Phases 4 and 5 hold the live-execution material. They do not open because you finished Phase 3, and they do not open because you paid. They open when your own recorded results say they should.
Because the expensive mistake in this field is not building a bad scanner. It is believing a good-looking number.
Every condition below exists because a specific way of fooling yourself has been documented — in this programme, on real student data, more than once. Each one closes a door that has actually been walked through.
Clear all six and you will not have proved you will make money. You will have proved your measurement is honest enough to be worth acting on. That is a smaller claim, and it is the only one anyone can truthfully make.
The same six apply to every student. Nothing here is secret, negotiable or discretionary. No package, tier or additional purchase unlocks them, speeds them up or replaces them.
Your scanner has recorded at least 200 signals carrying a real verdict — WIN, LOSS or FLAT — under the version you are running now.
Below 200, the noise is wider than any effect you could detect. A number built on 40 signals swings 30 points on luck alone. You would be reading weather, not climate.
Not win rate. Money.
Add up what your wins actually earned. Subtract what your losses actually cost. Subtract the round-trip fee on every single trade. The total must be above zero.
Break-even is not 50%.
It is calculated from the real sizes of your wins and losses against the cost of trading. Here are one student's own measured numbers:
Her strategy needed a 71.6% win rate just to break even. Her live scorecard read 61%, which felt like a strong result — and at 61% the strategy still loses money. She measured 48.6% across 5,473 simulated trades. Every coin negative.
A win rate alone cannot tell you whether you are making money. It is one of three numbers. The other two are the average size of your wins and the average size of your losses.
Being barely above break-even is not enough. Your whole uncertainty range has to sit above it, without touching.
Count only the signals that moved. Flats don't count here — nothing happened. Your real result is your win rate plus or minus that wobble.
At 45 moved signals the wobble is about ±15 points. Your "37%" is really somewhere between 23% and 53%. That is not a measurement. It is a rumour.
A result is not a signal until it is more than three wobbles from where it needs to be. Not one. Three.
And the standard does not change direction. One student read 61.8% and was excited. At 1.68 wobbles, not a result. Another read 37.8% and was worried. Same distance, also not a result. If you only apply the rule to numbers you dislike, you haven't learned it.
Split your signals down the middle by date. First half, second half. The edge has to appear in both.
An edge that lives entirely in one week is not an edge — it is a market condition you happened to be present for. Markets change. A strategy that only worked during one trend stops working when the trend does, and you won't notice until it has cost you.
What failure looks like: strong in the first half, nothing in the second. Or the reverse. Either way the pooled number is hiding a story rather than telling one.
You did not adjust a threshold, a hold time, a coin list or any other setting because it made your number look better.
This one needs written confirmation from your mentor. Software cannot check it, because software cannot see why you changed something.
With a few hundred trades and enough settings to try, something will always come up green by luck. Test twenty combinations and one will look excellent for no reason at all. Keep that one, discard the rest, and you have not found an edge — you have found the luckiest setting, and it will not repeat.
Two numbers in your code disagreed and one was wrong. You fixed the wrong one.
Your win rate was 48%. You raised a threshold. Now it reads 54%.
If you are unsure which you did, say so. A student who reports a change honestly keeps their progress. A student who hides one is measuring nothing and doesn't know it.
The ratchet — selftest.py — runs on your machine, passes every check, and has not been modified by you.
The ratchet is an automatic check for every bug this programme has ever been bitten by. Duplicate signals from a broken cooldown. A coin whose smallest price step is bigger than your flat zone. Two files disagreeing about what counts as a real move. A verdict branch that can never fire. Version labels that lie.
Never edit a check to make it pass.
A test suite that only ever goes green is decoration. If your ratchet fails, it has found something — and finding it before you risk money is the entire point.
Two things that are not the same: a ratchet that passes, and a ratchet too old to contain the check that would have failed. The second has happened, to a real student, and it is why she believed a file was correct when it wasn't. Current means which checks are present, not how big the file is.
Nothing bad.
You keep your scanner, your data, your record, your mentorship and every phase you have already completed. You keep collecting. Conditions 01 and 03 both improve on their own as your count grows, provided your grading is running properly.
And if you never pass — that is a real answer, not a failure.
Finding out in three months, on paper, that a strategy has no edge is the most valuable thing this programme can give you. It costs you a course fee. Finding out the same thing with real money costs considerably more, and most people choose that route.
Every measurement this programme has produced says the taught momentum strategy is a coin flip once fees are counted. Eight of them, on different machines, in different countries, by different students. We publish that. It is the result.
The skill you are buying is the ability to reach a conclusion like that honestly, and to know the difference between a number that means something and one that merely looks good.
Can I pay to unlock Phases 4 and 5 early?
No. No package, tier or additional product changes these conditions.
Can I get an exception?
No. The conditions are the same for every student.
Can I use my old data from a previous version?
It is kept and labelled, and you can read it. It does not count toward Condition 01.
My scanner seems to be finding a real edge, but I only have 150 signals.
Then you don't know yet. Keep collecting.
What if I disagree with one of these conditions?
Say so. Ask your mentor to justify it. Every condition here should survive being questioned — if one cannot, it does not deserve to be a rule.