Research · 4Xedge Bias
My Account My Terminal
FOLLOW
{{ brand }} BIAS TERMINAL
Home Framework Watch Research Support
Free account!
NEW · IN YOUR ACCOUNT · INCLUDED FREE

Ask Your History a Question

& Read the Answer as a Report

Research asks the past five years a question written in the numbers your own configuration already prints — your studies, your timeframes, your levels. It answers with a report: how often that reading occurred, when, how long it lasted, and which of your slots were behind it. It never tells you what happened next*.

Create your free account See the five tabs
* A historical count is a measurement of what your studies printed, not a forecast and not a recommendation. Trading involves substantial risk of loss and is not suitable for all investors. Past performance is not indicative of future results.
RESEARCH · IN {{ appName }}

Not a backtester. The same measurement, pointed backwards.

One symbol at a time, five years back, and never a verdict.

A backtester grades a strategy. Research does something narrower and more honest: it finds every bar where the reading you described was true, and describes those bars. There is no entry, no exit, no equity curve and no accuracy figure, because the moment a tool scores an idea it has started holding an opinion about your trading.

What you get instead is context for a reading you already trust. You have watched your stack line up at +90 and wondered whether that is a monthly event or a Tuesday. Research answers that in your own numbers, on the instrument you actually trade, over a window you choose — and it shows the arithmetic underneath every count so you can argue with it.

It lives in your account rather than in the terminal, on the rule that holds everywhere else: live market data on the terminal, everything else behind the account nav. Research is not live.

HOW A QUESTION RUNS
3 STAGES
01
THE QUESTION Your study set, then up to four conditions that all have to be true on the same bar, then the timeframes and the window.
bias ≥ +90 AND agreement ≥ 5 of 6 HELD FOR 3 BARS
02
THE ANSWER Two counts, neither derived from the other: a run of forty qualifying bars is one occurrence to a trader and forty to a naive counter.
47 EPISODES
312 QUALIFYING BARS
8/10 REPORTING HOLES NAMED, NEVER FILLED
03
THE COMPARISON Run the same question on a second symbol and set the two reports side by side. One thing varies, everything else holds.
EUR/USD · 47 XAU/USD · 61 ONE AXIS · SYMBOL

Five tabs, no step numbers

BUILD, CONDITIONS, REPORT, COMPARE, SAVED. Numbering them would only be a sequence on your first visit — after that you usually arrive through a question you already saved, which would make you start at step three. So the order is taught by state instead: REPORT stays dim until something has run, COMPARE until a comparison is open, and each then carries its count.

BUILD THE QUESTION
EUR/USD
STUDY SET
Session Range RSI 14 B Bands 20 CUSTOM · FROM TREND STACK
CONDITION
01 Bias score at or above +90 ⇄ HELD FOR 1 BAR
TIMEFRAMES
M15 H4 H12 D1 W1
02 Sep 24 → 31 Jul 26 4,180 / 6,000 BARS RUN QUERY
1 TAB · BUILD

Build

Five blocks in the order the sentence reads: symbol, study set, condition, timeframes, range. The study set sits above the condition because the condition is written in terms the set produces — load a Matrix preset as a copy or build one here, one or the other, never half of each.

The window is capped in bars per timeframe, never in calendar days: a single date limit would be generous for D1 and absurd for M15. Go past it and RUN QUERY becomes OVER BUDGET, states the arithmetic and the two ways out, and applies neither.

ILLUSTRATIVE · NOT LIVE MARKET DATA
CONDITIONS YOUR LIBRARY
8
bias ≥ +90 2 QUERIES
agreement ≥ 5 of 6 bullish 1 QUERY
RSI crosses above 70 RSI overbought
reporting ≤ 5 NOT USED
WHAT YOU ARE ASKING ABOUT The bias score One study in the set WHAT YOU ARE LOOKING FOR It was lined up It turned
RESOLVES TO Bias score above or below a level
2 TAB · CONDITIONS

Conditions

A condition is one testable statement — a level, a band, an agreement count — and it is yours, reusable across every question you ask. No symbol, no window, no answer attached. Your library shows how many saved queries ask each one, so correcting a level once corrects all of them.

You never have to know the mechanics. Say what you are asking about and what you are looking for — the bias score, it was lined up — and the page works out which kind of statement that is and shows you, read-only. Where a pairing has no honest reading it says so and names where that question belongs instead.

