Support, resistance and role reversal

The answer

Support is a price where enough buy orders exist to stop a fall. Resistance is a price where enough sell orders exist to stop a rise. Both are real when real orders sit there, and both are imaginary when you have simply drawn a line through a place the price visited once.

Why this costs you money

You mark a support level. The price comes down to it and you buy, because that is what you are supposed to do at support. The price goes straight through and you lose 8% before you accept it.

That is the small version of the mistake. Here is the expensive version.

Levels look reliable because you only remember the times they worked. Your eye finds the 2 bounces and skips the 3 pass-throughs. Nobody circles a level that did nothing, so the chart you remember is a chart where support always holds. Then you size a position as if the level were a fact.

There is a second, quieter cost that is specific to support. A great many people place their stop just below an obvious support level, because it feels like the logical place. This means a large cluster of sell orders sits just below the level. Those orders are not protection. They are fuel. When price reaches them they execute automatically, which pushes price lower, which triggers more of them. The most obvious level on the chart is therefore the most crowded place to put a stop, and crowded stops get taken.

You can lose money twice at the same level. Once by buying at it, and once by placing your stop where everybody else placed theirs.

How it works

There are 4 reasons a price level can genuinely affect trading, and 1 reason it often only appears to. Learn to tell them apart, because 3 of the 4 real ones can be checked and the fifth one cannot.

Reason 1: resting limit orders, which you can see

An exchange order book holds orders waiting to be filled. If 200,000 shares are waiting to buy at ₹1,000 and only 4,000 are waiting at ₹1,001, then ₹1,000 is a real floor until those 200,000 shares are consumed. This is not a theory. It is visible in the market depth window of your broker app right now.

This is the most mechanically true form of support that exists, and it is also the most short-lived. Orders can be cancelled in a millisecond. A wall of orders in the book tells you about this minute, not about next Tuesday.

Reason 2: option open interest, which has a hedging mechanism behind it

When many option contracts are open at one strike price, the people who sold those options must hedge. A market maker who has sold call options buys some of the underlying to stay neutral, and adjusts that hedge as the price moves. Near expiry, those adjustments push in the opposite direction to the price move: as price rises toward a heavily traded strike, the hedgers sell, and as it falls toward the strike, they buy.

The result is a mild pull toward strikes with large open interest, strongest in the final hours before expiry. Ni, Pearson and Poteshman documented this in US stocks in 2005 and it is one of the few chart-level effects with a clean, physical explanation. This matters enormously in India, where index option volumes are very large and expiries are weekly.

Be honest about the size, though. The effect is measured in fractions of a percent, it fades quickly as expiry moves further away, and it is a pull, not a wall.

Reason 3: round numbers, which are documented

Prices cluster at round numbers. Lawrence Harris showed this for US stocks in 1991 and the effect has been found in essentially every market examined since. People place limit orders at ₹500, not ₹497.35. Institutions set targets at $100, not $98.60. Option strikes are listed at round intervals, which concentrates activity further.

So a round number really is a place where more orders sit than at neighbouring prices. It is not mystical. It is a fact about how humans type numbers into boxes.

Reason 4: memory of a past price, which is behavioural

Investors are reluctant to sell at a loss. Odean measured this in 1998 in real brokerage accounts: people sell winners far more readily than losers. A consequence is that a price at which a large volume changed hands becomes a place where holders want to get out at breakeven. That produces genuine selling pressure when the price returns to it from below.

This is the mechanism behind role reversal: a level that acted as support, once broken, often acts as resistance later. The people trapped above it are waiting there. The mechanism is real, but note what it requires: large volume must actually have traded at that price. A level that price touched for 20 minutes 2 years ago traps nobody.

Reason 5: self-fulfilment, which is real but circular

Many people watch the same obvious level, so many people act at it, so something happens there. This is a genuine cause. It is also the weakest one, because it holds only for as long as the crowd keeps watching, and it gives you no way to know in advance whether the crowd's buying at the level will exceed the selling coming into it.

Lines against zones

A level is a zone, not a line. Prices are set by an auction and auctions are noisy. If you draw support at exactly ₹1,247.80 and the price turns at ₹1,239, did the level hold?

You cannot answer that unless you decided in advance. Fix a tolerance before you count anything. A workable starting point is 0.5% for a large index, 1% for a large stock, and 2% for a small one. Any tolerance is defensible. No tolerance at all is not, because without one you will judge every case after the fact and score yourself as right every time.

The 2 things that are decoration

A level marked from a single wick. One brief spike is one trade, not a crowd. If almost no volume traded there, nobody is waiting there.

A level that has never been tested. An untested level is a guess. It becomes evidence only when price returns to it and something measurable happens.

What it tells you, and what it does not

A support level tells you where a reaction is more likely than at a random price. It tells you nothing about the direction of the eventual break.

It does not tell you how strong the level is. Everyone can see how many times a level held. Nobody can see how many buy orders remain, and a level that held 4 times may have consumed almost all of the orders that were there.

It does not survive a change of information. A level formed by 2 years of accumulated opinion is worth nothing 30 seconds after a company reports a fraud or a central bank surprises the market. Prices are opinions, and levels are old opinions.

And critically: in a sustained downtrend, every support level breaks. That is what a downtrend is. Reading each broken level as an anomaly, rather than as the trend doing its normal work, is how people average down into a company that never recovers.

The decision rule

A level deserves your money only if you can name the mechanism.

