⚠️ Personal research and trading journal — not investment advice. The author does not provide licensed advisory services.
I have tested over thirty assumptions against real market data. Most failed. A few were borderline — the confidence interval straddled zero, meaning "probably noise."
One passed cleanly.
The RS line slope. Requiring it to be positive — meaning the stock is outperforming the index, not just in rank but in direction — was the first filter in my entire study to survive bootstrap walk-forward cross-validation with the confidence interval fully above zero.
This is the story of what that means and why it matters.
What the RS line is (and why it differs from RS rating)
The RS rating is a rank. It tells you: this stock is in the top X% of all stocks over the past 12 months. An RS rating of 90 means the stock outperformed 90% of the market. It's useful. It's a snapshot.
The RS line is different. It's a ratio: the stock's price divided by the index. When that ratio is rising, the stock is getting stronger relative to the market. When it's falling, the stock is underperforming — even if the price is going up.
A stock can have an RS rating of 85 but a falling RS line. That means: yes, this stock outperformed most stocks over the past year, but recently it's been losing ground to the index. The RS line is forward-looking in a way the rating isn't.
What I tested: among all stocks that meet the RS ≥ 80 threshold, does requiring the RS line to have a positive slope (rising over the recent period) further improve performance?
The answer is yes — and it's the most robust positive finding I've produced.
How I tested it
The standard problem with testing filters: most filters appear to work when tested on historical data but fail on new data. The data you trained on is already "contaminated" — you saw the outcome before you chose the test.
Walk-forward cross-validation is the honest version of backtesting. You test the rule on data that was genuinely unknown at the time the rule was applied. Each test window uses only information that would have been available historically, then measures what happened next.
Bootstrap resampling adds another layer: it asks "how stable is this result?" by generating thousands of simulated market histories using the same price data, re-running the test on each, and building a confidence interval. If the true effect is zero, a bootstrap CI centered near zero should straddle it — the CI should include zero. If the CI is entirely above zero, the signal is real.
The RS line slope requirement — requiring that the RS line be trending upward at the time of the breakout scan — passed this test. The confidence interval for the 30-day forward return improvement was entirely above zero: +1.09 percentage points, CI excludes zero.
In thirty-odd tests, this was the first time that happened.
What the result means practically
Starting from the RS ≥ 80 baseline, adding the RS line slope requirement:
| Baseline (RS ≥ 80 only) | With RS line slope > 0 | |
|---|---|---|
| 30-day forward return | Positive, CI includes 0 | CI excludes 0 |
| Improvement | — | +1.09pp |
| Interpretation | Signal, borderline | Signal, confirmed |
The improvement is not massive — 1.09 percentage points. But in a trading system where most assumptions fail entirely, finding one that survives the hardest test is significant. It identifies a real characteristic of the market.
The mechanism is plausible: a stock with a rising RS line is actively outperforming right now, not just historically. Momentum has a directional component. A rank tells you "was strong," but the slope tells you "is strengthening."
What I didn't test
I want to be precise about the scope of this finding.
This was tested on Thai stocks (SET/mai universe, 1990–2026). I have a parallel result for US stocks that is consistent in direction but was not bootstrap-validated with the same rigor. I'm reporting what I confirmed, not what I expect to be true.
The improvement is at the 30-day horizon. At shorter horizons (5d, 10d), the effect is noisier — the RS line slope matters more for medium-term continuation than for immediate post-breakout performance.
This is a signal filter, not a complete system. Adding RS line slope >0 to a baseline that already has flaws doesn't fix the flaws — it improves a valid baseline. If the underlying method doesn't have edge, no filter will rescue it.
How it changes my scanner
Prior to this finding, my scanner showed any RS ≥ 80 stock with a valid breakout pattern. The RS line was visible in charts but not enforced as a gate.
Post-validation, RS line slope >0 is a hard gate. Stocks that are ranking highly but losing ground to the index are filtered out before they reach the shortlist. The candidates that remain are both highly-ranked and actively strengthening.
This is the kind of methodological evolution that only comes from testing. You can't eyeball your way to confidence intervals. The pattern looks obvious in hindsight — of course you want a stock that's outperforming — but "obvious in hindsight" and "confirmed in data" are not the same thing. Before this test, I couldn't have told you whether requiring a positive slope would improve or degrade outcomes. Now I can.
The larger lesson
Most of what traders believe about filters is untested. Rules get passed down from one practitioner to the next, each one adding their confidence to the claim without verifying it. A rule that "worked" in one market, one decade, one practitioner's experience becomes doctrine.
The RS line slope is one of the few rules that held up when I actually looked. That makes it worth treating differently than the others.
Track. Study. Wait. Strike.
Personal research and trading journal — not investment advice. The author does not provide licensed advisory services. — MOEasymmetry
Draft 2026-06-11. Source: bootstrap walk-forward validation, Thai SET/mai universe 1990–2026. #RSL1 finding (2026-05-14). CI at 30d horizon: +1.09pp, CI excludes 0.