Point and Figure Charts: How the X's and O's Filter Out Noise
Point and figure charts plot price as columns of X's and O's with no time axis at all. How box size and the 3-box reversal decide what the chart shows, how to read column length and breakouts, the vertical and horizontal count price targets, and what the filter quietly deletes.
Most charts answer the question "what happened at 10:15." A point and figure chart refuses to. It has no time axis at all. Rising price stacks a column of X's, falling price stacks a column of O's, and a symbol is added only when price covers a set increment. A stock that spends four hours drifting inside that increment produces nothing. No bars, no gaps, no marks of any kind. The chart simply does not advance. That one rule is the whole method, and whether it is exactly what you wanted or a serious problem depends entirely on what you were using the chart for.
Quick Answer
A point and figure chart plots price as vertical columns of X's and O's and removes the time axis completely. X's mark rising price, O's mark falling price, and no column ever mixes the two. A symbol is added only when price covers one full box size in the direction the column is already running. A brand new column starts only when price moves against the current one by the box size multiplied by the reversal amount, conventionally three boxes. Quiet stretches leave no mark at all.
What Is a Point and Figure Chart, and How Is It Different From a Candlestick?
Take a stock at $46 and set the box size to $1. Price rises to $47 and an X appears. Rises to $48 and a second X stacks on top of it. Rises to $51 and the column is now six X's tall. That column is the entire record of the move. It does not say how long the move took, whether it happened in one violent minute or across three sessions, or what the volume was. It says price went up six dollars.
Then price turns. The column of X's does not become a column of X's and O's, because a column never mixes symbols. Instead the chart waits. Only once price has fallen far enough to satisfy the reversal rule does it step one column to the right and start stacking O's downward. Until that threshold is met, a pullback of any size is simply not drawn.
That is the part worth sitting with, because it is the difference that matters. On a 5-minute candlestick chart, forty minutes of directionless chop renders as eight indecisive bars you have to look at, interpret and dismiss. On a point and figure chart of the same forty minutes, if price never covered a box, it renders as nothing. The StockCharts ChartSchool introduction to point and figure charts frames this as filtering out insignificant price movements so that support, resistance and breakouts stand out. Accurate, and worth restating one degree less kindly: information is being deleted on purpose, and you chose the threshold. Almost every convention in the general technical analysis toolkit assumes an x-axis made of time, and on this chart type most of them stop applying cleanly.
One small mercy: because traders do occasionally need to know roughly when something happened, many platforms mark the passage of time by swapping the first symbol of a new month for a digit or a letter. That is an annotation bolted onto a chart which structurally has no calendar, and it tells you nothing about pace within a column.
A short note on where the method came from
Brief, because the history does not change how you trade it. The technique predates almost everything else on this site. A writer publishing as "Hoyle" documented the method in the 1898 book The Game in Wall Street, Victor De Villiers published the first dedicated manual on it in 1933, and A.W. Cohen developed and popularised the modern three-box reversal convention through Chartcraft Inc. in the 1940s. Wikipedia's entry on the point and figure chart carries that lineage. The reason the format looks the way it does is that it was designed to be maintained by hand, in pencil, on graph paper, from a printed list of prices. Columns of letters are cheap to update when the alternative is redrawing a bar chart every day.
How Do You Set the Box Size and the Reversal Amount?
Point and figure has exactly two settings and they do all of the work between them.
Box size is the minimum price increment required to add one more symbol to the current column. Set it to $1 and the chart ignores every move smaller than a dollar. Set it to 10 cents and it records ten times as much detail, including ten times as much of the wobble you were trying to get away from.
The reversal amountis how many boxes price must travel against the current column before a new, opposite-symbol column is allowed to start. Three is the modern convention. With a $1 box and a column of X's topping out at $51, a reversal needs price down at $48, and the new column then prints three O's at $50, $49 and $48. A fall to $49 does nothing. A fall to $48.50 does nothing. The chart is unchanged until the full three boxes are covered, which is precisely why a point and figure chart does not flinch at ordinary pullbacks.
The older one-box reversal method, where any single box against the column starts a new one, still exists and is still occasionally used. It is far more sensitive and far noisier, and the three-box convention exists because the sensitivity was not worth what it cost.