If the level is a round number, or a heavily traded option strike, or a price where very large volume actually changed hands, then treat it as a zone worth watching, with a defined tolerance.

If you cannot say which of those 3 it is, you have drawn a line at a place price visited. Watch it if you like. Do not size a position on it.

Unless the market is in a confirmed downtrend on your timeframe, in which case treat every support level as a place where a fall pauses, not a place where it ends. Buying support inside a downtrend is the most common way retail accounts lose money slowly.

And a specific instruction on stops: do not place a stop just below an obvious support level. Either place it far enough below the zone that ordinary noise cannot reach it, or accept that you will be taken out by the cluster.

Try this now

This takes 5 minutes and it will change what you believe about your own levels. Do it on a chart you already trade.

1. Open a daily chart, 2 years, of a stock or index you actually follow.
2. Pick one horizontal level you would genuinely use. Take the most obvious one on the chart. Write the price down on paper.
3. Write down your tolerance before you look at anything else: 1% for a large stock, 0.5% for an index. Write it down. This step is the whole test.
4. Now go left to right through the 2 years and count 2 numbers. Held: price entered the tolerance zone and turned away without closing beyond it. Broke: price closed beyond the far side of the zone.
5. Write the 2 counts as a fraction. For example, 3 held out of 7 tests.
6. Now do the same on a round number near the current price — ₹500, ₹1,000, $50, $100, or a round index level ending in 00 — and compare the 2 fractions.

What you should see. Two things, and both surprise people.

First, the total number of tests is small. Most levels are tested 3 to 8 times in 2 years, which is not enough data to conclude anything. If you have been trading a level as though it were reliable, you have been trading a sample of 5.

Second, for most readers the round number performs at least as well as the carefully drawn level, and often better. That is not a coincidence. The round number has a mechanism behind it. The drawn level usually does not.

If your carefully drawn level clearly beat the round number, keep it, and write down which of the 4 mechanisms above explains it. If you cannot name one, you found a coincidence in a sample of 5.

Three real cases

1. The BSE SENSEX at 21,000, January 2008 onward (India) — role reversal on a national scale The SENSEX reached roughly 21,206 intraday on 10 January 2008 and then fell by more than half over the following year. The 21,000 area then acted as resistance for years. The index came back to it in late 2010, failed, and did not move decisively and permanently above it until 2013 and 2014. That level combined all 3 real mechanisms at once: it was a round number, it was the level of a generational high, and enormous volume had traded near it. Millions of investors were holding losses from that zone and waiting to exit at breakeven. the exact January 2008 intraday high and the exact date of the durable break above it.

2. The S&P 500, 13 October 2022 (United States) — the stops below the level were the fuel The June 2022 low had been the most watched level in the US market for months. On 13 October 2022, a hotter than expected inflation report sent the index below that low at the open. Every stop resting under the obvious level executed. Then, within the same session, the index reversed and closed sharply higher, one of the largest intraday turnarounds on record for the index. the exact intraday low and the exact percentage of the reversal. The lesson is mechanical and it generalises: price is drawn toward clusters of resting stop orders, because those orders are guaranteed to execute. The most obvious level on the chart is the worst place to hide a stop.

3. Yes Bank, August 2018 to March 2020 (India) — support inside a downtrend is a sequence of failures The stock fell from roughly ₹400 in August 2018 to about ₹5.5 in March 2020, when the Reserve Bank of India placed it under a moratorium and a reconstruction scheme followed. Throughout that fall, level after level was identified as support, and every one of them broke. Each break looked like an anomaly on its own and each was in fact the trend working normally. the exact high in 2018 and the exact low in March 2020. The general lesson is the one that costs Indian retail investors the most money: a support level does not tell you the business is sound, and the chart cannot see the loan book.

The question that resolves it

A novice looks at a level and asks: how many times has it held?

An expert asks: what is actually sitting there?

The first question is answered by counting bounces, which is counting the past. The second question has only 4 possible answers — resting limit orders, option open interest, a round number, or trapped holders from a high-volume price. If none of the 4 applies, there is nothing at the level except your line.

What would make this wrong

If support and resistance carried no information, then price would turn at previously visited levels no more often than at randomly chosen prices, and round numbers would show no clustering. Round-number clustering is well-documented, so that part of the claim is safe. The general claim about drawn levels is much weaker, and the studies that test it produce mixed results depending heavily on how a "touch" is defined. Anybody who tells you support and resistance is proven is overstating; anybody who tells you it is worthless is ignoring order books and option hedging.

Three honest limits.

First, everything in this article is about probability, never about certainty. A level that has held 5 times may break on the sixth test, and there is no information anywhere on the chart that would have told you which test was the last one.

Second, the tolerance you choose changes the answer. Widen the zone enough and every level holds. Narrow it enough and none do. This is why the tolerance must be fixed before counting, and why comparisons between 2 people's level records are usually meaningless.

Third, the option-hedging mechanism is real but small and short-ranged. It is not a reason to expect an index to stop at a round strike 3 weeks before expiry.

Carry this

• Name the mechanism, or do not use the level.
• Fix your tolerance before you count, or you will score yourself right every time.
• In a downtrend, every support level breaks. That is what a downtrend is.

Knowledge check

Related

• Trends, trendlines and market structure
• Fibonacci retracements and extensions
• Reversal chart patterns
• Continuation patterns and gaps
• ← Trend, support and market structure