Our method

Published in full, before anything is commissioned. Seventeen steps, the thresholds that trigger a finding, and the points at which a conclusion is withheld for want of data.

Publishing a method invites copying, and that trade is worth making twice over. The method is easy to describe and slow to execute well, which is true of most analytical work. What it buys is the ability to check the reasoning before paying for it, and for a practice with no licence to point at, that check is the strongest thing on offer.

How to read this page

Steps are numbered because the reports reference them by number. When a finding in a review cites step 11, the working behind it is here. Stages group the steps by what they are trying to establish, and each stage discards work that the previous one has made unnecessary: there is no point measuring an edge on a record whose costs have already consumed it.

Stage one · Intake and reconstruction

01 · Ingest and normalise

Statements from MetaTrader 4 and 5, cTrader and TradingView, and CSV exports from exchanges, all reduced to one internal shape: entry time, exit time, direction, size, price in and out, and every cost line kept separate from gross result.

Platform exports disagree about almost everything, including whether partial closes are one position or several. Getting this wrong quietly corrupts every number downstream, which is why it is step one rather than an assumption.

02 · Reconcile against the balance history

The sum of normalised trades has to match the account balance curve. Where it does not, the difference is deposits, withdrawals, manual adjustments or trades missing from the export, and each has to be identified before anything else runs.

This step ends more reviews early than any other. An export that cannot be reconciled is not a record, and no amount of analysis fixes it.

03 · Sample adequacy

Before any metric is reported, the record is tested for whether it can support one. The test is run at whole-record level and again for each instrument, because a 600-trade account is frequently four samples of 150, and conclusions that hold for one do not automatically hold for the others.

Below roughly 200 closed positions, skill and sequence are often indistinguishable at the confidence levels worth acting on. That number is a guide rather than a gate: a method with a high win rate and tight result distribution becomes readable sooner than one whose returns depend on rare large winners.

04 · Period segmentation

The record is split at any structural break the trader reports, such as a change of venue, instrument set, or sizing rule. Comparing across an undisclosed break produces confident findings about two different methods averaged together.

Stage two · Decomposition

05 · Result by dimension

Gross result recomputed by instrument, by direction, by session and hour of entry, by weekday, and by holding-time bucket. Costs are held separately throughout, so that a weak instrument and an expensive one can be told apart.

Long against short on the same symbol is the comparison that most often surprises the person who produced the record.

06 · Distribution and concentration

Share of gross profit produced by the five largest winners, and the record recomputed with them removed. Mean result against median result, both in R multiples, since the gap between the two is itself a concentration measure.

A ratio that can be halved by deleting one row is describing a row, not a method.

Stage three · Costs

Costs come before edge measurement on purpose. Where they have already consumed the gross result, everything after this stage is a description of how the money was lost rather than why.

07 · Cost attribution per round turn

Spread at the moment of entry rather than the venue’s advertised average, commission against the actual size distribution, slippage measured separately for stop exits and market entries, swap on overnight holds, and funding on perpetual contracts.

Exchange exports frequently fold funding into realised result. Where that has happened it is reconstructed from the funding schedule and the position’s open interval, and the report says which of the two it was.

08 · Cost as a share of gross

The output of this step is a ratio computed per instrument, not a total in currency. A total says what was spent. The ratio says whether the method was working for the trader or for the venue, and once it passes 100 percent the answer is settled without any reference to strategy quality.

09 · Cost trajectory

The same ratio across time, since venues change conditions and traders change instruments. A cost problem that appeared in month nine is a different conversation from one that was there throughout.

Stage four · Edge

10 · Expectancy, trimmed and resampled

Expectancy in R multiples on the full record, then recomputed with the top and bottom deciles removed. Profit factor across a rolling window rather than reported once. The trade list resampled with replacement several thousand times, with the middle range of outcomes reported instead of a single figure.

The resampled range is the number that changes decisions. A point estimate invites a reader to act on two decimal places that the sample cannot support.

Schematic — illustrative only, not client data.

Stage five · Risk

11 · Realised risk against stated risk

Risk per position measured from the balance history at the time of entry, then compared with the figure the trader believes was used. The two differ in most records, and the direction of the difference matters more than its size.

