Building a Polymarket Copy Bot
Startup 1 of 12 Startups in 12 Months
My New Year's Resolution was to take more risk. For the month of January, I did so by building a Polymarket bot. $1000 in. $865 out by the month’s end. Below are the obstacles I tackled, what I wish I’d have done differently, and what I’d do to invest better.
Quick Intro to Polymarket
Polymarket is a crypto-native platform that lets you bet on real life events. You can either buy shares of Yes or shares of No. When one side wins, the capital from the other bet is transferred over to you. Here’s an example market:
Yes is at 9 cents and No is at 92 cents. This means that if I buy $9 worth of Yes, I’d get back $100 if the US confirmed that aliens exist before 2027.
And if I buy $92 worth of No, then I’d get back $100 if the US confirmed that aliens exist. The less likely you are to win, the more money you have to gain. Very straightforward.
You don’t have to wait for a market to resolve (ie. for the outcome to be decided) to sell your shares. Let’s say I buy $10 worth of Yes and bet that the US will in fact confirm that aliens exist before 2027. If Yes then moves to 18%, then the price of Yes would become $0.18 from $0.09, and the value of shares would become $20 from $10. At any time, I can exit my position by selling.
I first tried out Polymarket by buying equal amounts of both Yes and No. In theory, I shouldn’t lose any money by doing so. The gains from one side would come from the other side’s losses. That was the idea at least. I bought $10 worth of Yes (at $0.89) and $10 worth of No (at $0.13) from this Avatar market. Then when I could, I took profits. I ended up selling everything for $28.65, making $8.65 in profit (43% returns). Not bad! So I did it again with another market, making ~ $8 dollars in profit:
Buying both sides is called “market making”. For some reason, I concluded that this should end. Maybe it was too hands on. Maybe it was too risky. Earlier I thought I had lost money. Anyway, I decided that market making wasn’t for me.
What would be better? Find someone doing much better and copying them. I stumbled upon this tweet that looked for accounts with both a positive growth rate and little volatility. I tried to build a copy trading bot to copy them. Enter the problem.
The Litany of Problems I Encountered
The first problem I had was actually building the bot. ChatGPT refused to write code that let me gamble. Annoying. So I went to the one LLM known for its free speech - Grok. And it worked! It wrote the code for me.
Next, I had to figure out my environment variables. Apparently, Polymarket’s documentation is not up-to-date, so I relied on Grok and ChatGPT to write the code that would prompt PolyMarket’s API for my account variables and also for my Metamask’s variables. I needed to find things like my private key, my secret key, api key, passphrase, MetaMask wallet address, and wallet address. What all of these things are exactly? Couldn’t tell you. It took a bunch of back and forth with Grok and Polymarket’s API before I got it to start working. Finally.
My bot couldn’t copy all trades. The account I copied was buying thousands of dollars a week of positions. I had just $1000. The cash ran dry after a few hours. Not only that, but the API doesn’t let you sell positions worth less than a dollar. I rewrote the code to limit all buy orders to a max of $30, and raise all sell orders less than $1 to be $1. Then I let it run again.
It was thrilling to see that I had made 8% in one day. But every time I shut down my laptop for the day, it stopped working. I was missing a lot of buy orders and a lot of sell orders while my computer was off. Timing was key because this account I was copying rarely held positions till they were closed. He’d buy and sell after making a small profit, and did so non-stop. I couldn’t miss out on these, or else I’d have positions that would sink and go to zero. Not good.
But every time I uploaded my code to Railway to run, Cloudflare would block my bot because it thought it was bot activity. Argh! I thought switching to another platform (pythonanywhere.com) would fix this, but it didn’t - the same error would pop up. Grok also suggested a VPN, but I had lost trust because each new platform would charge me more. Fortunately, a software engineer I know who works at a trading firm confirmed that a VPN was exactly what I needed, so I installed NordVPN via the command line after moving to a third platform, Digital Ocean, which a friend recommended.
Well, after doing all of that, my original ~$950 went down to $840 where it hovers right now. I’ve lost 15% of my money. Was it worth this? Absolutely. I’ve learned so much. What are the next steps?
It’d be to find better accounts to copy. It’d probably be better to copy 5-10 people who are known for 100x wins in diverse fields than to try and find multiple stable earners to copy. That way, the losses even out but the wins are immense. I could also try finding industry insiders who I can copy, like people who trade on the price of eggs because they are farmers.