From PGAs to 1559 to ETH2 — How Everything (and Nothing) Changes With Gas Fees - Matt Cutler

uh happy east denver i'm so excited to be back this is my fourth east denver total third in person because we weren't here last year unfortunately it's super great to see the community and we're gonna spend the next 20 minutes or so going deep into the world of gas fees it's going to be a bit of a crash course i move very very quickly to try to cover as much possible um but i'm here with you for this reason uh boomers and ethereans alike like to compare complain about gas prices right and it never seems to go away and the question is why because i thought we were supposed to fix it or it's going to get fixed in the future we're going to talk today about how they work and why they work and what we can expect moving forward so we're going to talk about the past present and future of gas prices so prior to eip 1559 the original model was a very simple first price auction you set your gas price the miners prioritize the highest fees there was a fixed block size with a 15 million gas limit didn't didn't change and the miners received all of the fees at the end of it super straightforward uh it's simple and you had one variable to worry about very easy everybody got it but there was this issue where trading bots and other systems would often go to war with each other and they would escalate the gas price to win a certain opportunity and this would bid up the gas price of the entire network these are called priority gas auctions and you can see one of them this is a representative one here um if the gas prices for the network are low the bots go to war the pga happens it tabletops and then it sort of declines and if you're just a regular user doing a regular thing on your favorite d5 protocol or your favorite nft you suffered as a result of this so this was the big problem with this original fpa model where it was super difficult to figure out what you needed to do to get into the next gas the next block and you had these big surges okay and it made the network hard to use and people complained about gas prices and so as a result a lot of research went into what's known as eip 1559 which hopefully all of you are familiar with that significantly changed the nature of the gas market on ethereum okay and uh this was launched as part of the london hard fork in august of 2021 we at block native were very involved in this we were one of the first if not the first to support it at the london hard fork and it introduced a whole series of changes that really really affects the way that the network works first we now have dynamic block sizing so the blocks can actually change dependent on demand there's three variables that go into determining gas price a base fee a priority fee also called the minor tip and what's known as a max fee which is super disorienting for users the base fee is the minimum inclusion for the block and is set by the network now this is kind of a major change because prior to eip 1559 the network itself was not aware of what its gas fee was it was totally set externally by the miners and so the network actually starts to become self-aware here this is sort of interesting in its implications also the base fee must be paid it doesn't matter how your transaction gets into the block you could be a miner and you can shove it in there through a private network you still got to pay the base fee and the base fee must be paid in eth so it also establishes eth as the core currency of the ethereum network which strangely enough was not necessarily the case up until eip 1559. there's a lot happening here the base fee is then burned so the network charges the base fee and it goes away which exerts deflationary pressure on the circulation of eth which should in theory increase price as network activity increases and then finally the priority fee which is set by the user is paid to miners so you have these two variables base fee set by the network priority fee set by the uh the user the base fee gets burned the priority fee goes to the miner got it it's more complicated okay now uh we talked about paying that the base fee has an algorithm which is targeting 50 full blocks okay so the variable block size is 0 to 30 million in this model but if a block is 100 full the base fee will increase by the maximum in the next block of 12 and a half percent so block to block the most that the base fee can increase is twelve and a half percent if that blocks a hundred percent full it'll go up by twelve percent similarly the most the base fee can be reduced in between blocks is minus twelve and a half percent but it only does that with a zero percent full block empty blocks which are now a thing with miners actually mine blocks that have no transactions in them when the base fee goes out of control there's no demand and the base fee uh the base fee goes down by twelve and a percent blocks are empty okay when you set your priority fee you will always be charged that that is guaranteed to go to the minor it's technically optional but generally it needs to be at least two gray and you set this max fee but you're almost never charged the max fee and this is super disorienting for users because they put this number in that says the maximum amount your transaction is going to be charged but that's what you're authorizing it's sort of like a credit limit in a way and only under very specific network conditions will your max fee actually go into effect however setting the max fee incorrectly can reduce transaction performance and predictability so this max fee thing is a little bit uh confusing the big risk is the base fee is the minimum price you set for inclusion in the next block so let's say the base fee is 50 and you're like great i'm going to price my transaction at 50 but before your transaction can get included the base fee rises to 55. your transaction is therefore