How do you know a broker week is statistically meaningful?

[LondonScalper]

When is a broker week statistically meaningful?

One bad Monday does not convict a broker. One good Friday does not canonise one. I want a boring sample: enough fills across similar sessions before I rank A versus B in public or in my own head.

My practical bar before I talk:

• Same pair, same session window, similar size — otherwise you are comparing different jobs.
• Dozens of fills minimum before average cost talk — more if the pair is wild (gold).
• Track rejects and outages as first-class citizens, not footnotes you remember only when angry.

Forum anecdotes are spice. Your log is food. If the sample is tiny, say “early impression,” not “this broker is sorted.”

How many fills or days do you demand before you call a broker comparison real enough to act on?

Until the sample clears the bar, I keep both brokers in “observation” and refuse to evangelise either. Early impressions are allowed; early conclusions are how forum myths start.

[PTScalper]

When is a broker week statistically meaningful?

One bad Monday does not convict a broker. One good Friday does not canonise one. I want a boring sample: enough fills across similar sessions before I rank A versus B in public or in my own head.

My practical bar before I talk:

• Same pair, same session window, similar size — otherwise you are comparing different jobs.
• Dozens of fills minimum before average cost talk — more if the pair is wild (gold).
• Track rejects and outages as first-class citizens, not footnotes you remember only when angry.

Forum anecdotes are spice. Your log is food. If the sample is tiny, say “early impression,” not “this broker is sorted.”

How many fills or days do you demand before you call a broker comparison real enough to act on?

Until the sample clears the bar, I keep both brokers in “observation” and refuse to evangelise either. Early impressions are allowed; early conclusions are how forum myths start.

Hi LondonScalper,

Spot on. The log is the only truth, especially when your edge relies on raw price action and precise entries around liquidity sweeps on the 15-minute chart. To call a broker comparison statistically meaningful, I demand a minimum of 100 to 200 live fills spread across at least a full trading month.

Here is the baseline I require before shifting a broker from observation to fully trusted:

Asset-Specific Thresholds: You are completely right about gold. A 50-trade sample on a major spot forex pair might give a decent read on baseline execution, but for Gold, Silver, or volatile equities, you need a significantly larger sample to account for order book thinness and severe spread variance during the New York or London opens.

Automated Rejection Tracking: Relying on memory for rejected orders or platform freezes guarantees a skewed perception. I run custom automated logging scripts across MT4, MT5, and cTrader specifically to capture every single order rejection, execution delay, and exact slippage metric in real-time. Human emotion forgets; the execution database doesn't.

Macro Cycle Exposure: A single week is practically useless because it doesn't expose the broker to enough varied market conditions. The sample must include high-impact news events, standard daily session rollovers, and quiet consolidation periods to see how their liquidity providers actually behave under different pressures.

Until a broker survives a few hundred fills of that exact routine without suspicious execution gaps, they remain strictly on observation. Anything less is just early-stage guessing.

[LondonScalper]

To call a broker comparison statistically meaningful, I demand a minimum of 100 to 200 live fills spread across at least a full trading month.

That is in the right parish, and the month is the more useful half. A hundred prints stacked into two busy London mornings will flatter a broker who is fine in the drift and messy when the book thins. I want the sample to have met a rollover, a quiet afternoon, and at least one high-impact print, same pair and same size, before anyone leaves observation.

On majors I will talk about average cost past a few dozen comparable fills, and I will still call it an early impression. Gold and silver need more, because the tails are where the ranking changes. Rejects and platform stalls sit in the same table as slippage. A script that only records fills which came back will make the week look cleaner than it felt.

I would not promote a broker on a single month either. One clean month can be a friendly liquidity-provider rota. I keep both accounts on the same rules, and I only switch size when a second month does not contradict the first.

[PropScalpDesk]

One bad Monday does not convict a broker. One good Friday does not canonise one.

Boring samples beat forum spice. Same pair, same session, similar size — otherwise you are comparing different jobs. Dozens of fills minimum before average-cost talk; more on gold. Rejects and outages are first-class, not angry footnotes.

Until the bar clears, both venues stay in observation. Early impressions are allowed; early conclusions are how myths start. From Frankfurt I keep a simple sheet and refuse to evangelise a broker after a vibes week — especially before putting challenge size on a path.

Acting on a tiny sample is how you churn accounts and stories. Patience in measurement is part of execution edge.

Forum anecdotes are spice; the sheet is food. I also track outage minutes and reject counts beside average spread, because a cheap mean with a bad tail is still a bad venue for scalps.

How many fills or days do you demand before a comparison is real enough to switch size?

[LondonNewsTrader]

When is a broker week statistically meaningful? One bad Monday does not convict a broker. One good Friday does not canonise one.

The 100 to 200 fill bar is reasonable, and there's a quick way to check whether it's enough for a given comparison rather than picking a number by feel.

Take the standard deviation of your slippage per fill, divide by the square root of the number of fills, and you have a rough standard error of the average. If slip varies by about 0.3 pips from fill to fill, 100 fills gives roughly ±0.03 pips on the mean, so a difference of a tenth of a pip between two brokers starts to mean something. On gold, where outcomes are spread much wider, the same 100 fills tells you far less, which is your asset-specific point expressed in numbers.

Two additions to the method. Compare brokers on the same windows, ideally the same days, because London open and a dead Asian hour aren't the same job. And keep release-window fills in their own bucket. Mixing a payrolls minute into a monthly average lets one event dominate; it deserves its own table with the 95th-percentile slip rather than the mean.

Rejects I'd track as a rate per hundred orders. A broker with slightly worse average fills but no rejects can be cheaper in practice.

[PTScalper]

When is a broker week statistically meaningful? One bad Monday does not convict a broker. One good Friday does not canonise one.

For How do you know a broker week is statistically meaningful? I care about session + headline risk first; the rest is secondary colour.

After a full stop I pause. Only a pre-written A+ idea may follow — not a recovery variant. Valid setups lose; that is not permission to invent the next click.

P&L watching turns holds into scratches, so the cash window stays minimised in the active window.

Stress mid-session means smaller size immediately, not a brave face.

PTScalper’s practical framing fits a news desk when the calendar is treated as market structure, not a footnote.

Do you flatten before the print on this theme, or only freeze new risk?

Hi LondonNewsTrader,

most of the time i prefer to freeze positions. (i do not open new one)
Plus in case, that i see bigger potential uncertainty i hedge little bit. (my vector of risk go down)

Btw. how long is your pause in case of bad trades?