The assumption underneath every trading checklist ever written: more reasons, better odds. It is testable, it has never really been tested, and when you test it the line does not go up.
Confluence is the practice of requiring several independent reasons to agree before taking a trade, on the assumption that each additional reason improves the odds.
Do not take a trade for one reason. Wait until several line up — the level, the zone, the session, the round number, the higher timeframe — and take the trade where they meet. Every checklist ever written rests on this, and it feels unarguable. More reasons, better odds.
It is also testable, and testing it needs a base zone that can be found by rule plus factors that are yes-or-no rather than matters of opinion. Fair value gaps are the base: there are thousands of them and their edges are arithmetic. The four factors are the ones a beginner is actually taught to look for.
Score every zone from zero to four, then compare hold rates. If confluence works, the line goes up.
A restaurant has great reviews, a full dining room, an award in the window and a queue outside. Four reasons, all real.
They are also all the same reason wearing different hats — the place is popular. And popular restaurants are noisy, rushed and slow to seat you. Stacking four correlated signals does not multiply your information; it repeats it, while quietly selecting for the conditions those signals share.
The raw answer is dramatic and it is partly an artefact, so both versions are printed here.
| Instrument | 0 factors | 1 factor | 2 factors | 3 factors | Change |
|---|---|---|---|---|---|
| XAUUSD | 85.7% | 80.4% | 79.7% | 75.4% | -10.3 pts |
| GBPUSD | 81.7% | 82.9% | 79.9% | 73.4% | -8.3 pts |
| EURUSD | 82.7% | 76.9% | 82.8% | 73.9% | -8.8 pts |
| US30 | 76.5% | 79.7% | 78.1% | 80.3% | +3.8 pts |
| NAS100 | 81.6% | 79.7% | 77.2% | 66.7% | -14.9 pts |
What was counted: Every hourly fair value gap is scored against four factors a beginner is taught to look for: a round number inside the band, a prior-session high, low or close inside it, the band sitting in the discount half of the last 48 hours for a bullish gap (premium half for bearish), and a displacement candle having created it. Each is a yes or no. Held means price entered the band and was pushed a full band-height back the way it came before reaching the far edge — the identical rule used for the balanced price range.
These are the size-matched figures — only zones in the middle half by width, which is the comparison that survives the confound described below. Four-factor zones are omitted because once the sample is narrowed there are too few of them to report honestly.
On gold, size-matched: zones with no extra factors held 85.7%. Zones with three held 75.4%. Adding reasons was worth -10.3 points — in the wrong direction.
4 of the five instruments declined as factors were added. The exception was US30, which rose 3.8 points. Nowhere did the line climb the way a checklist assumes.
The unmatched figures are far more dramatic: on gold, 86.2% at zero factors falling to 57.3% at 4. That is a -28.9-point collapse, and it would make a much better headline than the real result.
It is mostly about width. A wider band is more likely to contain a round number or yesterday's low simply by covering more prices — gold's four-factor zones had a median width of 75.3 pips against 9.1 for zero-factor ones. And a wider band faces a harder test, because both thresholds scale with it. So factor count and difficulty were entangled.
Restricting to the middle half of zones by width removes most of that. The decline survives at about a third of its raw size. That surviving third is the finding; the other two thirds were an artefact of the measurement, and publishing them as though they were real would have been the most impressive wrong thing on this site.
A decline still needs explaining, and the likely mechanism is not mysterious.
The four factors are not independent. Displacement candles leave big gaps; big gaps span more prices, so they catch round numbers and prior-session levels; and prices that move far enough to do all that are trending. Stacking these does not add four separate pieces of evidence — it selects for one condition, a fast directional market, and that is precisely the condition in which price runs through a zone instead of respecting it.
Look at the factors individually and the same story shows: each one on its own sits close to the others in hold rate. None is a strong filter alone, and combining them mostly compounds what they share.
Dev finds a gold zone he loves. It was created by a huge candle, it holds the 3,840 round number, yesterday's low sits inside it, and it is in the discount half of the week's range. Four out of four. He takes triple his usual size.
