Point and Figure — the oldest chart, and its false targets
The answer
A Point and Figure chart plots columns of X for rising prices and columns of O for falling prices, and it adds a mark only when price moves by an amount you chose. It is a noise filter with 2 numbers you set yourself, which is more honest than most chart types. Its famous price objectives, counted off the width or height of a column pattern, have no mechanism behind them at all.
Why this costs you money
A trader draws a Point and Figure chart, finds a wide congestion area, counts across it and gets a price objective 22% above the current price. He buys. He decides to hold for the objective, because the objective is "what the chart says".
The stock rises 14%, stalls, and gives it all back over the next 4 months. He never took the profit, because the target had not been reached.
Here is what that target actually was. He counted 11 columns across the congestion area, multiplied by the reversal amount of 3, multiplied by the box size of 2 rupees, and added the result to the bottom of the pattern. The number was 11 × 3 × 2 = 66 rupees.
Nothing in that arithmetic contains anything about the company, about how many shares are available to buy, about who was selling, or about the price at which sellers would return. It is the width of a shape multiplied by 2 settings he chose himself.
The test is simple and it is devastating. Change the box size from 2 to 4 and redraw. The congestion area now has about half as many columns, but each column is worth twice as much, so the target lands in roughly the same place. That sounds reassuring. Now change the reversal from 3 to 5 and redraw. The target moves a long way. Two traders, both following the standard method correctly, get different objectives on the same stock, and neither can say why one is right.
The cost is not the wrong target. The cost is holding a position for a number that was never a forecast.
How it works
The grid
You choose 2 numbers before you draw anything.
The box size, B. The smallest price move the chart will record. It can be a fixed amount, or a percentage of price, or set from a traditional scale.
The reversal amount, R. How many boxes price must move against the current column before a new column starts. Almost everybody uses 3, which is why the method is often called the 3-box reversal method.
The chart is a grid. Each square is B tall. Only 1 column is active at a time.
Drawing it
In an X column, which means the price is rising, you check the day's high first. Every time the high reaches another full box above the top X, you add an X.
If the high did not add a box, you then check the low. If the low is at least R × B below the top X, the column is finished. Move 1 column right, start 1 box below the top X, and fill in O boxes down to the low.
In an O column you do the same in reverse: check the low first, then the high.
That is the whole construction. Notice the 2 things it removes.
Time is not on the axis. A column can take 2 days or 5 months. Some platforms print a small number or letter in the column to mark the start of a month, which is a partial repair.
Small moves are not recorded at all. A move of less than 1 box in the direction of the column adds nothing. A move of less than R × B against the column adds nothing. On a 3-box reversal chart with a 1% box, the price must reverse by about 3% before the chart notices.
The 2 versions, and why charts disagree
The original method, sometimes called the book method, used closing prices only. A single price per day.
The modern method, which every platform now uses by default, uses the day's high and low. It is more sensitive, so it produces more columns.
If your chart looks different from somebody else's chart of the same stock, check this setting before you argue about anything else.
Box scaling
The traditional Chartcraft scale set the box size in dollars according to the price of the stock: small boxes for cheap stocks, larger boxes for expensive ones.
Percentage scaling is better and you should use it. A fixed box of 1 rupee is a 2% filter on a 50 rupee stock and a 0.05% filter on a 2,000 rupee stock. Those are not the same chart type. They are 2 different instruments.
The signals
Point and Figure has a small, fixed catalogue of signals, and their precision is its real advantage. There is no argument about whether a signal occurred.
| Signal | Definition |
|---|---|
| Double top buy | An X column rises 1 box above the previous X column |
| Double bottom sell | An O column falls 1 box below the previous O column |
| Triple top buy | An X column exceeds the 2 previous X columns |
| Bullish support line | A 45-degree line drawn up from the lowest O column |
Every other named Point and Figure pattern is a longer version of these.
The counts, which is where it goes wrong
The vertical count. Take the length of the first column after a reversal, multiply by R, multiply by B, and add it to the bottom of that column.
The horizontal count. Take the width of a congestion area in columns, multiply by R, multiply by B, and add it to the bottom of the pattern.
Read those 2 rules and ask what makes them true. There is no mechanism. There is no reason a shape 11 columns wide should produce a move of 11 × R × B rather than 11 × 2 × B or 11 × 7 × B. The 3 in the formula is the reversal amount, which is a display setting, and it has been promoted into a forecasting constant.
A congestion area does contain real information: it is a price region where the market changed direction repeatedly, so it is a region where buyers and sellers have previously been closely matched. That is worth knowing. Turning its width into a number of rupees is a different claim entirely, and nothing supports it.
What it tells you, and what it does not
Point and Figure tells you where price has repeatedly stalled and turned. It is better at this than a candlestick chart, because time is removed. On a candlestick chart, 4 months of sideways movement is 4 months wide and easy to skip over. On a Point and Figure chart it is a dense block of alternating columns at the same price levels, and it is the most visible feature on the page.
It tells you whether a signal occurred, without argument. A double top buy either happened or it did not. Compare that with "is this a head and shoulders", which 3 people will answer 3 ways.
