Most traders arrive at this question after a bad month. The rules worked in testing, they still work on some days, and the damage seems to cluster at hours nobody can explain. So they search for the best time of day to trade their gold strategy, find a chart of average hourly ranges, and bolt a window onto the rules.
That is the wrong order. A session filter is not a way to delete losing trades; it is a statement about the conditions your setup needs, written in clock time. Get the statement right and the hours follow from it. Start with the hours and you are curve fitting with extra steps.
This guide covers four things: what a time window actually does to a setup, why gold's character shifts through the day, how to derive a filter from the mechanism of your setup, and how to tell a real filter from a fitted one.
A session filter is not a way to delete losing trades; it is a statement about the conditions your setup needs, written in clock time.
A Time Window Changes The Setup, Not Just The Count
The common assumption is subtraction. You had one hundred trades, you add a filter, now you have sixty, and the forty you removed were the bad ones. That is not what happens. A window does not subtract from a population, it selects a different one.
A breakout strategy restricted to the London open is a different strategy, not a smaller version of the old one. The trades that remain live in deeper liquidity, wider ranges and faster resolution. The ones you removed were a mix, and the mix mattered. What is left has a different shape: different distance between entry and the level that invalidates it, different time for a move to resolve, different cost to get in and out.
That has consequences you have to go back and re-read. If the trades that survive the filter develop faster, a trailing rule written for slow hours will behave differently. If they open with a wider initial swing, a stop placed by a fixed distance is now doing a different job. Volatility and stop distance travel together, and a session filter is a volatility filter wearing a clock.
So when you add a window, treat the result as a new strategy that deserves a fresh read of every other rule in it. Entry, invalidation, management and sizing were all written against the old population. They do not automatically carry over.
WHAT A WINDOW DOES
Selection
A breakout strategy restricted to the London open is a different strategy, not a smaller version of the old one.It forces you to re-read entry, invalidation, management and sizing against the population that is left.
Subtraction
You had one hundred trades, you add a filter, now you have sixty, and the forty you removed were the bad ones.The trades removed were a mix. What remains has a different shape, so the old rules no longer automatically fit.
Why Gold's Character Shifts Through The Day
Gold trades around the clock, but the people trading it change. That is the whole mechanism behind every session filter, and it is worth describing plainly rather than as a rule about direction.
Through Asian hours participation is thinner and ranges tend to be narrower. Levels can hold for hours and then give way without much warning. At the London open, European desks arrive, ranges expand, and the night's range is often tested early. The London and New York overlap carries the busiest participation of the day, and it is where most scheduled releases land. Late in the US day participation thins again and moves can drift or reverse without an obvious cause.
Cost moves with this too. The spread you pay and the slippage you absorb are not constant across the day, and they are part of whether a setup is worth taking at all.
None of this tells you where price is going. It describes conditions. Your setup depends on conditions, which is why it depends on hours — indirectly, and only through the conditions. Keep that chain intact and your filter stays honest. Break it, and the clock becomes a superstition.
Logic First: Write The Reason Before The Hours
Before you look at a single hourly breakdown, finish this sentence: my setup needs this condition present, and that condition tends to be available when this is happening in the market.
A sweep-and-reverse setup needs an obvious pool of resting orders to have built, and then enough participation arriving to take it and reject. That implies a quiet period that builds the pool, followed by an arrival of size. A trend continuation pullback needs a directional push already established before the pullback has anything to continue. That implies you are not the first participant of the day. A mean-reversion fade needs a range that participants are actually respecting, which is a different condition again.
Each of those sentences implies a window on its own, before any data. Write the window down. Only then go to your records — and go there to argue against the window, not to find the hour that scores best. Falsifying is a different activity from discovering, and only one of them survives contact with next month.
Two practical points. Keep the boundaries wide. A mechanism that only exists between 13:00 and 14:00 is almost certainly not a mechanism. And anchor the window to a session, not to a fixed server hour. Clocks shift twice a year and your broker's server clock does not necessarily shift with the session you care about.
HOW TO DERIVE A WINDOW
Curve Fitting Looks Like A Filter
The test is one hour. Move each boundary an hour earlier, then an hour later. If the whole thing collapses, you did not find a condition — you found noise in a small sample and gave it a name.
The tells are consistent. Boundaries on oddly precise minutes. Several windows carved out of the day rather than one. A weekday excluded with no reason attached to it. A filter discovered after reading results rather than before. And the quiet one: very few trades inside each hour bucket, so that two or three outcomes decide which hour looks best.
The cure is not more data. It is the sentence. If you cannot say why 13:00 and not 12:00 in terms of what your setup needs, you do not have a filter. You have a preference you are dressing up.
THE ONE-HOUR TEST
You find an hour in your records where the setup appears to behave better than elsewhere.
You can state what the setup needs and why that part of the day tends to supply it
ThenWrite it as a wide window anchored to the session, and keep the boundaries loose
Moving each boundary an hour earlier or later collapses the whole thing
ThenTreat it as noise in a small sample and leave the rules alone
You cannot explain the hour in terms of conditions at all
ThenKeep the rules unfiltered and keep journalling the hour until the records can argue
Writing The Window Into A Strategy You Can Follow
A session filter that lives in your head is not a filter. It is a mood that shows up when you are already tempted.
A session filter that lives in your head is not a filter. It is a mood that shows up when you are already tempted. The window belongs in the strategy document, not in your head.
Inside the MMFX strategies, the filter sits with the rest of the conditions rather than off to the side: the setup, the entry trigger, the level that invalidates it, the management rules. One line, written plainly enough that someone else could apply it without asking you what you meant.
It then has a specific home in the routine. The Daily SOP lesson in the Managing and Routine module lays out an eight-step ritual for every trading day — mark, bias, classify, wait, confirm, execute, manage, journal. The window lives in wait. It is the step that decides whether the clock has earned you the right to be looking for the setup at all. The Complete Workflow lesson runs the same sequence end to end, from blank chart to closed trade, which is where you will notice whether the filter actually changes your behaviour or just sits in the document. The Trade Management lesson covers break-even triggers, partial profits, trailing and scaling — all of which are worth revisiting when you change windows, because the hour changes how a move develops.
Finally, journal the hour on every trade, filtered or not. Then in three months your own records can argue with the filter instead of flattering it. If you want to run the same setup in two windows, keep them as two strategies with separate records, so the review is readable.
Trading carries risk of loss. Every decision you take is yours, and nothing on this page is financial advice.