A uniswap V2 clone-agnostic arbitrage bot

This bot is set for use on a private node on Polygon network. RPC nodes hosted by Alchemy or Infura don't have the throughput. You'll be locked/burn though your daily limits in about 10 minutes.

It is set to use Uniswap V2 flash swaps for liquidity across all assets on all exchanges across the network. You can use DeFi Llama to find more UniswapV2 clones.

The current database (/data/v2/matches/) has plenty, but you can run the scipts below to fetch new pairs and filter out dead pairs by volume and timestamp of last trade in each pool.

The bot prefers a private node because proprietary RPCs rate limit and the bot has no rate limit set. This makes trades faster for chains on which you have unrestricted access to a node.

Unless you have access to a private RPC for every chain you want to run it on, a rate limit should be set somewhere.

The bot is designed as a basic framework for defi arbitrage trades, and over time, more DeFi protocols should be added to increase opportunities (Balancer, Aave, Kyber, etc.)

Notes:

I wrote (almost) every line in this bot in an ambitious project to learn Typescript to a highly developed level, as well as several blockchain protocols I was utilizing, as well as how to aggregate and filter data.

By the time I had launched this bot on my dedicated Hetzner server, most of the liquidity from Uniswap V2 had gone and I had to begin work on a Uniswap V3 bot, which is pending, though may not be completed as I'm beginning to consider that the pace of software in that particular space (decentralized finance) moves at a pace fast enough to require a team of developers now. Gone are the days when a solo dev could make a mark on the various pools across the chains.

I recommend Hetzner Robot for your server. I ran this on AWS and burned through about 600 gbp in a month. Hetzner requires more Linux and server hosting knowledge, but it is worthwhile for the steep discount. I ran a bot on Hetzner for roughly 200 gbp/month with all NVME drives.

Moving on...

Issues:

The bot runs and parses all data, but there are still some blocks to clear before it actually sends transactions.

Current issues can be manually copied to ERROR.txt from app.log for ease of viewing as the logs are long.

Commands:

After installing, you should 'apt-get update' & 'npm install'. Also, you should compile the hardhat contracts.

For your IDE I recommend downloading EVM compilers and using local compiling.

I've provided '.env.test' and '.env.live' examples, for which variables need to be set for live/vs test environments as needed.

Test wallet keys can be generated using 'npm run hhnode' which will start a node and create a log for the node output.

Testing operations (detailed below):

1. Start the local Hardhat network with forking enabled.
2. Deploy test contracts to local fork;
3. Run the arbSim.ts script to create the arbitrage opportunity. (Optional, possibly deprecated)
4. Run this bot on the local network.

1. Start v2 bot in test mode:

You'll need to run a hardhat localhost node. You can change output pipe in pacakge.json. You'll need this for troubleshooting solidity contracts. The following starts a hardhat test node by forking the blockchain to localhost and pipes output for troubleshooting. You can change forked block in hardhat-config.ts.

npm run hhnode 

To deploy test contracts on localhost hardhat polygon fork:

In test, package.json commands set NODE_ENV=test and runs bot against localhost. Without specifying environment to cross-env, the bot will not recognize a provider. All logs can be found at ~/logs/app.logs Overwritten each run.

npm run v2DeployTest

To create arb opportunity in test (not always necessary depending on chosen forked block):

npm run arbSim

This empties all wallets generated by local fork into one wallet and trades on an exchange, causing volatility. This state is retained until a transaction runs against it.

Run the bot

npm run v2BotTest 

ex: exclude empty eth calls from hh results:

script -c "npx hardhat node" | grep -vP "eth_call\s+Contract call:\s+<UnrecognizedContract>\s+From:\s+0x[a-fA-F0-9]{40}\s+To:\s+0x[a-fA-F0-9]{40}" > output.txt

Deploying live to Polygon

npm run v2DeployLive

Start v2 bot in live mode:

npm run v2BotLive

Update or get pairs. (The following operation may be time consuming):

Get new pairs for v2 :
	npm run get2

Get new pairs for v3:
	npm run get3

Match v2 pairs
	npm run match2

Match v3 paris
	npm run match3

script -c "npx hardhat node" | tee output.txt

Notes:

I've moved smart contracts to a new repo because neither VSCode, nor Hardhat can handle multiple different versions of solidity docs and compilers, resulting in insane amounts of wasted time.

re: tradeSize // Would be good to have a strategy that takes into account the reserves of the pool and uses the min of the three below, but that adds a lot of complexity.

