A horizontal line drawn across an old low looks convincing when the next price move is already visible. The harder question is whether another analyst, with the future hidden, would draw the same line and reach the same conclusion. CME Group explains that prior highs and lows, moving averages and trend lines can suggest support and resistance, but the market need not respect any exact price. This guide turns the idea into a small, falsifiable educational exercise. It contains no live market signal, backtest claim or promise of profitable trading.
1. Define a level before you see what happens next
Choose one instrument, one consistent candle interval, and a historical development window followed by a later evaluation window. Record the time zone, data provider, missing-candle policy and any adjustments to historical prices. For this exercise, a local swing high is a candle higher than the two candles on either side; a swing low is the inverse. The last two candles must close before the swing can be confirmed. Marking it earlier would introduce look-ahead bias.
Propose a support zone around two nearby confirmed lows or resistance around two nearby confirmed highs. This is our experimental convention, not an exchange rule or evidence that the zone will hold. Record the date when the zone first became knowable, then hide the remaining chart. Do not move the boundaries each time price crosses them. If another person cannot apply the written instructions and produce similar zones, the apparent precision is probably a product of hindsight.
2. Use a zone, and choose its width in advance
CME's educational material emphasizes that support and resistance should be treated as areas rather than prices guaranteed to hold to the penny. Suppose two fictional swing lows appear at 99.80 and 100.20 price units. You might predefine a zone from 99.70 to 100.30, a total width of 0.60 unit. This is a fabricated illustration, not an observed trading setup. A price unit is not necessarily a pip, tick, dollar or percentage point.
Check whether the proposed width makes sense relative to recent volatility. A narrow zone may register ordinary price noise as a break, while a very wide zone may count almost every visit as support. One possible rule is to use half an average true range calculated entirely from past candles. Write down the rule before examining outcomes, and do not optimize it repeatedly on the test window. A moving average or round number can provide context, but cannot independently certify the next reversal.
3. Give touches, bounces and breaks separate definitions
Make your categories operational. A touch occurs when a candle's high-low range overlaps the zone. In this example, a bounce means price closes away from the zone within the next three candles. A break means two consecutive closes beyond the opposite boundary. Anything else is unresolved and remains in the dataset. These definitions are deliberately explicit, not universal technical-analysis laws; another researcher may compare alternative definitions in a separately registered test.
An individual candle can enter and leave a zone in ways that the candle summary does not reveal. Do not assume the best possible intrabar execution, and never act on a closing price before that candle has closed. CME describes how an old support area may later function as resistance after a break, but that is a possible pattern rather than a guarantee. Record gaps, major announcements, liquidity conditions and overlapping events so that a tidy chart does not hide difficult cases.
4. Show the denominator and include transaction costs
Imagine twenty independent touches classified under the rules above: eight bounces, seven breaks and five unresolved cases. Bounce frequency is 8 divided by 20, or 40%; break frequency is 7 divided by 20, or 35%; unresolved is 5 divided by 20, or 25%. These counts are invented solely to demonstrate arithmetic. Removing unresolved observations and reporting eight out of fifteen, or 53.3%, changes the denominator. That can be useful for a conditional question only if the exclusion is disclosed.
Consider a second hypothetical paper ledger showing thirty units of gross gains and twenty-five units of gross losses. Add eight units of estimated spreads, commissions and slippage: net outcome equals 30 minus 25 minus 8, a loss of three units. A high bounce rate therefore says little about net results. These figures are not a real strategy backtest, and a paper fill is not evidence that a broker would execute at that price during a volatile session.
5. Separate rule design from evaluation
Split the observations chronologically. Use the first period to decide how swings, zone widths, touches, outcomes and time limits are measured; lock those choices before opening the later period. Keep an event log containing the candle timestamp, boundaries, known information, outcome, assumed costs and a chart snapshot. Do not redraw a zone after a failed trade and then apply its new location to the earlier event. That is data leakage disguised as better analysis.
Check dependence between observations. Five touches during one continuous range may represent one prolonged visit rather than five independent trials. Introduce a documented cooldown or group them into episodes. Compare results with a basic reference case, such as the frequency of similar moves during randomly selected periods of equal length. Small samples, changing volatility and multiple parameter searches can produce unstable patterns. Even a positive out-of-sample observation is not proof of future performance.
6. Make the exercise useful without treating it as advice
Charles Schwab's educational disclosure cautions against relying on technical analysis alone and notes that investing can lose principal. Stops may execute at worse prices during gaps or thin markets, and leverage adds financing, margin and liquidation risks. A diagram showing a tidy bounce cannot establish an executable opportunity. Review instrument specifications and your own ability to absorb loss before considering any real exposure. This article intentionally avoids entry recommendations or return forecasts.
A practical final exercise is to give a colleague the same price history with the future portion hidden and ask them to draw zones independently. Compare their boundaries with yours before revealing the next candles. Large disagreements mean the definition needs work, not that one analyst should quietly adjust the chart. The goal is a transparent, testable question: under stated conditions, how often did a predefined zone coincide with a particular outcome, and what remained uncertain?
Official resources
No online income or trading system guarantees profits. Results depend on skills, costs and circumstances.
Disclaimer ↗