Technical Analysis · Chapter 4 of 12
Support, resistance and supply-demand zones
Why some levels genuinely matter, why most drawn ones do not, and the control that separates the two.
Support and resistance is the oldest idea in technical analysis and the one where the gap between the folklore and the evidence is widest. There is a real effect underneath it. It is much narrower than what is usually taught.
The three mechanisms that actually produce a level
- Resting orders. If a large participant has a limit order at a price, that price genuinely resists. This is the only mechanism that is unambiguously real, and it is invisible — you cannot see the order book history, so you are inferring it from price behaviour after the fact.
- Coordination. If enough participants watch the same obvious level, their collective orders create the effect. This makes the level self-fulfilling — and it has a sharp implication that most people get backwards: the more obvious a level, the more likely it is to matter. Round numbers, the prior day's high and low, the 52-week high, the opening range, the previous close. A clever level that only you have drawn has no coordination behind it and therefore no mechanism.
- Position memory. People who bought at a price and are now underwater tend to sell when they get back to break-even. This produces resistance at prior highs and is a genuine behavioural effect with reasonable support in the literature on the disposition effect.
The control that most level analysis fails
The claim is “price reacted at my level”. The control is: how often does price react at a randomly chosen level a similar distance away?
The answer, in almost every market, is: quite often. Price oscillates. Any level within normal daily range will be approached and departed from repeatedly, and a ‘reaction’ defined loosely enough will be found at nearly all of them.
So the test is not ‘does price react at my level’ — it is ‘does price react more at my level than at a distance-matched random one?’ When people run that test properly, the effect at prior extremes and round numbers usually survives at a reduced size, and the effect at drawn, subjective levels usually does not.
Being able to describe this control is worth more in an interview or a research conversation than any level-drawing technique.
Zones rather than lines
A single price is almost never the right object. Markets are noisy, and a level that must be hit to the tick will be missed by two ticks and then respected. A zone — a band — is both more honest and more tradable.
The important consequence is about stop placement, and it is the most practical thing in this chapter. If your level is a zone from 1,480 to 1,492, your stop does not go at 1,479. It goes below the zone plus a volatility buffer — typically some fraction of ATR — because the noise around a level is exactly where stops get taken before the level holds. Placing a stop at the obvious price is placing it where everyone else did, which is precisely where liquidity is sourced.
Supply and demand zones, defined mechanically
The supply-and-demand framing improves on classical support and resistance in one specific way: it gives you a rule for which level to mark, rather than leaving it to the eye. The construction is:
- Find a strong impulsive move — a run of large-bodied candles in one direction, ideally defined by a threshold such as a multiple of ATR rather than by eye.
- Mark the base immediately before it — the consolidation, or the last opposite-direction candle. That is the zone.
- The logic: an imbalance existed there large enough to move price forcefully, and unfilled orders may remain.
- The classifications — rally-base-rally, drop-base-rally, rally-base-drop, drop-base-drop — describe what happened before and after the base. They are worth knowing and worth treating with care: a purely mechanical before-and-after comparison misclassifies zones whenever the path within the comparison window is not monotonic. If you build this, look at the path, not just the endpoints.
What the evidence supports, honestly
| Level type | Support in testing | Note |
|---|---|---|
| Prior day's high / low | Reasonable | Obvious, universally watched, mechanically defined. The coordination mechanism is real here |
| Round numbers | Reasonable | Documented clustering of orders at round numbers across many markets |
| 52-week high / low | Reasonable | Widely watched; also entangled with momentum, which may be doing the work |
| Opening range high / low | Moderate, intraday | Mechanically defined and heavily traded, so partly self-fulfilling |
| Prior swing extremes | Moderate | Depends entirely on the swing definition, which brings back the lookback problem |
| Supply/demand zones | Weak to moderate | Often survives a naive control and frequently fails a dip-matched one — because the entry is itself a pullback, and pullbacks mean-revert regardless |
| Fibonacci retracements | Not supported for the specific ratios | Price retraces; that 61.8% beats 60% or 65% has not survived testing |
| Trendline touches | Not testable as usually drawn | Subjective point selection makes the result unfalsifiable |
The dip-matched control deserves its own paragraph, because it is the one that catches serious people.
Zone entries are, by construction, entries after a pullback. If you compare a zone strategy against a control that buys at the close on random days, the zone strategy will look good — but so would any rule that buys after a fall, because short-horizon mean reversion exists independently of your zone.
The correct control buys after a similar-sized pullback at a random level. When you make that comparison, the zone's apparent edge frequently shrinks to nothing.
Match the control to the entry. If the entry is a dip, the control must be a dip. This principle generalises far beyond zones and it is the most valuable single methodological habit in this entire book.
Check yourself