Part 4: The system under test · Chapter 13 of 14 · 8 min

A month of work: September 2026 with the system

A whole month followed with the journal: 32 points of interest, 12 for version B and 10 trades (+4.82 R, 5 targets), against the baseline of 27 trades (−6.46 R). The routine of a day, the weekly review and a warning: September weighs more than half of B's result, so it is not out-of-sample proof.

The routine of a day

  1. 00:00 UTC. Write the previous day's high and low and the D1 bias (chapter 3).
  2. 07:00. Mark the Asian range (book 3, chapter 11).
  3. At every H4 close. If the candle completes a POI (order block or FVG), fill in the filters: D1 trend, half of the range, a level taken just before, structure break. If the version you use says yes, compute entry, stop, target and size (chapter 10) and place the limit order.
  4. During the day. Do not change stop and target. Cancel limit orders after 40 candles or after a close beyond the far side.
  5. Evening. Write in the journal every POI completed, including the discarded ones, with the filters and the outcome.

The weekly review counts POI, trades and the sum in R, and compares with the baseline. It does not change the rules.

September 2026

In September 32 points of interest complete: 12 order blocks and 20 FVGs. Version B (D1 trend aligned and right half) selects 12 of them, which give 10 trades (two orders never fill): 5 targets, 5 stops, +4.82 R. The baseline, with all POI, gives 27 trades, 7 targets and −6.46 R (mean −0.24). The 17 trades B discards are worth −11.28 R. With 1% risk per trade, uncompounded, B would give +4.8% on the account and the baseline −6.5%.

WeekPOIB: trades, sumBaseline: trades, sum
1-6 September51, −1.01 R5, −2.04 R
7-13 September96, +5.87 R8, +3.85 R
14-20 September62, −2.01 R6, −3.08 R
21-27 September81, +1.97 R6, −3.16 R
28-30 September402, −2.02 R

Figure 1 shows the month on the daily chart, with the significant D1 swings (highs 4,700.63, 4,514.81, 4,405.75 and 4,228.70 dollars, all falling; lows 4,288.00, 4,243.81 and 4,116.60 falling) and the trades of B at their exit. The bias is bearish for much of the month, and the trades of B are all supply except one.

Figure 1. September on the daily chart: the D1 swings are falling (highs from 4,700.63 to 4,228.70, lows from 4,288.00 to 4,116.60) and the trades of version B, nearly all supply, exit with +2R or −1R.Data: gold, XAUUSDT perpetual future (real candles) · daily (D1) · from 17 Aug 2026 to 5 Oct 2026.
Read the chart in words

D1 gold chart from 17 August to 5 October 2026. Arrows mark the significant swings, falling. Labels mark the trades of version B at their exit, at +2R or at −1R.

WhenWhatPrice
10 Sept 2026−1R4,315.55
8 Sept 2026−1R4,446.54
7 Sept 2026−1R4,432.49
10 Sept 2026+2R4,438.79
8 Sept 2026+2R4,446.54
14 Sept 2026+2R4,359.96
14 Sept 2026+2R4,359.96
17 Sept 2026−1R4,387.02
17 Sept 2026−1R4,387.02
23 Sept 2026+2R4,372.54

Times in UTC.

Figure 2 shows the week from 7 to 14 September on H4: six trades of B (coloured rectangles) and two POI discarded by the filter (grey rectangles). The six of B give +5.87 R: four targets (two with overlapping zones) and two stops in the first two days. The two discarded ones were stops (−1.01 R each).

Figure 2. One week: the trades of version B (coloured rectangles) and the discarded POI (grey rectangles), with the outcome at the exit.Data: gold, XAUUSDT perpetual future (real candles) · 4 hours (H4) · from 7 Sept 2026 00:00 UTC to 14 Sept 2026 04:00 UTC.
Read the chart in words

H4 gold chart from 7 to 14 September 2026. Coloured rectangles mark the order blocks and FVGs of version B, grey rectangles the discarded POI, each with a label at the exit: targets at +2R and stops at −1R.

WhenWhatPrice
8 Sept 2026 00:00 UTCB −1R4,444.09
7 Sept 2026 08:00 UTCB −1R4,426.89
7 Sept 2026 12:00 UTC−1R4,422.40
10 Sept 2026 12:00 UTCB +2R4,382.86
8 Sept 2026 16:00 UTCB +2R4,407.95
9 Sept 2026 04:00 UTC−1R4,415.70
14 Sept 2026 04:00 UTCB +2R4,343.74
14 Sept 2026 04:00 UTCB +2R4,343.74

Times in UTC.

The journal

The 12 POI of version B, in order:

DatePOIZoneOutcome
2 SeptemberDemand FVG4,337.96-4,369.63stop −1.01
7 SeptemberSupply OB4,431.00-4,439.50stop −1.03
7 SeptemberSupply FVG4,416.39-4,420.49stop −1.04
8 SeptemberSupply OB4,418.93-4,444.09target +1.99
8 SeptemberSupply FVG4,411.63-4,418.93target +1.98
10 SeptemberSupply FVG4,382.86-4,408.83target +1.99
13 SeptemberSupply OB4,347.01-4,360.27target +1.98
16 SeptemberSupply OB4,330.53-4,368.12stop −1.01
16 SeptemberSupply FVG4,282.03-4,330.53stop −1.01
22 SeptemberSupply OB4,369.59-4,373.71target +1.97
27 SeptemberSupply OB4,283.23-4,287.26not filled
28 SeptemberSupply FVG4,265.39-4,283.23not filled

What the month says, and what it does not

It says the procedure is workable: 32 POI in a month, 12 pass the filter, 10 become trades, about one every three days. It also says that, even in a good month, results come and go: −1.01 R in the first week, −2.01 R in the third. Five trades out of ten are stops.

It does not say B works. September is part of the sample on which B was put together: it holds 10 of its 25 trades and +4.82 R of the roughly +8.67 R total, more than half. The trades of B are nearly all supply in a bearish month (demand has 4 trades in the whole period). Six of September's 10 trades come from three moves with two overlapping zones, and they are not independent cases. An out-of-sample month, with a rising D1 trend, is the test that is missing.

A four-step routine

  1. Fill in the routine every day: levels, bias, Asian range, POI filters.
  2. Write all POI in the journal, discarded ones included, with the outcome they would have had.
  3. Close the week with three numbers: POI, trades, sum in R. Compare with the baseline.
  4. Do not touch the rules before a hundred trades.

Exercises

Chapter exercises

1. On the chart the coloured rectangles are the POI of version B and the grey ones are discarded by the filter. Which letter is a discarded POI?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 4 hours (H4) · from 7 Sept 2026 00:00 UTC to 14 Sept 2026 04:00 UTC.
Read the chart in words
WhenWhatPrice
7 Sept 2026 00:00 UTCPoint A4,432.49
7 Sept 2026 08:00 UTCPoint B4,426.89
8 Sept 2026 08:00 UTCPoint C4,411.63
13 Sept 2026 20:00 UTCPoint D4,353.73

Times in UTC.

2. Why does September not prove that version B works?

3. In September B gave +4.82 R and the baseline −6.46 R. What can you NOT conclude?

4. Why must the journal also include discarded POI?

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.