Reviewed October 3, 2026 · Chart-pattern education
A convincing chart is an example. A reproducible test explains what was knowable, what was traded and what happened after costs.
This guide covers research design for ordinary stock-price patterns. Use the visual library for definitions; use this page to evaluate a proposed rule without confusing hindsight with evidence.
Record the data, geometry, timing and execution assumptions.
Include failed signals, costs and uncertainty.
Write a definition another reader could reproduce
“It looks like a triangle” leaves the most important choices unstated. Specify the instrument universe, observation period, bar interval, session, price adjustment and which highs or lows qualify as turning points. Then say exactly what makes a candidate complete, invalid or too ambiguous to use.
Scroll across the table to see all columns.
| Decision | Record before examining outcomes |
|---|---|
| Data | Symbols, dates, bar interval, sessions, adjustment convention and source. |
| Geometry | Allowed boundary slope, number of turning points and tolerance for near-equal levels. |
| Timing | When each turning point and closing signal becomes knowable. |
| Execution | Order type, entry delay, exit conditions, position sizing and unfilled-order treatment. |
| Evaluation | Costs, benchmark, development period and separate evaluation period. |
A turning point may become identifiable later
A high drawn on a completed chart is not necessarily a signal that was available at that high. For example, TradingView documents five bars on either side for its triangle indicator’s 5/5 pivots. The right-hand observations are needed before that particular pivot rule can be confirmed.
Original timing example
If a high occurs at bar 20 and the rule requires five later bars, it cannot be confirmed under that rule until bar 25 has completed. A test that enters at bar 20 using the confirmed pivot has used information from the future. This is an illustration of the timing issue, not a reproduction of a platform strategy.
A rule triggered by a closing price likewise cannot assume the decision was made earlier in that bar. Record the first eligible order time and a plausible fill convention. If a charting tool redraws candidate formations, preserve what it displayed at the time, not only its final historical display.
A high win rate is not the whole result
This invented four-trade sample uses dollar profit or loss per trade. Costs are a flat $5 per completed trade purely to make the example transparent. It is not a backtest, broker fee schedule or expected result.
Scroll across the table to see all columns.
| Trade | Gross result | Assumed cost | Net result |
|---|---|---|---|
| 1 | +$80 | $5 | +$75 |
| 2 | −$60 | $5 | −$65 |
| 3 | +$40 | $5 | +$35 |
| 4 | −$90 | $5 | −$95 |
| Total | −$30 | $20 | −$50 |
Two of four trades win, a 50% win rate, yet the combined net result is a $50 loss. Win rate needs the size of wins, losses and costs alongside it. Report exposure, trade count, drawdown and how much a small number of observations contribute to the total.
Keep development separate from evaluation
QuantConnect’s research guide describes overfitting, look-ahead bias and survivorship bias. Repeatedly changing a rule after seeing the same results can tune it to that sample; selecting only surviving or memorable stocks can distort the evidence.
- Keep a dated specification. Save every version and the reason for each change. Treat a revised definition as another experiment.
- Retain every eligible observation. Include failures and ambiguous cases under a stated inclusion rule. Do not select examples because a later chart looks clean.
- Reserve an evaluation period. Once you use it to tune the rule, it is no longer untouched evidence.
- Model implementation. Include applicable fees, spread, slippage and any financing or borrow constraints. A chart price is not a fill.
- Report uncertainty. A small or highly concentrated sample cannot establish a stable probability. Compare nearby parameter choices and document what was tried.
Separate a planned exit from an executable exit
FINRA explains that a stop-market order can execute away from its trigger price; a stop-limit order may not fill. A planned distance between entry and exit is therefore an assumption, not an enforceable maximum loss.
For a worked distinction between a measured objective and an assumed trade, see the ascending-triangle case study. The position-size tool can explore arithmetic, but it cannot verify a strategy or secure an execution price.
Continue learning
Browse pattern definitions · Work through an ascending triangle
Sources and method
Substantively reviewed October 3, 2026. Diagrams and numerical examples are original, synthetic teaching material. They do not reproduce a security’s trading history, a backtest or a validated strategy. Pattern names describe charting conventions; this article assigns no success probability.
Primary references: TradingView triangle-indicator timing; QuantConnect research guide; FINRA stop-order guidance.
Advertising is separate from this guide. Read our editorial and advertising standards.
What changed in this update
Replaced a duplicate pattern overview with a focused testing guide. Removed unsupported claims of enhanced timing and increased success probability. Added original checked examples for pivot timing and net results. The original URL and publication date are preserved.
Educational information, not personalized investment advice or a recommendation to trade. A diagram or measured-move projection does not guarantee an outcome.