What the data says about that zone. Nothing good and nothing terrible — but nothing better than a zone with none of those features. Once width is accounted for, the four-factor group was the weakest one measured on gold, and it had too few members to quote with confidence at all.
What went wrong in his thinking. He counted four reasons. He had one: gold moved fast and left a big imbalance. The round number and yesterday's low are inside the zone because the zone is wide, and it is wide because the candle was big.
What he should have done with the observation. Kept the zone, used his normal size, and used yesterday's low as a target on the way out rather than as a reason to get in.
The habit worth taking from this. Before adding a reason to a checklist, ask whether it can be true when the others are false. If it cannot, it is not a new reason.
Score your setups zero to four on your own criteria before you know the outcome, then compare the hit rates by score. If your line is flat, every hour spent hunting for the fourth reason was spent on nothing.
Score your setups in the backtester →Run the drill in the free backtester → Free, no sign-in to begin.
Does confluence actually improve your win rate?
Not in this measurement. Matched for zone width, gold zones with no extra factors held 85.7% of the time and zones with three held 75.4% — a decline of 10.3 points. Four of the five instruments went the same way.
Why would more reasons make results worse?
Because the reasons are not independent. Big candles leave wide gaps, wide gaps catch round numbers and prior-session levels, and all of it happens in fast trending conditions — which is exactly when price runs through a zone rather than respecting it. Stacking correlated signals selects for one market condition instead of adding four pieces of evidence.
What was the size confound?
Wider zones catch more factors by covering more prices, and they also face a harder test since the thresholds scale with the band. Unmatched, gold fell 28.9 points; restricted to the middle half of zones by width it fell 10.3. Two thirds of the dramatic version was an artefact.
Does this mean I should trade every setup I see?
No. It means the number of reasons is not the thing to filter on. One zone you can define precisely, with a stop you can defend, measured at least as well as an elaborate one — and it will appear far more often.
Should I stop using round numbers and prior-day levels?
Not at all — but use them for the job they do. Prior-session levels are strong targets, which the Key Levels Calculator quantifies. Being useful as a target is a different claim from being useful as a filter.
How do I test my own checklist?
Write your criteria down first, score every setup before you know the outcome, record zone width alongside the score, then group by both. If width explains more of the variation than the score does, your checklist is measuring size rather than quality.
Everything on this site is a reason to like a setup. This page says the way to use them is one at a time, well defined, with an invalidation you can state — not stacked into a checklist that mostly measures how fast the market was moving.
Open the free backtester →Beginner exploration
Open each answer for a plain-language way to read Confluence, test it carefully and decide what to explore next.
This page focuses on “Confluence”.Every trading checklist assumes reasons stack. Scored zero to four across thousands of zones on five instruments and matched for zone width, hold rates fell as factors were added — 85.7% with none against 75.4% with three on gold.For “Confluence”, a beginner should identify what the learning guide measures, assumes or teaches before acting on its conclusion.Treat this page's account of “Confluence” as a learning reference rather than a prediction, signal or promise of future performance.
For “Confluence”, translate the idea into a definition you could apply the same way on two different charts.While exploring “Confluence”, work through one example slowly and record which inputs or observations determined the result.Keep your “Confluence” record honest: list the limitation or counterexample before using the concept in a trading plan.Before leaving “Confluence”, practise the definition on unseen history and review consistency before judging performance.
Turn one idea from “Confluence” into a rule with explicit inputs, dates, costs and pass-or-fail conditions.Ask AI to expose missing assumptions in that “Confluence” test, not to guess the next market move.Use the FXAbsolute AI Backtesting Lab to inspect calculations connected to “Confluence” and the assumptions behind them.Reproduce any important “Confluence” result and reserve unseen data before deciding that an apparent pattern is useful.
Continue your exploration of Confluence with the beginner AI prompt guide, or inspect public calculations in the AI Backtesting Lab.