Module 3E
Probability thinking: conditional probability and the confluence stack
Leads into 3F, which governs what happens once you are in the position.
Grimes, A. The probabilistic framing follows his published work. One passage below is quoted directly; where his argument is used without quotation marks, that is deliberate and explained in the note on sourcing.
The most honest starting point in the curriculum
Adam Grimes opens his statistical framework with a claim that deflates most retail trading mythology: most of the time markets are very close to efficient, which means most price movement is random and there is no technical reason to be involved in it.
That is not pessimism. It is the foundation of sound probabilistic thinking, and it changes your whole orientation. You are not looking for trades. You are waiting for the specific conditions where randomness is temporarily overcome by something detectable and repeatable.
There are repeatable patterns, which is the good news and the reason any of this works. The edge is real. It is also small, and understanding exactly how small is what this module is for.
- Edge
- A slight, repeatable tilt in probabilities. Not a guarantee, and it only shows up over a sample.
- Sample size
- The number of trades before results mean anything. Twenty is a floor, not a target.
- Hypothesis vs prediction
- A setup is something you test with defined risk, not a forecast you are owed.
- Confluence stack
- Several aligned conditions raising the odds. Three to four beats any single signal.
- Gambler's fallacy
- Believing past results change the next independent trade. Runs both directions.
- Waiting from conditions vs fear
- Not trading because criteria are unmet, versus not trading because you are scared.
- Volatility clustering
- Big moves follow big moves, so a fixed stop can be too tight after an expansion.
What an edge actually is
An edge is not a system that always works and it is not a setup that guarantees a win. It is a slight tilt in the probabilities: a condition under which the expected outcome of a defined action is slightly better than random, over a large enough sample.
"All we have, all we can possibly have, is a slight tilt in the probabilities, a suggestion that there may be a slight departure from randomness over the lifetime of any trade we put on."
Adam Grimes
The practical implication is that a single trade proves nothing. Winning one does not confirm your edge is real. Losing one does not confirm it is broken. The edge only reveals itself across a meaningful sample, and Grimes puts the floor at around twenty occurrences before drawing any conclusion at all.
This is why the journal is not optional. Without documentation of every trade and its conditions, you cannot tell whether your results are tracking a real edge or variance. A trader who wins six of ten cannot distinguish skill from luck. A trader who wins 58 of 100 with consistent condition logging is starting to see something.
Hypothesis, not prediction
A beginner treats a setup as a prediction. This looks like the setup, it is going to go long from here.
A probabilistic trader treats it as a hypothesis. This has the structural conditions that have historically produced a better-than-random outcome. I am taking a position with defined risk to test that in this instance.
Those sound similar. The behavioural difference is enormous.
The prediction-based trader feels validated when they win and betrayed when they lose. The market should have done what the setup suggested, so a loss means something was wrong: the analysis, the setup, the market, the timing.
The hypothesis-based trader gets neither validation nor betrayal from a single outcome. A clean trade that hits its stop is a correctly executed test. A sloppy trade that happens to profit is contaminated data. What matters is whether the conditions were present and whether the execution was correct, independent of outcome.
This is the concrete mechanism underneath "process over outcome." It is not an aspiration or a slogan. It is how probabilistic thinking actually operates, and it is why the trade log grades decision quality rather than results.
The confluence stack
Conditional probability asks: given that X is true, what is the probability of Y? In trading: given a specific set of conditions, what is the probability of a favourable outcome?
A single signal in isolation has weak edge and sits close to random. Aligned signals compound.
This stack mirrors the five-layer read from 3C. Context and location are evaluated in the pre-trade write-up. Confirmation is evaluated in real time when price arrives.
| Condition | Layer | What you are checking |
|---|---|---|
| Higher timeframe bias | Context | Structural read aligned with the trade direction |
| Profile position | Context | Open relative to prior value. Inside or outside value, inside or outside range |
| Opening type | Context | Identified from the first thirty minutes, and supporting the expected delivery |
| Session context | Context | Inside an active window, not mid-session drift. News calendar checked |
| Location quality | Location | A mapped location: liquidity pool, prior value level, gap, order block, VWAP |
| 1. Delta direction | Confirmation | Immediate aggression aligned with the trade direction at the level |
| 2. Cumulative delta | Confirmation | Making new extremes with you, or diverging against the prior trend |
| 3. Volume and size at the level | Confirmation | Real volume present, and large prints showing meaningful participation |
| 4. Absorption or compression | Confirmation | Defence visible at the level, or price compressing before directional commitment |
| 5. Auction state | Confirmation | Exhaustion supports reversal. Strength supports continuation |
| 6. Tape signature | Confirmation | A speed change confirming participation has shifted |
| 7. Structure shift | Confirmation | The candle chart showing a change of character. One factor, never a gate |
| Risk placement | Execution | Stop behind a structural reason, not a round number of points |
| Path clarity | Execution | Fixed target by default. Extension only when the path read justifies it |
This is the canonical list, and the numbering matters. There are seven confirmation factors. That is your denominator, and you need a fixed one, because a threshold of "three to four" is meaningless if the list keeps changing size.
Other modules describe some of these at finer grain, splitting volume from large prints, or absorption from compression. Those are the same seven factors described in more detail, not additional ones. When you count, count against this table.
A trade with all four context conditions, a clear location, and three or more aligned confirmation factors is a materially different proposition from one with two context items and no confirmation read at all.
