6 / 6 · 5 min
The «Break | Bounce» statistics and the place of patterns in your process
The table under the feed is the only judge of all previous lessons. The final lesson teaches reading it without illusions and assembles a working process out of patterns: filters, plan, risk, journal.
The «Statistics» tab shows seven days per kind and timeframe: «Found» and two columns per horizon — «Break» and «Bounce» over 1 hour and 4 hours. One hundred percent is split between those two columns: only DECIDED findings count, and cases where price never left the level by half a typical bar do not enter the denominator — a «draw» does not masquerade as either. The number of the move ALONG the pattern is underlined: for a level breakout that is the «Break» column, for a retest or a liquidity sweep — «Bounce». Next to it in brackets — how many findings are decided: a share of 40 cases and a share of 1,400 are evidence of very different weight.
How to read the table without fooling yourself
First: look at the underlined number, not the bigger one. A 63% bounce on a BREAKOUT is movement against the pattern — an argument to fade breakouts or skip them, not proof that «breakouts work». Second: compare timeframes within one kind — a skew that exists on 1h and vanishes on 5m says more about lower-timeframe noise than any textbook. Third: 52–55% over hundreds of cases is already a real edge by market standards, but it only pays if your risk is smaller than your take; 50% is an honest zero on which nothing is built. Fourth: seven days is one kind of market weather; a skew that survives into next week is worth ten pretty numbers of one.
From table to shortlist: one pass
Open the statistics and look for cells where the underlined number is ≥ 55% with ≥ 200 decided. Say you find «Liquidity sweep · 15m — Bounce 59%» and «Level retest · 1h — Bounce 56%». That is your list for the week: two kinds, two timeframes. Select these kinds in the feed, volume from $5M, your exchanges — and the feed turns from a noisy stream into a queue of candidates, five to ten a day. Each candidate then goes through the plan from the approaches lesson: two branches, stop beyond the level, fixed risk. Everything else in the feed is other people's trades.
The whole process, step by step
- Once a week: walk the statistics, refresh your «kind × timeframe» list with an underlined skew and enough decided cases.
- Every day: the feed filtered to that list; fresh cards are candidates, old ones are material for training your eye.
- Per candidate: a plan before the resolution — both branches, entry, stop beyond the level (past the tolerance, not «by eye»), size from risk, not from confidence.
- After the trade: a journal entry — kind, TF, level kind, plan branch, result. In a month your journal becomes your own personal «statistics».
- Once a month: reconcile your journal with the table. If your results are worse than the kind's statistics — the problem is execution, not the detector.
Rebuilding the list after every red day
A 56% skew means: four trades out of every ten are losers, and they come in streaks. After three losses in a row a beginner decides «the kind broke», changes the list — and goes in circles, always one week late. The cure happens before the trade: compute in advance that at 56% a streak of four losers is routine, and put it into your position size. The list is changed on schedule — weekly, by the table — not by pain.
Open the statistics and do the pass from the example: find every cell with an underlined number of 55%+ and 200+ decided. Write them down — that is your starting list. If there are no such cells today, that is a result too: a week without a skew is a week of watching, and the skill of not trading through it is as real as the skill of entering.
Open pattern statistics«Level breakout · 15m» shows: Break 25%, Bounce 75% (570 decided). What does that mean in practice?
After a 15m breakout, price returns to the level three times out of four instead of continuing. Trading «the breakout» here means paying systematically; the skew itself, though, is material for the opposite trade: entering on the return to the level after the break, in the bounce's direction. That is how the table reads: it does not ban a kind — it tells you which way its gravity pulls.
Why are «draws» — cases where price never left the level — dropped from the denominator?
Because the table's question is «how does this kind resolve: break or bounce», and undecided cases do not answer it. Include them and both shares sag towards forty percent, the sum stops being a hundred, and the table starts answering a different, less useful question — «how often does anything happen at all». One hundred percent split two ways is the choice to count only the duels that actually took place.