A database is a decision tool
A folder full of spectacular winners can motivate you, but it cannot estimate how a setup behaves. The useful database includes the clean examples, the failed versions and the candidates you rejected. Together they train two separate skills: recognizing a pattern and deciding whether the current version is good enough.
Start with one setup. Mixing breakouts, episodic pivots and parabolic reversals into one unstructured collection makes comparison difficult because each has different context, triggers and failure modes. A focused sample lets repeated details emerge.
Record the same fields every time
Consistency matters more than the note-taking app. A spreadsheet, local folder or notes database can all work if each record answers the same questions.
Identity and date
Ticker, market, setup type, trigger date and broad market condition.
Precondition
Prior move, base duration, percentage gap, extension or catalyst relevant to the setup.
Participation
Volume relative to normal, liquidity, spread and whether the sector was acting in sympathy.
Decision levels
Planned trigger, stop reference, stop distance, target logic and actual execution if traded.
Outcome
Maximum favorable and adverse movement, time to failure or follow-through, and rule adherence.
Sample the setup, not only the outcome
If you search for stocks that rose 200% and work backward, you will learn what historic winners looked like but not how often similar-looking stocks failed. That is useful discovery work, not a complete test.
Build a forward sample too. On a fixed day each week, save every candidate that meets your written criteria before you know the result. Review them after five, ten and twenty sessions. This reduces the temptation to quietly exclude weak outcomes.
- Keep successful and failed triggers in the same collection.
- Label rejected candidates and state the reason for rejection.
- Separate market regimes so a strong year does not dominate the sample.
- Do not change the setup definition halfway through a sample without starting a new version.
A manageable weekly routine
- Choose one setup and one historical period.
- Review a fixed universe rather than jumping between memorable stocks.
- Capture ten to twenty candidates with identical chart scales.
- Complete the standard fields while the context is fresh.
- Tag one recurring strength and one recurring warning.
- At the end of the month, summarize the group rather than the best trade.
Volume beats intensity. One hour repeated every week produces a more reliable sample than a single weekend of saving hundreds of charts without notes.
Questions the database should answer
After enough records, search for differences between clean follow-through and immediate failure. Did the best breakouts have shorter bases? Did successful EPs arrive after quiet months? Did parabolic reversals require a failed VWAP reclaim? Were losses concentrated when the broader market was weak?
The answers are hypotheses, not laws. Write them as testable conditions and review the next sample. The database becomes valuable when it changes a future watchlist or prevents a low-quality entry, not when its chart count becomes impressive.