You do not need every factor. The skill is knowing which three or four are doing the work at this location for this setup.
The dangerous case, and it is worth reading twice. Low-condition trades will occasionally win on random variation. That is the worst possible outcome for a developing trader, because it reinforces taking low-quality trades on the grounds that sometimes they work.
The sample needed to reveal that the edge is not there is far larger than the run of wins that felt like confirmation. You can be wrong about this for a long time and be paid for it the whole way.
Waiting is a decision
Choosing not to enter when conditions are not met is a form of execution, and it is one of the hardest to practise consistently.
Grimes puts it about as directly as it can be put: a specialist who trades one setup has to accept that only a few market environments favour it, and their job is to wait for those environments. He goes further, and says you can redefine your job description to include not trading most of the time.
That is one of the most liberating reframes available. Your job in a session is not to find a trade. It is to assess whether the conditions for a qualifying trade are present. Most sessions they will not be, and "no trade today, conditions not met" is the job done correctly rather than a day wasted.
Two kinds of waiting, and only one is discipline
Waiting from fear. "I see a setup but I am afraid it will not work, so I am not entering." Outcome attachment producing avoidance. Emotionally driven and inconsistent, and it will let you skip good trades and take bad ones in the same week.
Waiting from conditions. "The setup does not meet the criteria I wrote down. No trade." Process execution. The decision was made before the session, in a neutral emotional state, and the live session is only observing whether the market supplies what you specified.
The write-up exists to produce the second kind by design. Without it, the two are indistinguishable from the inside, which is precisely the problem.
The gambler's fallacy, in both directions
After losses. Three losers in a row, and a trader sizes up or lowers their threshold because they feel due. The market has no knowledge of your recent results. Each trade is structurally independent, and the probability of the next setup is exactly what its conditions suggest.
After wins. Three winners in a row, and a trader sizes up or relaxes conditions because they feel hot. Equally false, and more dangerous, because it feels like confidence rather than delusion.
The second version connects to something 1D already showed you. The moment you start trying to maintain peak performance, you have disrupted it, because trying to hold onto a state is a different activity from being in it.
The correct calibration is boring: every trade at the same size with the same conditions standard, regardless of recent results. The conditions qualify the trade. Nothing else does.
Volatility clustering, and one honest complication
Large moves tend to follow large moves, and quiet periods follow quiet periods. That is a structural feature of real markets rather than a pattern you have to detect.
It has a direct implication. After a session with an unusually large directional move, expect elevated volatility to continue. Not a prediction of direction, an expectation about magnitude.
And here is the complication this curriculum has to state plainly. A four-point stop that is appropriate in a normal session may be structurally inadequate where the average range is three or four times that. In an elevated-volatility environment the same stop gets hit by noise rather than by being wrong.
The honest response is not to widen the stop, because 3D showed exactly what that costs. It is to recognise that some conditions do not suit a fixed-stop structure at all, and to stand down. Volatility assessment belongs in your context read, not in your stop placement.
Self-check
Grade yourself honestly
1. You take a trade with two confirmation factors aligned and it wins. What have you learned?
A correct answer is nothing, and that is the dangerous part. Low-condition trades win sometimes on variance, which reinforces the habit that produced them. The sample needed to prove the edge is absent is much larger than the run of wins that felt like proof. If your answer was that it validates the read, this is the exact trap the module was built around.
2. How many confirmation factors are there, and what is the threshold?
Seven factors, three to four aligned. The number matters because a threshold without a fixed denominator is not a threshold. If you were not sure of the count, go back to the table, because you cannot apply "three of seven" while quietly working from a list of five or nine.
3. In the moment, how do you tell waiting from conditions apart from waiting from fear?
A correct answer points at the write-up. If the criteria were specified before the session, you can check whether they are met and the answer is not a matter of feeling. Without that, the two are indistinguishable from the inside, which means you will always be able to call your avoidance discipline.
4. It is the session after a huge trending day and range is running several times normal. Your setup appears. What is the problem?
A correct answer says the fixed stop may now be inside the noise, and that the correct response is to stand down rather than widen it. Widening changes the ratio for every trade you take. This is a condition problem, and it belongs in your context read.
If question 1 caught you
It catches nearly everyone, and it is the most expensive idea in this module. A win is the most persuasive possible evidence for a habit, and it arrives whether or not the habit is any good. That is why decision quality gets graded independently of result, and why the log records the conditions rather than only the outcome. Without that separation you will be taught by your own worst trades, in the wrong direction, for years.
One rep before you continue
Score three setups you did not take
Find three setups from the last two weeks that you looked at and passed on. Not trades you took. Ones you skipped.
For each, score it against the table. How many of the four context conditions were present. Was there a clear location. How many of the seven confirmation factors aligned.
Then answer one question honestly for each: did I pass because the conditions were not there, or because I did not want to be wrong?
Three setups, four numbers each, one honest sentence. The scoring is easy and the last sentence is the rep.
What the next module does with this
3E judged the setup before entry. 3F governs what happens once you are in it, which is where most of the damage in this curriculum's own trading records was actually done.
Four valid exits, and everything else is emotion wearing a reason. It is a shorter module than this one and it protects more money.
Log the module
Where did you get stuck?
Not a test, and nothing is marked. This module has the confluence table that everything downstream counts against, so I want to know if the denominator landed. About a minute.