underpriced and is not a candidate for inclusion there's no way your transaction can get into a block until the base fee comes back down to the 50 that you set it at so what you want to do is set a max fee that has a safety factor in it that basically can account for hey what happens if the base fee um goes up so this whole thing about max fee settings a little bit tricky in the world of eip 1559 the common heuristic is your max fee should be twice your base fee plus your priority fee with the idea being if you double the base fee you're covered for the next six blocks because at a maximum rate of increase the base fee will double in six blocks okay but what's disorienting for the user is now the transaction looks like it's going to be super expensive and then the transaction gets on channel we don't celebrate to say hey you paid way less than your max fee okay so we work with folks across the ecosystem all the time and and it's interesting how persistently confusion on what the maxi actually represents when it comes into play why you want to set it different ways is actually part and parcel of what's going on so i really encourage everybody to get familiar with these concepts as part of the current network with eip1559 so the good news with eip1559 is we have variable block sizes so the network can adapt it's much easier to estimate what's going to happen because it's not supposed to move around so much and we have this deflationary pressure on the network there are other benefits to eip 1559 transactions also called type 2 one they actually tend to cost less if you're submitting a transaction that's type 0 the old model on average you pay about 14 more okay now do users care about overpaying for gas absolutely they do so we highly encourage whatever you're building to be type 2 or eip1559 compliant so that your users don't overpay unnecessarily two an unexpected side effect is type 2 transactions tend to go on chain slightly faster so by having type 2 transactions they get included slightly faster than type 0 transactions so another benefit of being eip 1559 compliant now however in the wild these factors interplay with each other in interesting and unexpected ways and so this is some real world data from an nft launch galaxy eggs on uh september 14th of 2021 and and here's what the base feed did during this galaxy eggs nft launch so the nft launches everybody's super excited surge of transactions in the network and we can see the base fee is going up we have full blocks at the 12.5 percent rate up up up up up up up up and then it gets to about 2500 gray kind of tabletops and then it descends down okay nice smooth algorithmic and predictable but remember under eip 1559 there are two factors which you have to include you have your base fee and your priority fee what was going on with the priority fee oh oh interesting so basically we saw this huge surge in the minimum priority feed minimum in the block the bottom of the block right these big jumps that's 2 000 gray and the time here is sort of set to the beginning if i overlay these two look at this when the base fee is low the priority fee goes high and as the base fee increases the priority fee declines they interact with each other there's actually an algorithm at play here the algorithm of eip1559 and there's the psychology of the users who are trying to participate in the nft drop and these two combine to look like this this was the actual network conditions during this drop and if we combine them what does that look like it's the exact same table top we had before rapid rise flat across the top and a decline so even under eip 1559 priority gas auctions are still a thing they just have more variables they're a little less likely to occur but we must pay attention to what's going on with the network we must be aware that there can be sudden changes in the conditions to get transactions included and that the psychology and these algorithms interact in ways that are often not obvious okay and this is probably the most important element that we pay attention to when it comes to predicting gas fees is it's pretty easy to predict the algorithms it's incredibly hard to predict the psychology okay and what's coming so there's some cons here there's some changes coming with the merge that we believe will change some of the psychology and this psychology changes will therefore have impacts on the state of the gas marketplace so oh uh here's another interesting thing that today even after all this time of eip1559 being on mainnet still only 75 percent of transactions are type 2. so fully 25 of transactions on ethereum are overpaying by about 14 on average they're getting included slower the network has to do more work so we still have work to do to drive adoption of type 2 and make eip1559 completely uh transparent through the whole network you guys with me so far i told you this is going to be a crash course right so we went through uh the beginning of where we started fpas first price auction and that turned into pga priority gas auction we reboot the model we moved to eip1559 we now have three variables at play how those variables get expressed and which actually get charged become a lot less clear so the eip1559 is more efficient it's more predictable but it's harder to understand okay so as we look ahead what's coming next is the merge so what's going to change with gas fees during the merge so estimate mid this year coming up pretty quick ethereum mainnet will merge with the beacon chain and you know what almost nothing changes that we still have dynamic block sizing we still have base feed priority max fee but instead of a minor tip it now becomes a validator tip okay we still have the base fee set by the network priority fee set by users it's still burned and the validators are going to receive the priority fee now the uh i won't go too far into this but the validator model and the