Part 2: The filters, one at a time · Chapter 3 of 14 · 8 min
The bias from the daily chart: what it is worth
The bias written as a rule on D1 swings (two highs and two lows rising or falling), two real days in which it holds and does not hold, and measurements over 260 days: in agreement 56-58% of the time, but "always bearish" would have scored 54-56%, and aligned POI did no better than the others.
The rule
Step 1 of the system asks for the direction of the daily chart before looking at the POI.
The confirmation filter has a cost: a swing exists only after price has moved away, and the book waits one more day. The bias always arrives late relative to price.
Over the 260 days used (from 14 January 2026, after a 30-day warm-up) the bias is bullish for 78 days, bearish for 100 and undefined for 82.
Two real days
Figure 1 shows a day when the bias holds, on D1, on 17 March 2026. The highs are $5,420.00 (2 March) and 5,236.74 (10 March), falling. The lows are 5,101.20 (24 February) and 5,004.60 (3 March), falling. The bias is bearish. The day opens at 5,013.59 and three days later closes at 4,646.00: −3.19 ATR (D1 ATR of 115.2 dollars), in the direction of the bias.
Read the chart in words
D1 gold chart from 19 February to 23 March 2026. Four arrows mark the last two significant highs and lows, all falling. Two labels mark the bias day and the three days after, with price lower.
| When | What | Price |
|---|---|---|
| 17 Mar 2026 | today | 5,043.75 |
| 19 Mar 2026 | 3 days later | 4,514.11 |
Times in UTC.
Figure 2 shows a day when the bias does not hold, on D1, on 28 January 2026. The highs are $4,555.55 (28 December) and 4,645.15 (14 January), the lows 4,280.96 (31 December) and 4,540.03 (16 January): all rising, a bullish bias. The day opens at 5,175.10, far above the last swing high, and three days later closes at 4,903.03: −3.01 ATR (D1 ATR of 90.5 dollars), against the bias. The swings were two weeks old: price had already moved by more than 500 dollars.
Read the chart in words
D1 gold chart from 23 December 2025 to 3 February 2026. Four arrows mark the last two significant highs and lows, rising and well below price. Two labels mark the bias day and the three days after, with price lower.
| When | What | Price |
|---|---|---|
| 28 Jan 2026 | today | 5,625.34 |
| 30 Jan 2026 | 3 days later | 4,740.90 |
Times in UTC.
What the data says
For each day we look at the move from the day's open to the close of the third day (also at one and five days), in the direction of the bias. "In agreement" means price moved in the direction of the bias. The other definitions (20-day average, close 5 days earlier, colour of yesterday's candle) are controls.
| Bias definition | Days | In agreement at 1 day | At 3 days | At 5 days | Move at 3 days (ATR) |
|---|---|---|---|---|---|
| D1 swings (the book's rule) | 178 | 102 (57%) | 103 (58%) | 100 (56%) | +0.25 ± 0.11 |
| Close above or below the 20-day average | 260 | 135 (52%) | 143 (55%) | 143 (55%) | +0.20 ± 0.09 |
| Close above or below the one 5 days earlier | 260 | 127 (49%) | 131 (50%) | 137 (53%) | +0.06 ± 0.09 |
| Colour of yesterday's candle | 260 | 135 (52%) | 127 (49%) | 131 (50%) | +0.03 ± 0.09 |
| Control: "always bearish" over the 178 days | 178 | 96 (54%) | 99 (56%) | 100 (56%) |
The margin of error is 6-7 points. The swing bias agrees in 56-58% of cases, against 50% for chance. The "always bearish" control over the same 178 days would score 54-56%: the period was falling, and the swing bias did not beat a rule that looks at nothing. The 20-day average scored 52-55%, the other two definitions 49-53%.
In the two halves of the period: the first 98 days with a bias give 53% (±10 points), the 80 of the second half 64% (±11 points). The difference does not leave the error, but the result is not stable.
On the book's POI the bias changed nothing.
| POI | Trades | Mean R (2R) |
|---|---|---|
| Aligned (118 POI) | 86 | −0.07 ± 0.15 |
| Against (73 POI) | 52 | −0.09 ± 0.19 |
| Undefined (100 POI) | 78 | −0.14 ± 0.15 |
By side: aligned demand −0.15 ± 0.25 (31 trades) against counter-trend demand −0.34 ± 0.24 (27); aligned supply −0.03 ± 0.19 (55) against counter-trend supply +0.19 ± 0.30 (25). For supply the direction is the opposite of the expected one, inside the error.
A four-step routine
- Mark the last two significant highs and lows of the daily chart and write rising, falling or undefined.
- Look at how far price has already moved from the last swing: if more than 2 ATR, the bias is old.
- Compare your bias with a fixed control ("always up", "always down"): if it does not beat it, it tells you nothing.
- Log the bias of every POI in the journal, even when you do not use it.
Exercises
Chapter exercises
1. On the chart the four arrows are the last two significant highs and lows of the daily chart. Which letter is the last significant high, meaning the most recent of the two highs?
Read the chart in words
| When | What | Price |
|---|---|---|
| 24 Feb 2026 | Point A | 5,101.20 |
| 2 Mar 2026 | Point B | 5,420.00 |
| 3 Mar 2026 | Point C | 5,004.60 |
| 10 Mar 2026 | Point D | 5,236.74 |
Times in UTC.
2. Under the book's rule, when is the D1 bias bearish?
3. The swing bias agreed on 58% of days at three days and "always bearish" on 56%. What do you conclude?
4. Why does the swing bias always arrive late?
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.