Part 4: The system under test · Chapter 11 of 14 · 8 min
The whole system tested: five versions and a comparison with chance
The steps of chapters 3-9 put together in five versions of the system, with two real weeks of signals and the measurements: D1 trend plus right half gives +0.35 ± 0.30 R over 25 trades (2.6 a month), the two-out-of-four version −0.06 ± 0.14 over 104. A random subset of the same size does as well 4% of the time for the first, 34% for the second.
The five versions
The system has four filters (D1 trend aligned, right half of the range, a level taken just before, structure break) and a fixed order (limit at the near edge, stop 0.25 ATR beyond the far side, 2R, cost 0.30 dollars). The versions are ways of combining the filters.
For comparison with chance, for each version we draw 10,000 random subsets of POI of the same size from the 291 and count how often the mean R is at least equal to the version's.
Two real weeks
Figure 1 shows the signals of version B in the two weeks from 1 to 17 September 2026, on H4: order blocks and FVGs with the D1 trend aligned and in the right half. There are nine, and they come from six moves (three times an order block and an FVG overlapping).
- 2 September, demand FVG 4,337.96-4,369.63 dollars: the order fills on 8 September and touches the stop on 10 September (R −1.01).
- 7 September, supply order block 4,431.00-4,439.50: stop on 8 September at 00:00 (R −1.03). The same day a supply FVG 4,416.39-4,420.49 touches the stop in the same candle as the fill (R −1.04).
- 8 September, supply order block 4,418.93-4,444.09: target on 10 September (R +1.99). A supply FVG 4,411.63-4,418.93 reaches the target at 4,385.48 in the 16:00 candle (R +1.98).
- 10 September, supply FVG 4,382.86-4,408.83: it fills on 11 September, target at 4,313.89 on 14 September (R +1.99).
- 13 September, supply order block 4,347.01-4,360.27: target on 14 September at 04:00 (R +1.98).
- 16 September, supply order block 4,330.53-4,368.12 and supply FVG 4,282.03-4,330.53: two stops on 17 September (R −1.01 and −1.01).
Four targets and five stops: +1.84 R in all over the nine trades. Counting moves with two overlapping zones once, the cases are six.
Read the chart in words
H4 gold chart from 1 to 17 September 2026. Nine rectangles mark the order blocks and FVGs, one demand and eight supply, each with a label at the exit of the trade: five stops at −1R and four targets at +2R.
| When | What | Price |
|---|---|---|
| 10 Sept 2026 16:00 UTC | −1R | 4,319.39 |
| 8 Sept 2026 00:00 UTC | −1R | 4,444.09 |
| 7 Sept 2026 08:00 UTC | −1R | 4,426.89 |
| 10 Sept 2026 12:00 UTC | +2R | 4,382.86 |
| 8 Sept 2026 16:00 UTC | +2R | 4,407.95 |
| 14 Sept 2026 04:00 UTC | +2R | 4,343.74 |
| 14 Sept 2026 04:00 UTC | +2R | 4,343.74 |
| 17 Sept 2026 12:00 UTC | −1R | 4,387.02 |
| 17 Sept 2026 04:00 UTC | −1R | 4,341.45 |
Times in UTC.
Figure 2 shows a signal of B that loses, on H4, from 25 to 27 March 2026. The supply FVG between $4,459.72 and 4,497.79 (0.52 ATR, ATR of 72.7 dollars) has the D1 trend aligned and sits in the right half. The order fills on 26 March at 12:00 at 4,459.72, the stop is at 4,515.96 (risk 56.24 dollars) and the target at 4,347.25. On 27 March at 12:00 price touches the stop: R −1.01.
Read the chart in words
H4 gold chart from 25 to 28 March 2026. A rectangle marks the supply FVG between 4,459.72 and 4,497.79. Horizontal lines mark the entry at 4,459.72, the stop at 4,515.96 and the target at 4,347.25. A label marks the stop hit.
| When | What | Price |
|---|---|---|
| 26 Mar 2026 04:00 UTC | Supply FVG | 4,459.72 – 4,497.79 |
| 26 Mar 2026 12:00 UTC | Entry | 4,459.72 |
| 26 Mar 2026 12:00 UTC | Stop | 4,515.96 |
| 26 Mar 2026 12:00 UTC | Target 2R | 4,347.25 |
| 27 Mar 2026 12:00 UTC | stop hit | 4,556.63 |
Times in UTC.
What the data says
| Version | POI (a month) | Trades (a month) | Targets | Mean R (2R) | Random subsets with R at least equal |
|---|---|---|---|---|---|
| A. Baseline | 291 (30.4) | 216 (22.5) | 62 | −0.10 ± 0.09 | 95% |
| B. Trend and half | 37 (3.9) | 25 (2.6) | 11 | +0.35 ± 0.30 | 4% |
| C. B plus take | 21 (2.2) | 13 (1.4) | 6 | +0.46 ± 0.42 | 5% |
| D. Two out of four | 145 (15.1) | 104 (10.8) | 31 | −0.06 ± 0.14 | 34% |
| E. Three out of four | 35 (3.7) | 22 (2.3) | 8 | +0.26 ± 0.31 | 8% |
Versions B, C and E, with 13-25 trades, give positive results that a random subset of the same size reaches 4-8% of the time: a low value, but to be read together with the number of combinations tried (chapter 12). Version D, which has 104 trades, gives −0.06 ± 0.14: like the baseline. The full version with four filters has 4 POI and 1 trade.
Entry variants. On version B, market entry after the touch gives +0.61 ± 0.34 (19 trades) and H1 confirmation +0.27 ± 0.45 (10). On D, market +0.11 ± 0.16 (80) and confirmation +0.10 ± 0.21 (32).
Version D by side, kind and period. Demand −0.22 ± 0.20 (44 trades), supply +0.05 ± 0.19 (60); order blocks −0.13 ± 0.19 (51), FVGs 0.00 ± 0.19 (53). In the first half of the period (up to 24 May) D gives −0.18 ± 0.21 (41 trades) against −0.10 for the baseline, in the second +0.01 ± 0.18 (63) against −0.10. D signals per month: from 10 to 22 in full months.
The number that weighs most is frequency: B fires 2.6 times a month. Reaching a hundred trades takes about 38 months, more than three years.
A four-step routine
- Write the system's rules before looking at the results, including the filter combinations you will use.
- Count signals per month: under 10 a month your journal answers after years.
- Compare each version with random subsets of the same size: if chance does as well, the version says nothing.
- Do not change the rules after a stretch of losses: change them only with a new sample.
Exercises
Chapter exercises
1. On the chart the rectangles are the signals of version B in two weeks. Which letter is the only one of the four signals that reached the target?
Read the chart in words
| When | What | Price |
|---|---|---|
| 2 Sept 2026 16:00 UTC | Point A | 4,369.63 |
| 7 Sept 2026 00:00 UTC | Point B | 4,432.49 |
| 7 Sept 2026 04:00 UTC | Point C | 4,416.39 |
| 8 Sept 2026 08:00 UTC | Point D | 4,411.63 |
Times in UTC.
2. Version B gave +0.35 ± 0.30 R over 25 trades, and random subsets of 37 POI do at least as well 4% of the time. What do you conclude?
3. B fires 2.6 times a month. About how many months does it take to get a hundred trades?
4. Version D (104 trades) gave −0.06 ± 0.14 R, the baseline −0.10 ± 0.09. What do you conclude?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.