Forward Testing On Demo Without Wasting The Sample

You searched for how long you should demo a strategy before going live. The honest answer is that the question points at the wrong variable. Weeks are easy to count and tell you almost nothing. What decides whether a demo period was worth doing is not its length. It is whether you set the conditions, the logging discipline and the stop date before the first trade, so that at the end you are holding evidence rather than an impression.

Most traders finish a demo run able to say only that it felt alright, or that they were doing well until the last few days. That is not a result. It is a memory, and memory edits itself in exactly the direction that costs money later. A demo period is a sample, and a sample you did not define in advance is not evidence, it is a memory.

A calendar month is not a fixed amount of trading. Two traders can both run a month on demo and one sees forty entries while the other sees six, because their setups appear at different rates. If you measure the test in weeks, you are mostly measuring the weather. Count occurrences instead.

Decide, before you start, how many occurrences of your setup you want to see. The number is yours, and it should be large enough that one unusual run does not dominate what you conclude. Then do the arithmetic honestly. If your setup appears roughly twice a week and you want forty occurrences, you are committing to about twenty weeks. That is worth knowing before you start rather than after.

Counting occurrences also guards against the quietest failure in forward testing: taking trades the strategy never asked for, simply to fill the time. If the market does not offer your setup, the sample grows slowly. That is a finding in itself. A setup that appears rarely carries a cost in patience, and you want to meet that cost on demo.

A forward test is only a test if one thing is being tested. Write the fixed variables on a single page and do not touch them until the stop date.

Instrument and session. The setup definition, meaning what qualifies and what disqualifies, in words someone else could apply without asking you. Risk per trade as a fixed fraction of the demo balance, and a sizing rule that follows from it. Management rules decided in advance: when you move to break even, whether you take partials and where, whether you trail. News handling, meaning which releases you stand aside for.

Two more get skipped often. The demo balance should resemble the account you would actually trade, because sizing behaviour changes with scale. And the hours you sit at the screen should resemble the hours you will actually have, not a fortnight of annual leave.

Every one of these, changed mid-sample, splits the sample in two. Changing one is allowed. Pretending the earlier trades still count is not. Note the change in the log and start the count again.

The journal is the only output the test produces. One entry per trade, written at the time, not reconstructed on Sunday evening.

An entry holds the date and session. The setup, classified by name. The specific reasons it qualified, checked against your written checklist. Entry, stop and target as planned. What you actually did while the trade was open. The outcome expressed in multiples of risk rather than currency, so that trades of different sizes can be compared. A chart note or screenshot.

Then the column that matters most: followed or deviated. One word. If you deviated, a second line saying how. Entered early. Moved the stop. Traded outside the session. Skipped a trade that qualified.

That flag separates two questions people collapse into one. First, does the strategy describe something repeatable. Second, can you execute it as written. Without the flag, a sample full of deviations gets blamed on the strategy, and you discard a process that was never actually run. Skipped trades belong in the log for the same reason. A strategy you cannot bring yourself to take is not viable for you, whatever the other entries show.

Write the stop condition next to the start date. The sample ends when the occurrence count is reached, or on a hard calendar date. Pick one, write it down, and let it decide. Without it, demo periods end on emotion. A good run sends people live early. A bad run makes them abandon a process they never finished testing.

Write the review questions now, too, while you have nothing invested in the answers. How many trades. How many flagged as followed. The distribution of outcomes in multiples of risk, not just the total. Which setup variant produced most of the entries, and which produced most of the damage. Whether the losses cluster in one session or around one kind of news. Whether the rule you broke most often is a rule worth keeping at all.

Then decide in advance what each answer leads to: continue the sample, revise one variable and restart the count, or stop testing this strategy. Whether you later trade it with your own money is a separate decision that depends on your circumstances and your tolerance for loss, and nothing written here can make it for you. What a defined sample can do is stop you making it blind.

You cannot forward test a process you have not written down. That is the practical reason the strategies and course material in the MMFX member app are written as steps rather than as ideas.

Complete Workflow covers every step from blank chart to closed trade in one end-to-end session. Your Daily SOP sets out the eight-step ritual for a trading day: mark, bias, classify, wait, confirm, execute, manage, journal. Trade Management covers break-even triggers, partial profits, trailing and scaling, which are exactly the in-trade decisions that have to be fixed before the sample starts rather than improvised while a position is open.

For a forward test, the value is not novelty. It is that the steps are stated. When the process is written in that much detail, the word followed in your journal means something specific, and the word deviated points at a named step you can go back and read. A sample run against a vague idea produces vague conclusions no matter how many trades it contains.

None of this material tells you what the market will do next, and it should not be read that way. It gives you a stated process. Take it as a starting draft, adapt the parts that do not suit your hours or your temperament, then freeze it for the length of the sample and let the log do the arguing.

Trading carries risk of loss. Every decision you take is your own, and nothing on this page is financial advice.