Taking the Profit

The claim, stated the way its believers state it

You can never go broke taking a profit. It is the oldest piece of advice in the business and it is not stupid. A profit taken is a profit; a profit left on the table can walk back to zero and through it. Somebody who books every gain the moment it appears has a record made entirely of gains, and a record made entirely of gains cannot be the thing that ends a career. Compare that with the alternative on offer, which is sitting through a position that has been right and might stop being right, on a rule you wrote weeks ago, with money that is already yours in some sense and not in any sense you can spend.

The advice also has the one property that keeps bad advice alive: it is rewarded, often, and visibly. On the seven trades below it improves one of them outright. The second trade makes 2.00 under the rule and 3.00 under the habit, because the price the habit sold at was higher than the price the crossover eventually sold at. Anyone taking profits early collects that experience several times a year and files it under judgment.

The sheet this module starts from

Module 12 has one running object and this is it: a sheet of seven trades, the ones lesson 63’s best rule takes on the sixty closes this course has been using since lesson 32. Every lesson in this module adds one column to it. This lesson supplies the entries, the rule’s own exits and what a profit target does to them; lesson 87 adds the stop; lesson 88 adds the size; lesson 89 adds what you thought was going to happen. The convention is lesson 63’s and is repeated here so nothing has to be taken on trust: a two-bar average is compared with a five-bar average at the close of each bar, the comparison is not read before bar 5, and the position is put on or taken off at the close of the following bar. Costs are 0.1230 a share for the round trip, which lesson 85 makes 0.0796 of an R.

That is 10.70 gross, 0.861 in costs, and 9.84 net, on thirty of the fifty-nine moves in the series. Buying at the first close and selling at the last nets 5.68, so the rule’s own contribution is 4.16 and everything on this page is measured against that number rather than against 9.84. Six trades finish positive and one does not.

One sentence of a rule, run at every ordinary setting

Here is the habit, written as a rule so it can be tested: leave the trade at the first close at or above the entry price plus some target, and if that never happens, leave it where the rule would have. The target is a dial like every other dial in this course, and it has a natural set of ordinary values, because R is the unit this course measures distance in. Nothing, a quarter, a half, one, one and a half, two. Seven settings including the rule’s own exit, which is the same thing with the dial turned all the way off.

Every row below runs the same seven entries. None of them changes the trade count, so none of them changes the costs: seven round trips at 0.1230 is 0.861 in every row. The last column is what the row nets over buying and holding.

Read the fourth column first, all the way down. Six of seven, six of seven, six of seven, seven times. Then read the third: 7.64, 7.64, 8.84, 11.74, 11.04, 10.64, 9.84. The distance between the best row and the worst is 4.100 a share, and the whole rule was worth 4.162 over holding, so the exit setting alone accounts for 98.5 per cent of what the entry rule earned. The choice everybody argues about is when to get in. The choice worth 98.5 per cent of the result is one most people make by feel, and the column they check their feel against is constant down the page.

The first two rows are identical, and that is a check rather than a coincidence. A quarter of an R is 0.386 a share, and the smallest profitable move any of these seven trades makes is 0.40, on trades 3 and 4. So every target below 0.40 fills at exactly the same place as no target at all, and a reader who recomputes the table and gets two identical top rows has the arithmetic right rather than a copy-paste error.

Two further columns are worth a moment because they say what the win rate will not. Bars held falls from thirty to eight, so the habit spends a quarter as long in the market to keep four fifths of the gross — which sounds efficient until you notice that the money it gave up is money, and the time it saved is not spent on anything. And the t-statistic falls from 3.65 to 2.35, which is the same seven trades saying less about themselves than they did before, because the thing that shrank was the average and not the spread.

What survives

The habit is not simply wrong, and a page that stopped here would have cheated. The best row on the table is a target, not the absence of one: a one-R target nets 11.74 against the rule’s 9.84, which is 6.06 over holding rather than 4.16. Exiting on a fixed distance can genuinely beat exiting on the signal that got you in, and the reason is visible in the first table. Trade 3 runs thirteen bars and gives back some of what it made before the crossover finally fires; a target that is far enough out to let a move develop and close enough in to leave before the giving-back is a real thing and not a story.

What is wrong is the specific version everybody actually runs, which is the top row: out at the first sign of green. That row is the worst of the seven, by a margin larger than the entry rule was worth. Almost nobody you trade against has run this table on their own record, and running it costs an evening and a spreadsheet with three columns in it, because the entries are already in their statement and the only new column is what the price did afterwards.

