573572905a735d324724e83619001853e344e1092"],"id":1,"jsonrpc":"2.0"}' https://rpcapi-tracing.fantom.network

So far, we've spent 22560 gas units, so we have 1,000,000-22,560=977440, which is great, because that is exactly the gas value in the traces before the first opcode. So far so good.

Now I'm going to the bottom of the trace: when the STOP opcode (which has a gasCost of 0) was executed, the amount of gas left is 833,994. So the amount of gas used in total is 1,000,000-833,994=166,006 gas units

But this is weird - ftmscan says that the gas usage was 226,050, which is waaaaaay higher. Now, I know I didn't take into account refunds. Buy here the deal - refunds are gas units returned to the user, so any refunds (and there is some refund here) will actually decrease the gas cost even more.

So - here is finally the question - how come this transaction gasCost was 226,050 gas units on ftmscan, and much less when running with their own debug trace, and on my local simulator (which I built using Foundry)??

Answer 1

All your considerations are correct. The problem is that Fantom is only 99% EVM equivalent, you can find an explanation here: https://github.com/Fantom-foundation/go-opera/wiki/EVM.

In particular, on Fantom 10% of unspent gas gets spent as a disincentive to militate against excessive transaction gas limits. So the actual gas spent by that transaction would be: 833994*0.1+166006 = 249405 gas, according to the theory.

The final result is not the same as 226050 (but still it explains why we have an increase). I'm not sure how to interpret the discrepancy, one possibility is that the debug trace isn't trustworthy since it runs the tx as if it was at the beginning of the block, while the actual tx was in the middle (so the first txs may have tampered the state).