ILLUSTRATIVE · NOT LIVE MARKET DATA
REPORT FROZEN RESULT
ran 04 Sep 26 09:12 GMT
OCCURRENCES 47 EPISODES · 312 BARS
SHARE OF EVALUATED 2.4% OF 12,910 BARS
PER TIMEFRAME
H4 4,290 − 17 = 4,273
H12 1,430 − 6 = 1,424
D1 477 − 0 = 477
bars in range − holes = bars evaluated. Every rate on the page divides by the last of those.
NO OUTCOME COLUMN What price did next is not on this report, and not by omission.
3 TAB · REPORT

Report

Frozen at the moment it ran, carrying the study set, the parameters, the timeframes and the resolved window that produced it. Edit the question tomorrow and this answer does not move — re-running gives you a second report, because otherwise a bookmark quietly becomes a false statement.

Below the counts is the deep read: where your level sits in this instrument's own distribution, how far past it each episode peaked, which of your slots agreed on the matches and across the whole window with the difference printed, the per-row funnel showing which condition is doing the filtering, and how much of the finding rests on partial evidence.

ILLUSTRATIVE · NOT LIVE MARKET DATA
COMPARE NO RECOMPUTE
5 REPORTS
SYMBOL 5 values AXIS
CONDITION bias ≥ +90 HOLDS
TIMEFRAMES H4 H12 D1 HOLDS
ENGINE & BASIS local v1.4.2 / house / bid HOLDS
EUR/USD 47
XAU/USD 61
GBP/USD 29
ORDER, NOT RANKING — THE TOP ROW IS NOT A PICK
4 TAB · COMPARE

Compare

A comparison reads reports you have already run and recomputes nothing. It varies exactly one thing — usually the symbol, everything else holding — because if two dimensions vary, a difference in the counts cannot be attributed to either.

So the button carries the refusal rather than drawing bars with a caveat: NOTHING VARIES, SELECT TWO OR MORE, NO COMPARISON TO OPEN when two axes move, or BELONGS ON THE REPORT when it is really two timeframes of one instrument. Where the windows or the completeness differ, it compares rates instead of counts and says which it is doing.

ILLUSTRATIVE · NOT LIVE MARKET DATA
SAVED YOUR LIBRARY
4 · 8 · 2
SAVED QUERIES
Strong bull, all TF bias ≥ +90 RELATIVE 5 reports
Flip on thin evidence flip to bullish AND reporting ≥ 8 1 report
SAVED REPORTS
EUR/USD H4 H12 D1 · 02 Sep 24 → 31 Jul 26 47
XAU/USD H4 H12 D1 · 02 Sep 24 → 31 Jul 26 61
RELATIVE BY DEFAULT A question saved as “last 2 years” still means the last 2 years next spring. RUN ON… resolves it once, then hands the same absolute window to every symbol in the batch.
5 TAB · SAVED

Saved

Three libraries, because they are three different things. A query is a question with no symbol and no answer. A report is a question plus a symbol plus a window plus the result. A comparison is a list of reports and stores no figures of its own.

Ranges are relative unless you pin them, so a question you saved keeps asking about the recent past rather than about a fading window — and RUN ON… resolves that range once for a whole batch, since reports whose windows differ cannot be compared.

ILLUSTRATIVE · NOT LIVE MARKET DATA
ONE DENOMINATOR FOR ONE WORD

A hole is never a zero, and a share is never quietly inflated.

Feeds have gaps. Studies stop reporting. A tool that treats an unevaluable bar as a bar where nothing happened will tell you your condition is rarer than it is, and it will never mention why. So the three counts are separate, named, and printed together — and every rate and share on both tabs divides by the last one.

THE WHOLE DENOMINATOR bars in range − holes
———————————
= bars evaluated The per-timeframe table gives you all three, so the row reconciles on screen rather than in a footnote. Where a finding rests on partial evidence, the report says how much of it does.
× {{ p.t }} — {{ p.d }}
These are not features waiting to be built. Each one was considered and left out, because the moment a historical count is dressed as a result it stops being a measurement and starts being a claim about your trading.

Ask it about your own reading

Research is part of the free account, on the account page after Data source. Build the question in the studies you already run, and read the answer as a report you can keep, export and compare*.

* A historical count is a measurement, not a signal, and does not guarantee profitable or successful trades.

Trading involves substantial risk of loss and is not suitable for all investors. You should make sure you understand the risks involved, seeking independent advice if necessary. Past performance is not indicative of future results.

Get a free account

Straight answers

{{ q.a }}
{{ brand }} BIAS TERMINAL
A multi-timeframe bias terminal for currency traders. Information only — not financial advice, not a signal service, and never a recommendation to enter or exit a position. You trade at your own risk.
Trader Risk Warning | Privacy Policy | Terms of Service | Copyright © 2026 – {{ appName }} · every price, score and count on this page is illustrative placeholder content, not live market data.