It does not tell you when, or how long, or anything about volume. Traditional Point and Figure has no volume at all.
It does not tell you the extremes. A high that did not complete a box is not on the chart.
And its counts do not tell you a target. They tell you the size of a shape multiplied by your own settings.
The decision rule
Use Point and Figure for 2 things and refuse the third.
Use it to locate levels. The prices where the chart has many columns touching the same row are prices where the market has repeatedly changed its mind. That is a real observation about a real price series.
Use it to force a written definition. Setting B and R in advance is a commitment, and it stops you deciding after the fact that a move was meaningful.
Refuse the count as a target. If you want a target, take it from something with a mechanism behind it: a valuation, a level where sellers previously appeared in size, or a fixed multiple of the distance to your stop.
Try this now
This takes about 5 minutes and it proves that the box size is a choice, not a property of the stock.
- Open a Point and Figure chart of a stock you own or follow. TradingView, StockCharts and several Indian platforms all offer it. Set the reversal to 3 and the box size to a percentage of 1%. Set the range to about 2 years.
- Write down 3 things: how many columns are on the chart, whether the current column is X or O, and the price level of the most recent signal (a double top buy or a double bottom sell).
- Change the box size to 0.5%. Write down the same 3 things.
- Change the box size to 2%. Write down the same 3 things again.
- Now find the widest congestion area on the 1% chart — a run of columns that stay inside the same price region. Count its width in columns. Multiply by 3, multiply by the box size in rupees or dollars, and add it to the bottom of the pattern. Write down that number. That is a horizontal count objective.
- Repeat the count at the 2% setting and write down that objective too.
What you should see. Three findings, and each one settles an argument.
The column count changes by a large factor. Halving the box size usually more than doubles the number of columns, because more small reversals now qualify.
At least one setting disagrees with the others about the current direction. On most stocks, at least one of your 3 charts will be in an X column while another is in an O column. Same stock, same 2 years, opposite conclusions, and all 3 charts drawn correctly.
Your 2 objectives are far apart. It is common for the difference between them to be larger than the stock's entire range over the 2 years. If a target can move that far because you changed a display setting, it is not a target.
Keep the 3 numbers. The next time somebody shows you a Point and Figure objective, your first question is now automatic: what box size, and what reversal?
Three real cases
1. Victor de Villiers, 1933 — the formalisation, and what came before it Point and Figure is the oldest chart method still in common use in the West. It predates candlestick charts in English by about 60 years. Traders on the New York Stock Exchange were keeping price records by hand in the 1880s and 1890s, in what was called the book method, in the era when Charles Dow was writing. The method was designed for a person with a pencil and a ledger and no way to store a chart. That is why it is so compact: a month of trading can fit in 3 columns.
Victor de Villiers published The Point and Figure Method of Anticipating Stock Price Movements in 1933, which appears to be the first full written system, including the counts.
Notice the date and what surrounded it. This is a method built to survive without computers, in a market with no continuous data feed, no volume history and no adjusted price series. Almost every design decision in it is a response to a constraint that no longer exists.
2. A. W. Cohen, Chartcraft, and the Bullish Percent Index — the one derivative that is genuinely useful A. W. Cohen at Chartcraft standardised the 3-box reversal method in a book published in 1947. Chartcraft later became part of Investors Intelligence, which still publishes Point and Figure services. Thomas Dorsey's Point and Figure Charting, published by John Wiley & Sons in 1995, is the standard modern reference.
Cohen also created the Bullish Percent Index, which is the percentage of stocks in an index that are currently on a Point and Figure buy signal.
This one deserves attention, because it is the exception to everything else in this article. It is not a forecast from a shape. It is a breadth measure: a count of how many members of an index are in an uptrend by a fixed, mechanical definition. Its value comes from the fact that the definition is unambiguous and applied identically to every stock. You could compute the same thing with a moving average rule and get something similar.
That is the honest summary of Point and Figure's real contribution. The precise, countable signal definition is useful. The forecasting layer built on top of it is not.
3. Robert Earl Davis, 1965 — the study everybody cites and nobody reads Robert Earl Davis published a study of Point and Figure patterns in 1965, reporting very high success rates for certain formations across a large sample of US stocks.
This study is cited constantly in Point and Figure literature and almost never examined. Apply the 3 questions from earlier in this cluster to it.
Was the rule written down before the data was looked at? Unclear from the usual citations.
How many patterns were tried? Unknown.
Does it survive costs and a later period? No published replication on modern data that I am aware of.
I am not saying the study was dishonest. I am saying that a result from 1965, with no replication, no cost adjustment and no correction for the number of patterns tested, is the weakest kind of evidence, and that it is the main evidence offered for a method that is still sold today.
The question that resolves it
A novice asks: what is the price target?
An expert asks: what would have to be true about buyers and sellers for that target to be reached?
The second question is answerable for some targets and not for others. "Sellers appeared in size at 840 twice in the last year, so 840 is where I expect supply" is a statement about behaviour that can be checked. "The congestion was 11 columns wide, so the move is 66 rupees" is arithmetic on a picture. The first can be wrong in an informative way. The second cannot be wrong at all, because it never claimed anything about the world.