| Setting | How it is set | What it does well | What it costs you |
|---|---|---|---|
| Fixed absolute box | You pick a currency amount per box, say 50 cents on a $40 stock | Predictable. One box always means the same distance, so you can size risk against it | Goes stale. A box tuned in a calm week fills the chart the day the range doubles |
| Percentage of price | The box is a share of current price, so it scales as the stock moves | Travels between a $9 stock and a $900 one without retuning | Ignores volatility. Two stocks at the same price with different personalities need different boxes |
| ATR-derived box | Box size is computed from Average True Range over a lookback | Adapts to the current range on its own, and the usual default where platforms offer one | The box changes underneath you, so a column of six symbols no longer means a fixed distance |
| Traditional price-banded scale | Inherited from hand charting: the box widens in steps as the price level rises | Matches how most published P&F charts and the older literature were drawn | The band boundaries are arbitrary, and a stock sitting near one gets a different chart either side of it |
| Smaller box, any method | Tighten the increment so less movement prints a symbol | More detail, earlier breakouts, and more setups over a session | More reversals survive the filter. You put back the noise you switched to P&F to remove |
| Larger box, any method | Widen the increment so more movement is needed per symbol | Visibly cleaner structure and far fewer meaningless column flips | Lag. The breakout confirms after a bigger chunk of the move has already happened |
| 1-box reversal | The older method: any single box against the current column starts a new one | Catches turns almost immediately, and records more of the path | The column count explodes and most columns mean nothing. This is the setting the 3-box convention exists to replace |
| 4-box or 5-box reversal | Require more boxes against the column before it breaks | Only substantial turns earn a new column, so the trend read gets very stable | Paired with a wide box, a reversal can need a move most day traders would have been stopped out of long before |
Notice that every row trades the same two things against each other. Detail and lag are one property viewed from two sides, and there is no combination of box and reversal that is both responsive and quiet. The honest consequence is that two traders looking at the same stock on the same day, one running a 25-cent box and the other a $1 box, are looking at genuinely different charts and can reasonably disagree about whether a breakout happened. That subjectivity is structural, not a flaw in anyone's setup. If your platform does not offer the chart type at all, or offers it with the box locked to a preset, the survey of free charting software is a reasonable place to start looking for one that exposes both parameters.
How Do You Read the Columns?
There are only three things on the chart, so the reading is mostly about column length and column position.
Column length is the persistence read.A long unbroken run of X's means price travelled a long way without ever handing back three boxes. That is a stronger statement than it looks, because it rules out a whole class of behaviour: the move did not stall, it did not round over, and nobody managed to push it back a meaningful distance while it was running. Short alternating columns say the opposite, and they say it clearly. This is the closest thing point and figure has to a momentum reading, and it is worth being careful with. It is a statement about distance covered without a qualifying pullback, not a measurement of speed, and the sources that describe it as momentum are using the word loosely.
Column position against the previous same-direction column is the signal.The classic buy trigger is a column of X's rising one box above the high of the previous column of X's. The classic sell trigger is the mirror image: a column of O's falling one box below the low of the previous column of O's. That is it. No indicator, no close, no confirmation candle. Either the column exceeded the prior one or it did not, which is the same binary clarity that makes marking levels that actually hold much easier on this chart type than on a candlestick chart full of wicks poking through.
One box = $1. Reversal = 3 boxes. Nothing else decides what appears.
The third convention worth knowing is the 45-degree trendline. Because the grid is uniform, point and figure traders draw support and resistance lines at a fixed 45 degrees from the corner of a box rather than fitting a line by eye to a set of swing points. It removes the argument about where the line goes, which is most of the argument. It also removes the flexibility, so a trend running at any other angle is not describable by the convention. If you are used to fitting trendlines to actual swing highs and lows, this will feel arbitrary at first, and the fairest defence of it is that the arbitrary version is at least consistent between two traders.
What Patterns Show Up on a Point and Figure Chart?
Point and figure has its own named pattern vocabulary, and the first thing to know about it is that one of the names is a trap.
A double topon a point and figure chart is bullish. It describes a column of X's that pushes one box above the high of the previous X column, and it is the standard buy signal. A double bottomis bearish, describing a column of O's breaking one box below the previous O column low. If you learned the candlestick versions of double tops and double bottoms, where a double top is a reversal signal and bearish, the terminology here points the opposite way. Same words, different chart type, inverted meaning. Read the chart type before you read the pattern name.
The catapult is the other name that comes up constantly. The idea is a two-stage confirmation: an initial breakout, then a pullback that fails to undo it, then a second breakout in the same direction. A bullish catapult resolves upward, a bearish one downward. Definitions differ between sources on exactly what the first stage has to be, with some requiring a triple-top breakout before the pullback and others accepting a double top, so check which version your reference is using before you go hunting for one. The trading logic underneath is the same break-and-retest structure you would recognise from a candlestick chart, just expressed in columns.
The column broke out above the last one. That still leaves the question of whether the underlying setup is worth risking money on.
Upload the screenshot and SnapPChart reads that one image against a fixed rubric, then returns a setup grade, an entry, a stop with the reasoning behind the level, targets, and the reward-to-risk those levels imply. The grade is the part worth disagreeing with before the order goes in.
Grade this chartHow Do You Set a Price Target From a Point and Figure Chart?
This is the one genuinely distinctive thing point and figure does that other chart types do not, and it is the reason the format survived. The uniform grid means you can count boxes, and counting boxes gives you an arithmetic price target rather than an eyeballed one. There are two methods.
The vertical count
Measure the length in boxes of the breakout column, multiply by the box size, then multiply again by the reversal amount. A breakout column ten boxes tall on a $1 box with a three-box reversal gives 10 x 1 x 3, or $30 of projected move. Project that from the base of the column and a move starting at $40 has a $70 target.
