How to build a winning strategy
A repeatable method in six steps: start from a hypothesis, keep the filters few, demand a large sample, and let closing line value be the judge.
Open the builder, add a filter, watch the curve. Add another, watch it climb. An hour later you have a beautiful backtest and no idea whether any of it is real. There is no secret filter that prints money. But there is a method that separates real edges from lucky-looking noise, and it starts before you touch a single filter.
Step 1: Start with a hypothesis, not a filter
Don't toggle filters until the curve goes up. Start with an idea you can say out loud: "Draws are underbet in low-scoring leagues", "Big favourites are overbet on weekends". A hypothesis you can explain is one you can trust.
Step 2: Translate it into the fewest filters possible
Every filter you add is a chance to fit noise. Aim for two to four filters that directly express your hypothesis. If your idea is "home underdogs in top leagues", that's a league filter and an odds range, not ten conditions.
Step 3: Demand a large sample
Run the backtest and check the number of bets first, before the profit. A glowing yield over 60 bets is nearly meaningless. Look for hundreds or thousands of qualifying bets, so that variance has had time to average out.
Step 4: Read the P&L curve, not just the total
A good strategy climbs steadily. A curve that stays flat and then jumps in a single month is usually one fluke result carrying the whole thing. Zoom in and ask: would the strategy still be positive without its best week?
Step 5: Guard against overfitting
If you tuned the odds range from 1.90 to 1.94 because it "looked better", stop. That's curve-fitting. The honest test is out-of-sample: does the edge survive on data you didn't optimise on?
Step 6: Let CLV be the referee
Once live, the real scoreboard isn't this week's profit. It's whether your picks beat the closing line (CLV). Positive CLV over time means the edge is real, even during a losing streak.
The mindset
Most strategies you try will fail these tests, and that is the process working. The goal isn't to force one to pass; it's to keep only the ones that genuinely do.
Continue learning
Browse all articles
VideoBacktesting without fooling yourself
The backtest is the most powerful and the most dangerous tool you have. How overfitting sneaks in, and how out-of-sample testing keeps you honest.
VideoAnatomy of a strategy
What a strategy is actually made of: one sport, one bet selection, one filter tree. How groups combine on the canvas, and what gets frozen the moment you hit Save.
Reading a backtest: every metric explained
The backtest panel shows a dozen numbers and most people read one. What each metric measures, what good looks like, and the traps that make bad strategies look brilliant.