The Nifty Gas Station gives users the ability to mint NFTs at optimally low gas prices, thus saving dozens of dollars in fees and lowering the barrier for any artist to put their own creation on the Ethereum blockchain.
We believe this offers a permissionless, non-rent-seeking and decentralized mechanism for artists and creators to port their work into the metaverse.
Moreover, The Nifty Gas Station saves minters money and time. Here's how:
Rather than mint the NFT immediately, users send the NFT smart contract, metadata, as well as a pre-determined amount of ETH, to our oracle that will add it to the queue. A machine learning algorithm running off-chain will track gas prices and trigger the oracle to perform a batch transaction, minting the entire queue of NFTs, when gas prices fall beneath the target gas price.
While the Nifty Gas Station uses the gas fee prediction algorithm as a tool to plan batch mints, we see its potential to emerge as a novel and useful "lego" in the Ethereum development stack.
Many artists complain NFT minting fees (i.e. gas fees) fluctuate too much during the week and cause them huge economic losses in the long run. A NFT could be minted in the range $15-20 one day while the next day go up to $60-70 range.
It all starts with someone sending a transaction to our smart contract with the following data:
[optional] refund address
... along with with enough gas fees for minting.
Using a SARIMAX statistical model and hourly average gas fees from Etherscan a prediction on the trend of gas fees is made. Then, according to whether predictions show an uprising or downgoing trend, an Oracle is invoked to set the IsTimeToMint flag to
The smart contract will then batch mint all queued data in a single transaction using the ERC2309 standard built on top of the ERC721 token, mostly used for NFTs.
The resulting ERC721 will be transferred to the legitimate owners and any extra gas fee returned.
By combining batch minting with gas fee prediction we estimate to save on gas fees by as much as 20%.
The machine learning algorithm is still in an early stage and predictions may fluctuate during the day causing spurious activations of the smart contract. We aim at stabilizing predictive outcomes by using a more reliable autocorrelation model introducing elements like Google Trends and ETH/USD price ratio.