Part 3: Size, signals and outcome · Chapter 9 of 14 · 8 min
Too many signals
Adding signals raises the number of trades and lowers the quality of each. On the plan (80-candle breakout, 79 trades, +0.74 ± 0.43 R) whoever adds the same rule at 10 candles makes 289 with +0.36 R on average and +104.5 R in sum: the sum rises, but mean R halves and the added rule (+0.22 ± 0.24) cannot be told apart from zero. Entries without a signal, at random and with a narrow stop, make −0.10 R per trade: 961 trades and about −94 R.
The temptation to do more
The plan makes 8.1 trades a month, one every 3.7 days: for 77% of the hours it has no position open (chapter 11). The rest of the time you look at a moving chart. The temptation is to fill that time: add another "faster" rule, take an "almost valid" signal, enter because "it is moving". Three forms of too many signals:
- Extra rules, faster or less selective, next to the plan.
- Half signals: entries that resemble those of the plan but are not.
- Entries without a signal: entering for the sake of it, with narrower stops because "it is fast anyway".
Extra rules
Book 12 showed that, with the same exits (1.5 ATR stop, 3 R target, duration 48), the breakout length changes the result a lot:
| N-candle breakout | Trades | Mean R ± error | Sum | Total cost | Winning |
|---|---|---|---|---|---|
| 10 | 210 | +0.22 ± 0.24 | +46.4 R | 2.3 R | 35% |
| 20 | 178 | +0.15 ± 0.26 | +26.5 R | 2.0 R | 30% |
| 30 | 141 | +0.26 ± 0.31 | +37.2 R | 1.5 R | 33% |
| 50 | 118 | +0.28 ± 0.34 | +32.7 R | 1.2 R | 33% |
| 80 (plan) | 79 | +0.74 ± 0.43 | +58.1 R | 0.8 R | 46% |
The 10-candle rule makes 21.6 trades a month (about one a day) against 8.1 for the plan, and has a mean R of +0.22 ± 0.24: the lower limit is below zero, it cannot be told apart from a rule with no edge (book 12, chapter 2). Whoever adds it to the plan:
| Trades | Mean R | Sum | |
|---|---|---|---|
| Plan alone (80 candles) | 79 | +0.74 | +58.1 R |
| Plan + 10-candle rule | 289 | +0.36 | +104.5 R |
The sum rises (from +58.1 to +104.5 R) and the mean R per trade halves (from +0.74 to +0.36). The journal looks better (more gain), but each trade is worth on average half, and three quarters of the trades (210 out of 289) come from a rule whose edge is not proven. With a rule with no edge the added sum would be zero (as in book 11: more trades do not create an edge). The risk is doubling the work, the attention and the exposure (the number of trades almost quadrupled) for a gain that could be noise.
Read the chart in words
Daily gold chart with, below, three cumulative R curves: two rising (80 and 10 candles) and one, much steeper, falling.
Entries without a signal
The extreme case is entering without a signal. We simulate two versions, with random entries and random direction, on the same candles:
| Random entries | Trades | Mean R ± error | Sum |
|---|---|---|---|
| About one every 20 hours, same exits as the plan (1.5 ATR stop, 3 R target, 48 candles) | 120 | +0.08 ± 0.30 | +9.6 R |
| Very frequent, narrow stop (0.5 ATR, 1 R target, 12 candles) | 961 | −0.10 ± 0.06 | −93.6 R |
With the plan's exits (wide stop, large target) random entries are worth about zero (+0.08 ± 0.30): they do not lose, because a wide stop and a large target make the cost small (0.01 R) and a random direction has no expected value. But with a narrow stop (0.5 ATR) and many trades the result is clear: −0.10 R per trade, 961 trades, about −94 R. There the cost weighs more (0.03-0.05 R per trade, book 11, chapter 10), the ambiguous candle counts (book 12, chapter 8) and the number of trades multiplies everything. The sign and the size of the error (±0.06) make it certain: it is not bad luck, it is arithmetic.
The paradox is that with many small trades the result shows early and the feedback is frequent: every day there are winning trades, and the feeling is of being active and learning. The journal, after 961 trades, says −94 R.
When more signals are a choice
It is not forbidden to have more rules. A trader can have two plans. The difference is whether they are written, measured and tested (book 12), with a size that takes all of them into account (book 11, chapter 11: total open risk, correlation), and whether the addition improves mean R or only the sum. Adding a signal because "it is moving" is a deviation; adding a second plan after testing it out of sample, with a risk cap, is a choice.
What to do with it
- Count the plan's trades and the off-plan ones, separately, in the journal (chapter 2).
- Compare the mean R, not just the sum, of each group of trades.
- Put in the plan a maximum number of trades per day or per week, and a total open risk (book 11).
- Do not add rules to the plan without testing them out of sample (book 12, chapter 13).
- Consider "doing nothing" a part of the plan: for 77% of the time the plan does not trade.
Procedure in four steps
- Separate in the journal the plan's trades from the off-plan ones.
- Compute for each group trades, mean R ± error, sum, cost.
- Fix a maximum of trades and of total open risk in the plan.
- Before adding a rule, test it out of sample with the error and the correlation with the plan.
Exercises
Chapter exercises
1. Adding to the plan (79 trades, +0.74 R) the 10-candle rule (210 trades, +0.22 ± 0.24 R) gives 289 trades with mean R +0.36 and sum +104.5 R. How is it read?
2. Random entries with a narrow stop (0.5 ATR), 961 trades: −0.10 ± 0.06 R. Why is the result certain?
3. Random entries with the plan's exits (wide stop, large target) are worth +0.08 ± 0.30. Why do they not lose?
4. For what part of the time does the plan have no open positions?
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.