One rule, written down before any test, run on 5 markets, each with three-touch lines (the main rule) and with any two-touch line: 10 tests. 5 of 10 averaged above zero out of sample, none graded promising. Every test is shown below with equal weight.
The trend line strategy is taught by the trader and YouTuber Tori Trades, in her own public videos and blog. This page tests IDFX’s mechanical reading of what she teaches: her rules, made exact so a computer can run them. She has not seen or endorsed this test. She trades by eye, rates her lines and can skip a setup, so this tests the rules, not her.
All 10 rejected. 5 of 10 averaged above zero out of sample, about what chance gives when nothing works, and none did better than the same rules on reshuffled price. Before trading costs. Every test below has equal weight.
Her method, as she teaches it: price breaks a trend line, take the trade in the direction of the break (the action line); keep the stop on the other side of the opposing line (the safety line) and trail it along that line; no take-profit. Her A+ setup is a line with three touch points. These are the rules as written down and committed on 27 October 2026, before any test was run.
Start at the highest swing high of the last 13 weeks. Touch as many swing highs as possible, never cut through one, and start each steeper line where the last one ended. The newest, steepest line is the one traded.
The same on the swing lows, from the lowest low.
A 4-hour candle closes above the falling line (long) or below the rising line (short). The line needs at least three touch points (two in the variant), price must not already have closed through it, and the other line must exist with price on its safe side. Fill at the next bar’s open.
The opposing line is the safety line. The stop sits a quarter of an ATR beyond it and moves along it at every close, never loosened.
No take-profit and no time limit: the trade stays on until the stop is hit. One trade at a time, 1% risk.
Her three drawing rules (connect the lines, touch as many swings as possible, never cut through price) plus “the previous point B is the new point A” describe exactly one outline: the one a rubber band makes when stretched over the swing highs, known as the convex hull. That is how the engine draws her lines, so there is no judgement left in where a line goes. Both diagrams are illustrations, not data.
10 tests were run (5 markets x three-touch and two-touch lines). With that many, one can look good by chance alone even if none has an edge. All are listed with equal weight; read any positive result as '1 of 10'.
Each curve is the test windows stitched end to end, in R, on the same time axis. "Windows up": test windows with a positive average. "Shuffled price as good": how many of the engine's reshuffled price series did as well or better; the fewest for any of the 10 tests was 2.
| Test | Touch points | Trades | Win rate | Per trade | Total | Profit factor | Windows up | Shuffled as good | Verdict |
|---|---|---|---|---|---|---|---|---|---|
| Platinum · 3+ touches | 3+ (main) | 82 | 39.0% | −0.012R | −1.0R | 0.92 | 2 of 5 | 39 of 40 | Not this time |
| WTI oil · 3+ touches | 3+ (main) | 79 | 36.7% | −0.021R | −1.6R | 0.92 | 2 of 5 | 17 of 40 | Not this time |
| Gold · 3+ touches | 3+ (main) | 60 | 36.7% | +0.108R1 of 10 | +6.5R | 1.24 | 3 of 5 | 33 of 40 | Not this time |
| US30 · 3+ touches | 3+ (main) | 64 | 28.1% | −0.099R | −6.3R | 0.76 | 3 of 5 | 35 of 40 | Not this time |
| NAS100 · 3+ touches | 3+ (main) | 56 | 42.9% | +0.198R1 of 10 | +11.1R | 1.54 | 5 of 5 | 2 of 40 | Not this time |
| Platinum · 2+ touches | 2+ | 310 | 31.6% | −0.139R | −43.0R | 0.70 | 0 of 5 | 40 of 40 | Not this time |
| WTI oil · 2+ touches | 2+ | 307 | 36.2% | +0.126R1 of 10 | +38.5R | 1.30 | 4 of 5 | 5 of 40 | Not this time |
| Gold · 2+ touches | 2+ | 267 | 33.0% | +0.260R1 of 10 | +69.4R | 1.65 | 5 of 5 | 5 of 40 | Not this time |
| US30 · 2+ touches | 2+ | 315 | 30.5% | −0.078R | −24.7R | 0.82 | 2 of 5 | 17 of 40 | Not this time |
| NAS100 · 2+ touches | 2+ | 276 | 35.1% | +0.091R1 of 10 | +25.0R | 1.17 | 3 of 5 | 11 of 40 | Not this time |
The 3+ rows are her A+ setup, the main rule; the 2+ rows are the same rule taking any line with two or more touch points. No target in any of them. Read any positive row as 1 of 10. Every row is out of sample and before trading costs.
The closest call: every test window positive, and reshuffled price did as well only 2 times in 40. It needed 1 or fewer, and it is 56 trades and one of ten tests. The longs were flat; the shorts made it.
The biggest total. The longs made it while gold roughly doubled over the test windows, and the shorts lost slightly. A trend rule that earns on the long side of a market that went up is exactly what the shuffled-price check looks for: 5 of 40 reshuffled runs did as well.
Shown because they came closest, not because they passed: both were rejected, and with ten tests one or two are expected to look good by chance. Long and short splits from each run’s results file. Gold’s rise: about $1,960 to $4,016 on this price feed across the test windows.
Real out-of-sample trades from the main-rule tests, with the lines as the engine drew them on the signal bar: touch points circled, the stop trailing along the safety line. Which trades: the most recent main-rule trade of each kind (long / short, a full 1R loss / a stop trailed into profit) with a first stop of 1 to 3 ATR; to show the rules, not the result.
Nobody approved sample trades before the runs (the owner chose to review afterwards), so every trade was checked without the engine instead. A separate script redrew the lines from the raw 4-hour bars by another method and walked each out-of-sample trade bar by bar against the pre-registered rules.
The free TradingView indicator replays every trade of all ten tests identically. The strategy export matches too, except on 3 trades whose entry opened beyond the safety line after a weekend gap. The three-touch lines look like hers; the two-touch lines are often short and steep, which is why three touch points is the main rule.
All 10 tests in one file, published as it was written, with the engine run behind each row. Each main-rule test also has its own page and file.