How well does any.sender cope during events like YAM Wednesday?

Peak of the ICO boom in 2017 pales in comparison to the fees paid over the past few weeks.

any.sender is a non-custodial and high-throughput transaction relayer. Our goal is to optimistically pay the best network price while getting your transaction confirmed quickly, even during crazy and significant congested days like YAM Wednesday, so how did we do?

Users paid 4.8x fees on Ethereum compared to Bitcoin

The casino environment of degenerate and chad traders that fuelled the 2017 bull-market has returned. But this time, thanks to the rise of DeFi, it has bypassed the centralised exchange gatekeepers and a significant portion of the action is playing out directly on Ethereum.

So much so that Ethereum block producers collected over $6.8m in fees during the 36 hour YAM yield farming hype. The typical fast rate for a transaction increased from ~200 gwei to ~330 gwei.

To put the network fee costs into perspective:

• ETH transfer at 21k gas ~$2.45 (~0.0063 ETH),
• Uniswap trade at 100k gas ~$11.64 (~0.03 ETH),
• Deploying a gnosis safe wallet at 250k gas ~$29.20 (~0.075 ETH),
• Spend the entire block limit at 12m gas ~$1401 (~3.6 ETH).

While YAM ultimately failed due to a smart contract bug, it is still astonishing the impact the craze has had on the fee market.

Can any.sender help alleviate the network fees?

There are two approaches for beating the crazy network fees:

• Layer-2 solutions. Send your transactions via rollup,
• Smarter fee strategies. Optimistically send the transaction at a low fee and then strategically increase the fee until the transaction is mined.

While there are several really exciting layer-2 solutions in the works, they are not yet ready to support the DeFi craze. That leaves us with the smarter fee strategy option which is what we have focused on at any.sender.

What is our strategy? We send transactions to Ethereum at a fee below the fast rate. We re-evaluate every transaction in the queue when a new block is mined and then increase its fee according to a pre-defined fee strategy. The transaction is re-published when the fee estimation has increased by ~13%.

Our transaction fee bumping strategy is fairly simple, but it optimistically lets us catch the network fee lows while targeting a ~1–3 minute confirmation time.

Of course, our fee strategy must change over time to meet the new demand from the fee market, but for now that is how it works. So how well does it hold up?

The Life and Death of YAM

tldr; any.sender transactions outcompeted a popular exchange and popular wallet relayer by 56–77 gwei on average, and the estimated fast rate by 19 gwei on average. All transactions were mined in under 6 -7 minutes (vast majority <5 min) and the transactions with the largest fee savings were always mined the fastest.

Methodology. any.sender sent a new transaction every three hours using the fee strategy described earlier. We tracked transactions from a popular wallet relayer and a popular exchanges, and pick their transactions based on the time of inclusion in the blockchain. Finally, we captured the fast rate from etherchain which provides similar statistics to the popular ethgasstation.

Figure 1: Approximately 23 transactions during the Life and Death of YAM (Wednesday — Friday).

Figure 1 shows the list of gas prices paid for transactions from any.sender, the popular wallet relayer and the popular exchange relayer, alongside the estimated fast rate.

The range of fees is quite staggering. It ranges from 116 gwei (any.sender) to 432 gwei (popular exchange). The estimated fast rate steadily increased, drastically fell, and then rose at a meteoric rate. YAM Wednesday is a great demonstration of the volatile network fee that relayers must handle in real time. It is never a boring time on the Ethereum network…

Let’s walk through the timeline:

• 3pm 8/12/2020–11am 8/13/2020: The life of YAM.

It was an exciting time as farmers were actively managing their YAM crop, and then sending tractors to save the YAM, and then eventually harvesting the crop before the once-in-a-life-time-freak-YAM-killer arrived.

any.sender paid under the fast rate for 4/9 transactions during this period and only in one occasion did the popular wallet beat the any.sender price. Given that miners collected over $6m transaction fees during this period, it is remarkable that any.sender could still pay below the estimated fast rate.

• 11am 8/13/2020–2am 8/14/2020: The death of YAM.

We miss you YAM

any.sender paid below the fast rate for 5/7 during this period. None of the measured relayers beat the any.sender price.

• 2am 8/14/2020 — present: The rise of CURVE?

We caught the start of the CURVE launch… just an honest days work…

any.sender paid below the fast rate for 4/6 transactions during this period. None of the measured relayers beat the any.sender price.

Figure 2: any.sender relative savings (above zero) or cost (below zero).

Figure 2 highlights that any.sender beat the other relayers except for one occasion and mostly beat the estimated fast rate as well.

Let’s take a brief look:

Exchange savings. The exchange overpayment ranged from from 23 gwei to 227 gwei. any.sender beat the exchange relayer by 77 gwei (on average).

Wallet savings. The wallet overpayment ranged from from -13.81 (beat us once) to 152 gwei. any.sender beat the wallet relayer by 56.9 gwei (on average).

Fast rate savings. There were several times when any.sender beat (and got beaten) by the estimated fast rate. The estimates ranged -17.25 gwei to 79.03 gwei. Overall, any.sender still beat the estimated fast rate by 19 gwei (on average).

Figure 3: We target < 3 minute confirmations, but it was pretty tough on YAM Wednesday.

We have fine-tuned any.sender for <2–3 minute confirmations as that is applicable for most applications while giving us ample time to catch the super-lows.

While some transactions took slightly longer than expected for us, we never had any stuck transactions.

The data we have gathered over the past two days can be used to feed back into our design. So next time YAM² appears, we will be ready to beat the time-to-mine stats.

Why was the any.sender saving modest compared to the fast rate? Block producers collected over $6m in fees on YAM Wednesday which represents the largest fees paid ever in the history of Ethereum. It is pretty incredible that any.sender still managed to beat the estimated fast rate. We believe the modest saving was mostly due to the bumps over-jumping the fast rate, assuming the estimated fast rate is a reliable statistic.

Why was any.sender the only relayer to pay less than the estimated fast rate? The current strategy by the other relayers appear to simply set a fixed rate and then periodically adjust the fixed rate in response to the network fee volatility. Most services are happy to pay a higher gas price as a single stuck transaction will clog their entire queue and relayer software cannot typically cope well with that.

But who pays the extra fees? Ultimately it is you, the user, who is sometimes paying through the nose.

We built any.sender to save $$

We realised last year that sending transactions to Ethereum, reliably and at a fair cost, was a difficult technical problem.

No team had really solved it well. Instead most relayers just looked up ethgasstation, added +50 gwei on top of the estimated fee rate and then sent the transaction. If the network fees spiked during that period, then tough luck, you have a stuck transaction. Even worse, if you wanted to hire a third party relayer, it would add +200k gas to your transaction!

Our goal with any.sender was pretty simple:

any.sender was built to optimistically save users on transaction fees by catching the super-lows while at the same time achieving relatively fast confirmation times and guaranteeing no stuck transactions.

The exercise revealed about how we can improve any.sender, and even further tune it to get lower prices and faster confirmation times. Watch this space for more improvements.