My First Strategy Was Garbage — the Backtest Told Me So

I found the Strategy Tester tab about a week after I found Pine Script. That week was expensive to skip.

I'd built an EMA 5/15 crossover strategy with Gemini's help, wired it to fire a webhook, and I was ready to go — right up until I discovered TradingView has a Strategy Tester tab that runs your strategy against history and tells you, in cold numbers, what would have happened.

Mine lost money. Not dramatically — it wasn't a strategy that blew up an account — it just steadily gave back what it made, trade after trade, the way a strategy does when it has no actual edge. A crossover that catches every whipsaw in a sideways market and gets stopped out for a small loss each time, and only occasionally catches the trend it was designed for.

The part where I started tuning

So I did what felt like the obvious next step: I started changing the inputs. Made the fast EMA a bit faster. Made the slow one a bit slower. Added a filter. Removed it. Tried a different timeframe. Every change, I'd re-run the backtest and look at the equity curve.

At some point — I couldn't tell you exactly which change did it — the equity curve turned from a slow bleed into a climb. Profit factor above 1.5. Win rate over 50%. It looked, on the chart, exactly like a strategy that worked.

I had found a strategy that worked on the past. I had not yet learned that this is a different claim from a strategy that works.

I went live on Delta Exchange with that strategy, webhooks firing straight into real orders. And for a while, the discipline problem really was solved — the computer executed exactly what the backtest said to execute, with none of my hesitation. What I hadn't solved was a problem I didn't know existed yet: I'd built a strategy that was tuned to that specific stretch of history, not one that had a genuine edge in it. The losses that followed, and the term I eventually learned for what I'd actually done, are the next post.

Follow bots built the way this post describes

Walk-forward tested, forward-tested on paper, then run with real risk limits. 15-day free Platinum trial, no card required.