49,172 Real Bets and a 1,350-Unit Loss — Why I Kept the Data
There was a time when I would have looked at a trading strategy losing 1,350 units and thought there was only one sensible thing to do.
Bin it.
Instead, I kept every bet.
The Master Collector was an automated research strategy that placed 49,172 actual tiny-stake bets into the live Betfair market. These weren't historical results dropped into a backtest. Real bets were placed, at deliberately tiny stakes, and the results were collected.
And overall, they were terrible.
After almost 50,000 bets, the dataset was approximately −1,350 units.
But somewhere along the way, I stopped looking at it simply as a losing strategy. I started looking at it as 49,172 pieces of evidence.
What happens if we separate different conditions? Are some parts consistently worse than others? Are there areas where the behaviour changes? What should we reject? What deserves more testing?
So I started analysing, segmenting and refining.
Most ideas went nowhere. Some looked promising for a while and then fell apart. That's research.
At the Framework V1 checkpoint in August 2026, Dog Pound had reached +466.54 units after 2,100 actual live bets, after commission and using level staking.
That's why I kept the losing data.
Not because losing £1,350 units was secretly a success. It wasn't. But because a failed hypothesis and useless data are not necessarily the same thing.
Sometimes the thing that tells you where not to look helps you discover where you should.
The relationship between those two datasets is the research.
Want to see how we approached it? Explore the Trade Carefully Research Framework and download the free sample:
Sports trading involves financial risk. Historical performance does not guarantee future results.
Comments