Eight years of jumping between strategies. That's the whole origin story.

I’m Kevin Colbeck. I’m 24, I’ve been trading for more than eight years, and I’ve been writing software for most of that time too — five-plus years building my own indicators and trading algorithms, long before any of it looked like a company.

For most of those eight years I was doing what a lot of traders do: jumping from strategy to strategy. Read about a setup, try it, catch a couple of good weeks, hit a bad stretch, wonder whether the edge was ever real or whether I’d just been lucky — and move on to the next idea without ever really answering the question. Every jump felt like progress. Almost none of it was.

Eventually I started answering the question with code instead of with feelings. That was better, and it was still brutally slow. To prove a single strategy honestly you have to state the rules exactly, get clean historical data, build a simulator that doesn’t cheat, account for slippage, resist tuning it until the curve looks good, and then — the part almost nobody does — wait and watch it perform on data you didn’t have when you built it. Months of work per idea. I had a hundred ideas.

So I built the thing I needed

Chat·Backtest is that whole process compressed into seconds. Describe a strategy the way you’d explain it to a friend. The AI turns it into exact, inspectable rules — no black box, you can read every line and export it. A real engine runs it across two decades of market data and shows you what actually happened, including the parts you were hoping not to see.

The thing I care most about is the second half, because it’s the half the industry skips. Anybody can produce a beautiful backtest; with enough attempts, so can random noise. That’s why every strategy here can be frozen and deployed to a public forward-test ledger— the rules get locked and hashed, and from that day forward the strategy scores itself on data nobody had when it was written. The record is append-only. We can’t edit it, and neither can you. Losers stay on the board on purpose.

That’s the whole philosophy, and it’s deliberately the opposite of how this industry usually works. They sell the dream. We sell the proof.

What this is not

I want to be blunt about this, because the space is full of people who aren’t. Chat·Backtest does not send signals, does not tell you what to buy, does not predict prices, and does not touch your brokerage account. There is no “autopilot.” It is a research and education tool for testing ideas on historical data and keeping an honest record of how they hold up. What you do with what you learn is yours, and the risk is yours too.

I also don’t hide the ugly results. Some of the strategies in our own library lost money, and we published them with the numbers attached — including a strategy of ours that failed. A library where everything wins is evidence of curation, not skill.

Where this is going

The near-term goal is to make proving an idea completely frictionless: more markets and timeframes, deeper trade-by-trade forensics, and exports that let you run your strategy anywhere you like — it’s your work, not ours to lock up.

The longer-term goal is bigger, and it’s the reason the ledger exists at all. Right now the only way to judge someone’s trading is to trust their screenshots. I want a place where a track record is a timestamped, independently verifiable, publicly auditable object — where “I called that” can be checked by anyone in about four seconds. If that becomes normal, a lot of what the retail trading industry currently gets away with stops working. That seems like a good outcome.

Get in touch

I read everything. If something here is broken, wrong, or misleading — especially if you think a number is misleading — I want to know. Email kevin@chatbacktest.com.

We post new records and write-ups on X: