Getting Started — Part 2: What Is Algorithmic Trading? (A Non-Coder's Take)

When people hear "algorithmic trading," they picture hedge fund servers in Manhattan, PhDs in math, screens full of candlestick charts. Something serious and intimidating and definitely not for people like me.

My version looked a little different. For the first few weeks, my trading bot ran on my MacBook. Every time I closed the lid — lunch break, coffee run, falling asleep on the couch — the bot died. No trades, no monitoring, nothing. That problem lasted weeks before I moved everything to a cloud server.

This post is for anyone who's heard the term "algorithmic trading" and assumed it wasn't for them. I thought the same thing. Turns out the concept is simpler than it sounds — the hard part is everything else.

What Does Algorithmic Trading Actually Mean?

Strip away the jargon and it's just this: instead of a human deciding when to buy and sell, a set of rules makes that decision automatically.

That's it. The rules can be simple — "if the price drops below $10, buy" — or complicated enough to fill a textbook. But at the core, it's always the same idea. You define conditions. The code checks those conditions. If they're met, it acts.

The reason people do this isn't speed or sophistication. It's emotion. At 2 AM, staring at a position that's down 15%, your brain starts negotiating. "Maybe it'll bounce back." "I should double down." "I'll just hold a little longer." A bot doesn't do that. It follows the rules you gave it, every time, whether it's 2 PM or 2 AM.

That sounds like a huge advantage. And it is — until your rules are bad. Then the bot follows bad rules with perfect discipline, 24 hours a day. I learned that the hard way.

Why Would Anyone Bet on Tomorrow's Temperature?

Before I explain how the bot works, you need to know where it trades. Polymarket is a prediction market. People bet on the outcomes of real-world events — elections, sports, economic data, and yes, weather.

Here's how a weather market works. Polymarket asks: "Will the high temperature in New York exceed 60°F tomorrow?" You can buy YES or NO shares. If you buy YES at $0.15 and the temperature ends up above 60°F, your share pays out $1.00. That's a $0.85 profit on a $0.15 bet. If it doesn't, you lose your $0.15.

So why weather? Three reasons that made it click for me.

First, the data is free. Professional weather models — the same ones governments use — are publicly available through APIs like Open-Meteo. You don't need to pay for data or build your own forecasting system. The models already exist.

Second, the outcome is objective. There's no ambiguity. The actual temperature gets measured by weather stations, and that number decides who wins. No interpretation, no controversy.

Third, new markets open every single day. Every city, every day, a fresh set of bets. That means the bot always has something to evaluate.

Does this actually work at scale? There's an account on Polymarket called gopfan2. $1.48 million in verified profit from weather markets. A 60.8% win rate across 1,700+ positions. The PnL curve isn't a lucky spike — it's a steady climb over 18 months. That's what made me think automating this was worth trying.

How a Bot Decides What to Buy

No code in this explanation. Just the logic.

My bot runs through a loop every day for each city it monitors. Step one: it asks three different weather models — GFS, ECMWF, and ICON — "what will the high temperature be tomorrow in Ankara?" Each model gives a slightly different answer.

Step two: it checks how much those answers agree. If all three say 75°F, the bot is fairly confident. If one says 68°F and another says 82°F, something's off — stay out.

Step three: it looks at what the Polymarket price implies. If YES shares for "above 73°F" are trading at $0.15, the market is saying there's only a 15% chance. But if all three models predict 78°F, the bot sees a gap between what the models think and what the market thinks.

Step four: if the gap is big enough, buy.

That's the entire strategy. Models predict. Market prices. Bot compares. Gap exists? Buy. No gap? Skip.

Simple, right? In theory. In practice, I ran this exact logic for my first week and went 0-7. Seven trades, seven losses. Every actual temperature came in higher than my forecast, by a consistent margin. Something was systematically wrong with how I was reading the model data.

What I Got Wrong About Automated Trading

Before I started, I had this vague image of setting up a bot, turning it on, and watching money trickle in. The "automation" part felt like the hard work was front-loaded — build it once, collect forever. That turned out to be completely wrong, and I want to be specific about why.

The first thing: a bad strategy doesn't get better just because it's automated. It gets worse. My bot placed 163 trades in its first month. 125 of those were losses. It was executing the strategy perfectly — the strategy just wasn't good enough. The bot didn't know that. It doesn't have opinions. It saw a signal, it bought. Over and over.

The second thing: backtesting lies to you. I found two filters in the historical data that turned a 23% win rate into a 32% winner with +44% ROI. On paper, incredible. I shipped both filters at once. The very first live deployment had issues. Historical data is clean. Live markets are not.

The third thing: "automated" doesn't mean "unattended." Polymarket changed its fee structure for weather markets. A trade showed up in my wallet that my bot didn't place. A model that worked great for one city was terrible for another. The market shifts underneath you, and if you're not watching, the bot keeps trading on outdated assumptions.

Automated trading doesn't mean you stop paying attention. It means you pay attention to different things — strategy performance, edge decay, infrastructure, rule changes. The bot handles execution. You handle everything else.

That's the basics. If you want to see how all of this plays out in practice — the good trades, the bad trades, the pivots — the Weather Bot series starts here and runs for fourteen episodes so far. Part 3 of Getting Started covers something different: AI agents, how they work on-chain, and why they're not the same thing as bots.

More updates on the way. If you're working on something similar or found a smarter way to do it, drop it in the comments — the more we share, the faster we all move.

Disclaimer: This blog documents my personal learning journey. Nothing here is financial advice.

← Previous: Getting Started — Part 1: What This Blog Is About   |   Next: Getting Started — Part 3: What Are AI Agents? →