What would make this wrong
The claim here is that Point and Figure is a useful noise filter with a clear threshold, and that its counts are not forecasts.
The filter half would be wrong if a Point and Figure signal, defined precisely and tested with the box size fixed in advance, produced no better results than the same threshold applied without the chart. That is easy to test, because a Point and Figure double top buy is very close to a percentage breakout rule, and percentage breakout rules have been tested extensively.
The count half would be wrong if somebody took a large sample of congestion areas, computed the horizontal count objective for each, and showed that price reached the objective more often than a control target set at the same distance by a rule with no pattern in it. That is a fair test. It has an obvious control group and it requires no new data. As far as I am aware it has not been published.
Two honest limits on the criticism.
The counts might work as a rough scale rather than a price. A wide congestion area does take longer to form and does involve more accumulated activity than a narrow one. It is not unreasonable that bigger bases precede bigger moves. What is unreasonable is converting that into a specific number and holding a position for it.
Institutions do use Point and Figure, and not out of nostalgia. The reason usually given is objectivity: a signal definition that cannot be argued about is valuable when several people must agree on a process. That is a real benefit and it has nothing to do with prediction.
In India
Percentage scaling is not optional here. Indian share prices span an enormous absolute range, from stocks priced under 20 rupees to stocks priced above 100,000 rupees. A traditional box scale set in absolute amounts is meaningless across that range. Set the box as a percentage and the chart becomes comparable between stocks.
Price bands change the shape of the chart. Indian stocks carry daily price bands, commonly 2%, 5%, 10% or 20%, with a wider dynamic band for stocks in the derivatives segment. A band caps the day's high and low, which are the 2 inputs the modern method uses. So the number of boxes a session can contribute is capped by regulation. A stock under heavy selling pressure produces the same single O box on a band-locked day as a stock that drifted down mildly.
That has a direct consequence for the counts. A congestion area that formed during a period of repeated band locks is wider than the same amount of disagreement would have made it in an uncapped market, because the price could not travel and had to spend more sessions in the region. A wider congestion area produces a larger horizontal count objective. The band therefore inflates the target. This is the clearest available demonstration that the count is a property of the mechanism and not of the market.
Corporate actions need adjusted data. Bonus issues, splits and large special dividends are more frequent in India than in the United States. On an unadjusted series a 1-for-1 bonus produces an instant fall of about 50%, which on a 1% box chart is around 50 O boxes in a single column. Check that your platform uses an adjusted series.
Availability varies. Point and Figure is available on TradingView with NSE and BSE data and on some Indian broker platforms, but it is far less commonly offered than candlesticks, and the default box scaling on those platforms is not always documented.
In the United States
The traditional box scale was built for a market quoted in fractions. Until 2001, US stocks traded in increments of eighths and then sixteenths of a dollar . A sixteenth is 6.25 cents. The traditional scale, with its 0.25 and 0.50 boxes, was designed around those increments. After decimalisation the smallest quoted increment became 1 cent, and the traditional boxes stopped corresponding to anything in the market's structure. They are now simply inherited numbers.
Extended hours change the high and the low. The modern method uses the day's high and low, and on most US platforms the default chart uses regular trading hours only, from 09:30 to 16:00 Eastern time. A stock that traded much higher in the pre-market session after an earnings release contributes none of that range to the chart. Switching extended hours on can add or remove whole columns.
Point and Figure survives in institutional use. Investors Intelligence still publishes Point and Figure research, and the Bullish Percent Index remains a widely quoted breadth measure for US indices. Dorsey Wright, founded on the method, was acquired by Nasdaq in 2015. That is a genuine fact about institutional adoption, and it is worth separating from the question of whether the counts predict anything.
Where they differ, and what that tells you
The construction rule is identical in both markets. The inputs are not.
In India, the day's high and low are capped by the price band. In the United States they are not. Since the modern method reads the high and the low, the band is a direct limit on how much a session can contribute to the chart.
What that tells you is that box counts are not comparable across the 2 markets. A rule such as "buy on a triple top, and set the target from the horizontal count" behaves differently in India, because the width of Indian congestion areas is partly produced by regulation. The chart looks the same. The number it produces means something different.
There is a second difference and it runs the other way. In the United States, the extended session hides range from the chart. In India, there is no extended session, so the regular-hours high and low are the complete high and low. An Indian Point and Figure chart of a cash equity is complete in a way a US one often is not.
Put the 2 together and you get the useful rule. In India the chart is complete but truncated. In the United States the chart is untruncated but often incomplete. Neither is a better chart. But if you take a rule tested on one and run it on the other, the thing you are testing has quietly changed, and no warning appears on the screen.
Carry this
- Box size and reversal amount are your choices, not the stock's properties.
- The signal definitions are precise, which is Point and Figure's real strength.
- The counts are width multiplied by your own settings. They are not forecasts.
- Change the box size. If the current column flips from X to O, you have learned what the chart was actually telling you.