6 / 6 · 5 min
Your own selection and your own journal: the process around the feed
The feed gives raw material; the decision and the accounting stay with you. The last lesson assembles filters, entry rules and a journal into a working process — and explains why we do not show “how many percent worked out”.
The temptation of any formations service is to hang a number next to the card: “works out 62% of the time”. We computed exactly that and took it off the page, and here is why. The share of successes depends on where precisely you enter, where you put the stop and how long you wait. Our simulation of market entry showed one thing, a limit entry at the level quite another, and yours will be a third. A number that cannot be carried over to your trade works not as knowledge but as a promise. So the card carries only the map of the situation, and the measuring stick stays your own journal.
Step one: narrow the feed with filters
Hundreds of cards are born in a day, and trading everything is pointless. The filters at the top of the page really do narrow the stream: kind (start with one, not all five), timeframe, side, minimum turnover and minimum hourly NATR. Turnover cuts off coins where your own order moves the price; NATR cuts off those where reaching the level takes a week. A sensible start is one kind, one or two timeframes, turnover from five million, NATR from one percent.
What to look at on a card before deciding
- Distance to the level: if it is already more than a couple of NATRs, the situation has gone — this is a chronicle, not a candidate.
- Touches: a level tested twice or three times weighs more than a one-off.
- NATR: it also sets a sensible stop. A stop closer than one NATR to the level is knocked out by ordinary noise.
- Turnover: on a thin coin spread and slippage will eat half the idea.
- Market direction: a long on an altcoin while bitcoin falls is a trade against the current.
- Age and timeframe together: a five-minute card an hour old is almost always dead.
Step two: the plan before entry, not after
The card deliberately contains no plan, and that is your job: before entering, write down three numbers — where you enter, where you admit the idea is wrong, and where you are going. For a bounce the entry is usually a limit order at the level, the stop behind the level with about one NATR of room for a sweep, the target the nearest level in the direction of the move. For a break the entry is on confirmation past the level, the stop back behind it. If the three numbers do not add up to a sensible ratio, the situation is not yours, and that is fine: a skipped trade costs nothing.
Step three: a journal that answers questions
A journal is not decoration; it exists so that in a month you can answer which kind and which timeframe bring you money and which bring only fees. The minimum set of columns: date and time, coin, kind, timeframe, level, where you entered, where you exited, the result in units of risk, and one line of “why I entered”. Twenty or thirty entries are already enough to see that, say, hourly retests you hold calmly while five-minute bounces you close in panic.
Looking for the kind that “always works”
None of the five kinds works on its own: what works is the bundle of kind plus market plus your execution. Altcoin bounces live while bitcoin is calm and fall apart in a decline — that is exactly why the algorithm has bitcoin gates on the bounce. Someone hunting for an eternally working kind usually cycles through them and switches precisely when the previous one starts to work.
Tune the feed to yourself: one kind, one timeframe, turnover and NATR. The page address keeps the filters — bookmark it and open straight into your slice. Then start a journal: a table of nine columns is enough.
Tune your own feedWhy is there no success rate on the card if you have the data?
Because that rate depends on entry and exit rules, and everyone's are different. Our internal count exists to tune the algorithm's thresholds: it answers “is a card like this worth creating at all”. Carrying it over to your trade is not valid, and showing it as a promise would be dishonest.
How many trades are needed to tell whether a kind works for you?
Fewer than twenty is not a sample but chance; thirty to fifty entries of the same type already show a stable picture. More important than the count is uniformity: if across those thirty trades the timeframe, the size and the exit rule all changed, there is nothing to compare — you measured something different each time.