Where the gap concentrates in a subset, such as positions opened after a losing session, the finding moves to stage six and is quantified there.

12 · Drawdown mechanics

Depth and duration of the worst sequence, the gain required to recover from it, and what the same sequence would cost at the size currently traded rather than the size that produced it.

The recovery arithmetic is not decoration. A 30 percent drawdown needs 43 percent to get back, and traders who scale after a good period tend to have modelled the first number and not the second.

13 · Correlated exposure

Correlation between positions held at the same time, computed on the actual overlap intervals rather than on daily closes.

Four positions on correlated instruments is one position at four times the size, and sizing models that treat them as independent discover this during the week it matters most.

Stage six · Behaviour

14 · Grouping by preceding result

Positions grouped by the outcome of the previous session, then compared on median size, frequency and win rate. Where size moves and win rate does not, the record has documented a reflex.

The elevated subset is then re-run at baseline size, which converts the observation into a figure in currency. Findings in this stage are always stated as money, never as advice about temperament.

15 · Duration asymmetry and rule breaks

Median holding time on winners against losers, entry clustering in narrow windows, position count per day against the trader’s own stated limit, and the cumulative cost of documented breaches across the sample.

Stage seven · Writing

16 · Confidence grading

Every finding is graded before it is written, and the grade travels with it into the report. Nothing is stated more strongly than its evidence allows.

Grade	Meaning	What supports it
Firm	Acted on with confidence	Adequate sample, effect survives trimming and resampling
Directional	Points one way, treat as provisional	Sample adequate but effect sensitive to trimming
Contextual	Worth knowing, not worth acting on alone	Observed but not separable from variance
Withheld	Stated as not established	Sample too short, or data quality prevents the test

17 · Internal challenge and delivery

Each firm finding is argued against before the report is sent: what would have to be true for it to be wrong, and does the record contain that. Findings that do not survive the exercise are downgraded rather than deleted, and the downgrade is visible in the document.

Where the balance of evidence points to stopping a method, the report says so in those words. That outcome is priced into the engagement from the start, which is the practical meaning of taking no commission from anyone’s trading volume.

Thresholds published in advance

Fixed before a record arrives, so that a threshold cannot be adjusted to produce a more saleable conclusion.

• Sample adequacy is assessed at roughly 200 closed positions for whole-record findings, and per instrument for instrument-level ones.
• Cost ratio above 100 percent on an instrument with an adequate sample produces a firm finding regardless of anything else in the record.
• A size difference between behavioural subsets is only reported when the accompanying win-rate difference sits inside the noise band for those sample sizes.
• Resampled ranges are reported at the middle 90 percent of outcomes. Where that range crosses breakeven, the finding is graded no higher than directional.
• Any metric that changes by more than half when the top and bottom deciles are removed is reported with both values, never with one.

What the method does not include

• No forecast, no market outlook, no view on any instrument.
• No recommendation to buy, sell or hold, and no personal advice.
• No verification of the venue: we do not audit brokers, confirm custody, or check that reported fills occurred.
• No psychological assessment. Stage six measures behaviour and prices it. It does not interpret it.
• No optimisation service. Where a method needs rebuilding, that is a separate piece of work and it is not what a review delivers.

Data handling

Accepted: statement exports, CSV files, and read-only exchange API keys where an export is not available. Never requested at any stage: account passwords, keys carrying trade or withdrawal permission, wallet seed phrases, remote access to a terminal.

Read-only keys, where used, are revoked by the client on delivery. Working files are held for the period set out in the engagement terms and then deleted. Records are never used as examples, redacted or otherwise, which is why the sample report on this site is constructed rather than borrowed from someone.

Versioning

This page is the method as it currently stands, and reports cite the version they were produced under. Changes are additive where possible: a metric is added, a threshold is tightened, a step is split. Where a change would alter a past conclusion, that is noted rather than absorbed quietly, since a method that revises itself without saying so is not a method that can be checked.

Next

The sample report applies every stage above to a constructed record, and cites these step numbers as it goes. Reading the two together takes about half an hour and answers most of what a first conversation would otherwise cover.