The Pace You Actually Trade At

What a day contains

Start with what the rule reads, because everything else follows from it. Every cell in lesson 63’s grid compares two moving averages of the close. It reads one number a day. A session is six and a half hours and the rule consults one instant of it, so the remaining three hundred and eighty-nine minutes are not information the rule is ignoring, they are information it was never built to take. That is not an argument that watching is bad. It is the observation that for this family of rules there is nothing in the watching that reaches the decision.

So count the decisions. Here are lesson 71’s four rules over the twenty-nine moves on which all of them are defined, which is the same twenty-nine moves lessons 71 to 75 have been working on.

Read the first column down. Thirteen orders in twenty-nine days, and nine of the thirteen come from one rule. The 2-and-5 crossing generates 69.23 per cent of the book’s workload, and lesson 73 gave that same rule the largest weight it could hold, 0.711. Those two facts are the same fact. Look at the third column of lesson 73’s inputs: 2-and-5 had the lowest standard deviation of the four, 0.6015 against 0.7104, 0.7278 and 0.7082, and it had it because it holds a position on nineteen days of the twenty-nine while the others hold on twenty-two to twenty-six. Being out of the market more often is what made it quiet, being quiet is what won it the weight, and being out of the market more often is also what makes it flip. The rule with most of your money in it is the rule with most of your day in it, by construction rather than by luck.

Read the third column and the fourth together, and mistrust the fourth. Turning one entry in twenty-eight moves into nine entries a year is arithmetic performed on a single event, and no single row of that column should be quoted. The aggregate is what carries weight: across all 253 cells, on the stretch each is defined on, 203 of them fired once, thirty fired twice or three times, and twenty fired more than three, the busiest reaching fifteen. The distribution, not any row of it, is what says the pace is nearer half a dozen entries a year than forty a month.

Now the same grid on the shared window, which is where the day comes from. The 253 rules produce 456 orders over the twenty-nine days, an average of 1.80 each, and 80.63 per cent of them ask for at most one order in the whole stretch. And here is the number that stops this from being an argument for idleness: run all 253 and only one of the twenty-nine days is quiet. A full day’s work is available. It costs exactly what lesson 75 priced it at, which is a quartering of the size per trade for a day 4.61 per cent smaller.

Which turns the question from how to fill the day into which parts of it are a function of what is already stored.

The first four rows are twelve operations for a four-rule book: two averages and one comparison per rule, plus a division and a count on the days an order exists. On sixteen of the twenty-nine days all twelve return yesterday’s answer. Lesson 69 measured the delay a person inserts between the signal and the order; this table is what that delay is being inserted into, and it is twelve operations that a spreadsheet finishes before you have found the chart.

The last two rows are the ones that feel like the day’s real work and are not daily at all. Lesson 67’s 589 trades arrive at the pace measured above, which on the busiest rule in the grid is eight and a half years. A daily review cannot decide whether a rule still works, because the evidence needed for that decision does not arrive daily. What it can decide is whether the twelve operations ran, whether the order that should have gone out went out, and whether the count in lesson 75’s fourth column is still what you think it is.

Almost nobody has counted the orders their own rules asked for last month and divided by the trading days. It is one count and one division, and the answer is the pace lesson 65 asked you to supply and most readers never did, which means most readers are still carrying lesson 65’s waits with somebody else’s numerator in them.

Module 10 keeps a running list of what the day contains that is not the decision, and this page opens it with one item: the twelve operations, and the sixteen days in twenty-nine on which they change nothing.

Six days of nothing, then five in a row

The thirteen order days do not arrive evenly. They fall on moves 34, 37, 38, 40, 42, 43, then a gap, then 50 to 54 consecutively, then 56 and 57. So the window contains a six-day stretch with no order in it and, immediately after, five consecutive days each of which needs one. Take those two stretches against each other, on the 2-and-5 rule, which produced all five of the busy days.

Moves 44 to 49, six days: the position never changes, no order is sent, and the rule earns 1.90 in the instrument’s units, which is 2.65 typical daily moves.

Moves 50 to 54, five days: exit, entry, exit, entry, exit. Five orders in five sessions, and the rule earns 0.20, which is 0.28 of a typical daily move.

The six days that asked for nothing made nine and a half times what the five days that asked for something made, and that is before any of the five orders is paid for. The busy stretch is the fast rule being whipsawed: the two-bar average crosses back and forth over the five-bar average while the instrument goes almost nowhere, up 0.40 across the whole five days. Every crossing is real, every order is correct under the rule, and the week is worth a quarter of one day’s movement.

Now price the five orders, which is two and a half round trips, using lesson 12’s two ends. On lesson 12’s index ETF, where a round trip costs a fifth of one per cent of a typical day’s movement, the five orders cost 0.0036 against a gain of 0.20: 1.8 per cent of it, and the week survives. On lesson 12’s micro cap, where a round trip costs 52 per cent of a day’s movement, the same five orders cost 0.93, which is 4.7 times the entire gain. Identical rule, identical week, identical orders, and the only thing that changed is the instrument lesson 12 told you to choose first.

Read the two stretches against the feeling each produces. The six quiet days feel like a system that has stopped working and produce nine tenths of the eleven days’ money. The five busy days feel like a working week and produce a tenth of it, or a loss, depending on a decision made in lesson 12 before any of this started. The urge to do something on the quiet days is an urge to convert the first stretch into the second.

So count your own order days before you decide the day is empty, and price your busiest week before you decide it was productive.

What this does not settle

That one entry in twenty-eight moves is a pace. It is one event on one stretch of one instrument, and the sixty closes this course carries are far too few to estimate any single rule’s frequency. What survives that objection is the aggregate and its order of magnitude: 203 of 253 cells fired once and 86.17 per cent fired twice or less, on windows of twenty-eight to fifty-five moves. That is not consistent with forty trades a month by any reading, and the gap is a factor of tens rather than a rounding.

That the annual column should be quoted. It should not. Every row of it multiplies a count of one or six by a constant, and a count of one has a standard error about the size of itself. The column exists so that 480 orders a year, which is what lesson 65’s forty a month comes to, can be set beside 28.5 and 6.6 without arithmetic in the reader’s head. Use the distribution instead, and treat every yearly figure on this page as arithmetic rather than as measurement.

That the day is empty for everyone. It is empty for a daily-close rule family, which is what this course has been building since lesson 63. A rule that reads the session reads thousands of numbers rather than one, and has a genuinely different day; nothing here says intraday attention is worthless, only that these rules cannot consume it. The test of whether your own day is like this one is the count in problem one, not this page.

That a machine may run the first four rows unattended. It may compute them unattended, which is not the same permission. Lesson 69’s finding was that a loss limit reads and a position cap prevents, and lesson 75 made the cap the control; both of those still need somebody who is answerable when the count comes back wrong. What the twelve operations remove is the delay and the arithmetic error, not the responsibility.

That doing nothing and knowing nothing are the same. On the sixteen quiet days the twelve operations still have to run and the answer still has to be looked at, because the difference between a day with no order and a day where the order was missed is invisible from the outside and is exactly the difference this page cannot see in your record either. An idle day is monitored. An unmonitored day only looks idle.

And the concession that costs most: this page counted the decisions and said nothing about what has to be standing up for even one of them to reach the market. Twelve operations are worthless if the data feed is stale, the broker is unreachable, or the machine holding the averages is the same laptop that is running the charts. Lesson 77 starts there, on what redundancy actually buys and what it does not.