⚠️ Educational research and a personal trading journal — not investment advice. การศึกษาเท่านั้น ไม่ใช่คำแนะนำการลงทุน. The author does not provide licensed advisory services. Investing involves risk; consult a licensed adviser before acting.
Every trader believes their settings are almost right — that if they just tightened the stop, or filtered a little harder, the system would get better. So I did the honest thing: I took my own breakout method and tried to break every assumption in it. Decades of data, one knob at a time.
What I found is the most useful thing I can tell you about trading systems — and it's not a better setting. It's a warning about how easily you'll fool yourself.
The setup My method has maybe six tunable knobs: how wide the stop is, how close to the breakout you buy, where you take partial profit, how tight the base has to be, whether you demand a volume surge, and so on. I swept each one across decades of trades and measured the edge (net return per trade, after costs).
I was hoping to find a few free upgrades. At first, it looked like I had.
The seductive part Several tweaks made the number go up. Tighten the stop — better. Demand the pivot be near the highs — better. Require volume to dry up as the base forms — better, and the win rate jumped too. Stack a few of them together and the edge nearly doubled.
If I'd stopped there, I'd have rewritten my system on the spot. A lot of people would have — and would have started trading it, and posted the backtest as proof.
But a doubled edge from stacking filters is exactly what a fooled trader sees. So I ran the test that tells the truth.
The test that kills illusions: go year by year A single big number across 30 years can hide everything. So I broke every "improved" system down year by year and asked one question: does it beat the original in most years, or just in a few?
The answer was brutal. That nearly-doubled "super-system"? Its huge average came from a handful of explosive bull years — one of them booked a massive return on just six trades. In the typical year, it was actually worse than the original. The filters hadn't found an edge; they'd shrunk the sample down to a few lucky names and called the luck a strategy.
The "volume dry-up" filter — which I believe in, which Minervini teaches, which you can see on the chart — told the same story: gorgeous average, but it beat the baseline in only half the years. A coin flip dressed as an edge.
One by one, the upgrades died in the year-by-year test. The number went up; the consistency didn't. That gap is the signature of overfitting, and it's the single most common way traders blow themselves up with their own research.
The one that lived Exactly one change survived: requiring a genuine surge of volume on the breakout day. It improved not just the average but the typical year — better in roughly two-thirds of all years, across both the early decades and the recent ones. It held up because it has a reason to work: a breakout on real volume is real demand showing up, not a quiet drift over a line.
And here's the part that made me laugh: the volume surge is something I was already doing by eye. The data didn't hand me a new edge. It confirmed the one instinct I already trusted — and exposed every "clever" addition as a mirage.
(One honest footnote: this held on my home market. When I ran the same test on US stocks, the volume-surge rule didn't help — there, institutions accumulate quietly and good breakouts don't always shout. Same lesson, opposite detail: never assume an edge from one market transfers to another. Test it where you trade it.)
The lesson worth more than any setting 1. Your settings are probably fine. I went in expecting to find mistakes. I found that the system was already well-built — and that almost every "improvement" was noise. 2. A bigger backtest number is not a better system. If the average goes up but the year-by-year consistency doesn't, you've overfit. Always check the typical year, not the headline. 3. Trust changes that have a reason, not changes that just score higher. Volume-on-breakout works because of why it works. The stacked filters scored higher because of luck. 4. It's a discipline game, not a parameter game. The deepest finding: no setting changed the method's fundamental character — it wins in about 55% of years and makes its money in the tail, no matter how I tuned it. You don't win by optimizing. You win by executing.
I spent days trying to improve my system. The system's answer was: stop fiddling, and go follow the rules you already have.
That's the least glamorous conclusion in trading. It's also the true one.
Track. Study. Wait. Strike.
Personal research and trading journal, not investment advice. The author does not provide licensed advisory services. Markets carry risk; your decisions are your own. — MOEasymmetry