deflationary pressure should increase as a result of the merge so the good news is once you learn eip1559 you're gonna be well prepared for what's coming with the merge however there's this change and this change is very subtle but we think this change is going to be worth paying attention to and it takes a little explaining and this change is known as fixed block times so today under the current system proof of work you have variable block times the network targets about 13 and a half seconds per block on average but we all know a block can be confirmed in one second or 60 seconds that there's actually no way to know when the next block is going to be confirmed okay it's completely unknowable and so therefore if you're trying to get into the next block you don't know when the block is going to happen so every millisecond is equally valuable as the next millisecond because everyone's trying to get on the next block nobody knows when that block is going to get confirmed and everybody competes equally through a consistent time horizon okay that time is basically flat and linear and transactions flowing into the transaction pool or the mempool are can happen at any time but there's this subtle change which is coming with the merge and under proof of stake blocks will be confirmed every 12 seconds like clockwork and so what does that mean that means not all milliseconds are equal anymore because if i'm trying to get something on a ch on chain and i want to make sure i get it and someone doesn't beat me to it when am i going to transmit my transaction and i'm going to transmit my transaction in the first second after the block was confirmed or am i going to wait to 2nd 11th maybe second 11 and a half so when i transmit the transaction all my competitors out there have a smaller time window to respond and so what's significant here is this will impact intra block transaction velocity okay today in the current proof of work model we have a constant velocity all milliseconds are equal as we move to proof-of-stake as we move to fix block times we will have this surge effect where the second or two before the block gets confirmed more transactions will flood into the pool there'll be tighter time horizons for various automated systems to compete with each other there'll be a rush to lower and lower latency infrastructure in order to detect and respond and we believe that this will have an impact on gas fees right because the nature of transactions flowing in and out will be different but the question is what change what's going to happen and the answer is nobody knows yet right there are researchers who are working on this and what we're going to have to do is look at this theory and compare it to the reality okay that what we saw with the launch of eip1559 was a bunch of new behavior we hadn't seen before like the interaction and interplay between the priority fee and the base fee under periods of rapid congestion and i believe we believe at block native we will see some similar new phenomena uh post merge that will affect the entire network and the nature of the gas fee marketplace so this is super fun this is super interesting and it is super meaningful oh there's another thing looking past the merge where we look towards um sharding which will increase block space and we have a hint of what might happen here by looking at l2 and sidechain smart contract transaction volumes so this chart real quick it shows the blue is smart contract transactions and the um orange is just simple transfer transactions on ethereum mainnet this line is the polygon side chain simple transactions just simple transfers and this is polygon sidechain smart contract transactions so where we have polygon sidechain where we have more uh reduced gas fees where we have more block space theoretically what we see is significantly more smart contract transaction volume so as we look ahead to sharding which will be 2023 time horizon we may see massive increases in overall smart contract transaction volume so there's this thing that happens which is look if we just had enough block space the gas prices would go down it's sort of like saying if we just had enough highways there would be no traffic right it turns out that as you add highways you add traffic because more people will travel and so similarly as we look to add block space there's this going to be corresponding increase in demand for that block space and this will still matter so are we going to be talking about gas fees next year probably are we going to be talking about gas fees after that probably are we going to be talking about gas fees 10 years in the future probably why because block space is valuable and so in the past gas fees mattered in the present right now gas fees matter and in the future gas fees are going to matter so we all have to pay attention to that so this meme isn't going to change we're all going to still be talking about this and and what we need is continued research and we invite people in um we at block native publish real-time data on the state of the gas marketplace we encourage you to check that out and incorporate that find new things look through historic blocks there's tons of interesting stuff happening here we believe we can improve the user experience so users have a better idea of what they can expect for what their transactions are going to cost where they should put transactions on a side chain or l2 to have less and and what the outcomes are going to be and that fundamentally with more capable infrastructure we can improve the transaction experience across the board and as we do so we're here to help check out our gas estimator check out our gas api we're block data.

com and this has been super fun i'm matt cutler really appreciate your time and attention and please have a great rest of east denver okay thanks everybody [Applause] that was amazing thank you thank you thank you