			Wrong/deprecated:
				// This strategy attempts to use the biggest tradeSize possible. 
				// It will use toPrice, despite high slippage, if slippage creates profitable trades. 
				// If toPrice is smaller than maxIn(for slippage) it will use maxIn.
			According to Uniswap Analysis (see README), AMMs are resistant to large trades, thus the above strategy is flawed. 
			
			Instead this tradesize will ensure that trades are not > reserves, and above, size will be min(maxIn, toPrice)
			
			(Hopefully limiting by target.maxOut is redundant because maxIn works on slippage in the same way, and amountsOut will only calculate a number it will output.)

JETSWAP: Removing Jetswap for now as all trades seem to exit prematurely out of contract without reason.

It would seem like you want to 'buy' the cheaper token, but you actually want to 'sell' the more expensive token.

ex: A: eth/usd = 1/3000 = on uniswap B: eth/usd = 1/3100 = on sushiswap borrow eth on uniswap, sell on sushiswap for 3100 = $100 profit minus fees.

TODO: Make Univ2 clone contracts, but change name to Jetswap, so I can arb them.

Ideas:

Price difference between assets could be used as a 'slippage' parameter, to limit trade sizes to available profit, in-line with the 'optimal arbitrage 2' problem. (Though this may already be the effective result of the tradeToPrice function.)

References:

In depth study on Uniswap Markets particularly, including 'Optimal Arbitrage' algos:

https://arxiv.org/pdf/1911.03380.pdf Breakdown of 'Optimal arbitrage in Uniswap': K = constant product, or x * y = k, where x = reserves of Token0 and y = reserves of Token1 mp = reference/marketPrice (i.e. a price you know you can trade into elsewhere) α = tokenIn β = tokenOut ∆α = an amount of tokenIn ∆β = an amount of TokenOut

Rα = reserves from loan pool represented as loanPool.reserves0 * loanpool.reserves1 = loanPool.constantProduct (Kl) Rβ = reserves from target represented as target.reserves0 * target.reserves1 = target.constantProduct (Kt)

∆β' = loaned amountTokenOut + fee ∆β'- ∆β = profit

Maximize: mp∆α − ∆β (marketPrice * amountTokenIn - amountTokenOut)

Rα = reserves of tokenIn

if ∆α, ∆β ≥ 0 (Rα − ∆α)(Rβ + γ∆β) = k

General article on matrix/graph Arbitrage:

https://github.com/ccyanxyz/uniswap-arbitrage-analysis Note: According to the 'Optimal arbitrage in Uniswap' article, this is completely pointless, as on any robust market like Uniswap, the fees destroy the marginal profits arbitrage is constrained to.

ethers.parseUnits() example in BigInt:

const baseValue = BigInt(10) ** BigInt(18);
const humanReadableNumber = BigInt("10000");

const result = humanReadableNumber * baseValue;

console.log(result.toString()); // Print the result as a string

NOTES FROM UNISWAP DISCORD / CRYPTO RACHEL:

Price Impact Math:

##Based on 10,000 sell token0

orgTotalToken0 = 2000000 orgTotalToken1 = 1000

tokenAmount0selling = 10000

ConstantProduct = orgTotalToken0 * orgTotalToken1

orgPrice = orgTotalToken0/orgTotalToken1

token1change = (orgTotalToken1 - (ConstantProduct / (tokenAmount0selling + orgTotalToken0)))

PricePaidPerToken1 = tokenAmount0selling / token1change

priceDifference = PricePaidPerToken1 - orgPrice

intImpact = priceDifference / orgPrice

PercentImpact = intImpact * 100

Helfpul stuff:

regex to get rid of script/hardhat bumf: ^(.Mined empty block range #)(.)$

19 03 2024 DEVELOPMENT NOTES:

• check if approvals are needed in flashMulti.sol
• make testFunds.ts live version

done - check/change trade.tradeSizes.(loanPool/target.tokenOut) and make it so that you don't have to have the token in wallet for it to cnfigure tradeSize.