Trade 3, end to end

The rule goes long at the close of bar 26, at 98.8. It is the lowest entry of the seven and the trade the sheet turns on.

Bar 27 closes at 99.2. That is a profit of 0.40 a share, which is 0.259 of an R, and it is the first close above the entry, so the habit sells there. The trade is a winner. It goes into the record as a winner, it goes into the win rate as a winner, and it books 0.40 less the 0.1230 round trip, which is 0.277.

Now leave the position alone. Bar 28 closes at 99.6. Bar 29 at 100.0. Bar 32 at 100.3, and here is the boundary worth checking by hand: the one-R target on this trade is 98.8 plus 1.5443, which is 100.3443, and the close is 100.3. It misses by 0.0443 — four cents and a bit — so the one-R exit does not fire on bar 32. It fires on bar 33 at 101.5, for 2.70. Anyone reproducing this in a spreadsheet with prices rounded to one decimal will fill on bar 32 and get 1.50, and the difference between their table and this one is a rounding convention rather than a disagreement about the market.

The rule’s own exit comes at bar 39, at 101.3, for 2.50.

So the same trade, on the same seven decisions, is worth 0.40, or 2.70, or 2.50, depending on nothing but where you decided to leave. And in every one of those three worlds it is a win, recorded as a win, counted in the six of seven that the last four paragraphs of every trading journal are about.

A win rate cannot see the size of a win.

That is why the habit that costs the most is the one that leaves no mark. Three of the four numbers a trading journal actually reports come out unchanged: seven trades, six winners, and one loser at 0.30 in both worlds, because the losing trade never printed a profitable close for the habit to take. The fourth moves, and it is the one nobody writes down. The average winner falls from 1.833 a share to 1.467, which is exactly a fifth. A trader who halved their position size would see it in the account inside a month. A trader who did this sees a journal that still says six of seven, still says the system works, and it does still work — at 1.96 over holding instead of 4.16, which here is the difference between a rule worth running and a rule barely worth the costs it pays.

So take your own last twenty closed trades and add one column: what the price did after you left.

What this does not settle

That a one-R target is the right target. It is the best of seven settings tried on seven trades, which is a search, and lesson 63 exists to price exactly that: a search of 253 cells on this same series found a result that a search of the same size beats on 52.5 per cent of series with nothing in them. Seven cells is a much smaller search than 253 and seven trades is a much smaller sample than the search deserves. The honest reading of the table is the spread between its rows, which is 4.100, and not the location of its maximum, which is worth almost nothing.

That the win rate is useless. It is not; it is one of the two numbers expectancy is made of, and lesson 17 needs both. What this page shows is narrower and worse: the win rate is exactly the number that a change in exit policy cannot move, so it is the wrong instrument for the one decision most likely to be quietly destroying a record. A trader who watched their win rate fall would investigate. Nobody investigates a number that has not moved.

That seven trades can carry any of this. They cannot. Lesson 19 prices that directly: fifty trades cannot separate a method making 0.35R a trade from one with no edge at all, and seven is not fifty. The reason to believe the finding anyway is that it is not a claim about how much money the target makes, which would need a sample. It is a claim about what the win rate is arithmetically capable of registering, and that holds on any record, including one of seven.

That the habit is about fear. It might be, and the sources at the foot of this page make a serious case that it is, but nothing measured here shows a state of mind. What is measured is an exit policy and its cost. A reader who takes profits early out of a considered view about mean reversion, and a reader who does it because an open gain is unbearable, produce the same sheet and lose the same 2.20 a share, and this page cannot tell them apart.

And the concession that costs most: this whole table exists because the exits could be moved while the entries were held still, and that is not the position you are actually in. In a real record the two are entangled, because a trader who takes profits early takes different trades afterwards — the capital is free sooner, the next signal arrives with a position already flat, and the seventh trade might never have been taken at all. Everything on this page assumes the seven entries survive the change in exit policy, and on a live account they would not. Lesson 87 takes the mirror of this habit, which is where that entanglement first bites, because it is the setting that appears to do nothing at all: every stop from half an R outward leaves the record exactly as it is, all seven trades, 9.84 a share, six winners. Tighten it one notch to a quarter of an R and it fires on two of them, both winners, turning six winners into four and 9.84 into 3.94, which is 1.74 below simply holding.