Two caveats before you write that number down. TradingSim's walkthrough of point and figure price targets works the method through with a real chart, and sources differ on the anchor point, with some projecting from the base of the breakout column and others from the breakout level itself. Check which convention your platform uses, because the difference is not small. The second caveat is the multiplication itself: multiplying by the reversal amount triples an already large number, which is why vertical count targets land a long way out and frequently go unmet. The method works best after a sharp, momentum-driven move, which is exactly the situation where the breakout column is long enough to be worth counting.
The horizontal count
Same arithmetic, different input. Instead of measuring the height of the breakout column, you measure the width in columns of the sideways consolidation that preceded it, then multiply by box size and reversal amount and project from the breakout. A base eight columns wide, again on a $1 box with a three-box reversal, gives 8 x 1 x 3, or $24. The logic is the familiar one that a longer base implies a larger move, expressed as a number.
The horizontal count is the less reliable of the two, and it is worth saying that plainly rather than presenting both as equally solid. The weakness is that you have to decide which row of the congestion to count across, and the base is usually not a neat rectangle, so two traders counting the same consolidation will pick different rows and get different targets. TradingSim flags exactly this subjectivity. Use it after a long, well-defined base where the row choice is obvious, treat the number as a rough scale rather than a level, and lean on the vertical count when both are available.
Point and Figure vs. Renko vs. Candlestick
Point and figure and Renko get lumped together as "the chart types without a time axis," and they do share that headline property, but the mechanics diverge in ways that change what each one is useful for.
Renko drops the clock but keeps a fixed price-per-brick grid, and it keeps marching. Every brick takes the next position to the right, so a Renko chart is one continuous staircase flowing left to right and the horizontal axis, while meaningless as time, is at least always advancing. Point and figure drops the clock and keeps no per-symbol horizontal slot at all. A column grows vertically in place, and the chart only ever widens when a reversal fires. Ten boxes of uninterrupted upside occupy exactly one column of width. Point and figure also carries the vertical and horizontal counts, a native price-target technique Renko has no direct equivalent for.
The third member of this family works differently again. Heikin Ashi keeps the time axis entirely and smooths each candle by averaging it against the previous one, so you still get one candle per period and the noise reduction is arithmetic rather than structural. Three answers to the same complaint, three different things given up.
| Field | Candlestick | Renko | Point and figure |
|---|---|---|---|
| What creates a new mark | The clock. One candle per period, whether price moved or not | Price. One brick per fixed move in either direction | Price, in the current direction only. One symbol per box travelled the way the column is already going |
| Time axis | Evenly spaced and meaningful | Absent, but bricks still flow left to right in one continuous run | Absent, and there is no per-symbol slot either. The chart only widens when a reversal fires |
| Layout | One bar per period, side by side | A single staircase, each brick offset one step from the last | Vertical columns. A column grows in place, upward or downward, until the reversal rule breaks it |
| What starts a new column | Not applicable. Every period is a new bar | Not applicable. Every brick takes the next position | A move against the column of box size times the reversal amount, conventionally three boxes |
| Volume | Attached to each bar and comparable bar to bar | Dropped | Dropped |
| Wicks and the intrabar path | Visible. You can see what got rejected | Gone. Brick bodies only | Gone. A symbol says the box was covered, not how or in what order |
| Cost of a reversal | A reversal candle can print on any move, however small | Commonly framed as roughly double the brick size | Three boxes by convention, so at the same unit size a larger required move than Renko |
| Native price targets | Measured moves from pattern height, computed by hand | No standard built-in method | Vertical and horizontal counts, a technique specific to this chart type |
| Best environment | Anything. It is the general-purpose view | Sustained directional moves | Sustained moves and level work. Both degrade in a tight range |
| What you give up | Nothing, but you read the noise yourself | Time, volume and the shape of the path | Time, volume, path, and any view at all of a move smaller than one box |
The row that decides most of this is the reversal cost row. Renko asks for roughly double the brick to turn. Point and figure asks for three boxes. At the same unit size that makes point and figure the more stubborn of the two, which is an advantage when you are trying to hold a trend and a real cost when you are trying to get out of one.
What the Filter Actually Costs You
The benefits are genuine and easy to state. Support and resistance are obvious because the grid is uniform and there are no wicks to argue about. Trendlines are unambiguous. Breakouts are binary. Weeks of sideways drift collapse into a couple of columns instead of a hundred bars you have to scroll past. For anyone whose problem is that they see too many setups, the chart is an effective piece of self-discipline.
Now the bill. Time is gone, and with it every read that depends on it: session structure, the open, the lunchtime lull, how long a base has been building in actual days, whether a breakout happened in ninety seconds or over a fortnight. Volume is gone entirely, which is a harder loss than it sounds, because if your process relies on relative volume to separate a real breakout from a fake one, point and figure takes that input away and hands back nothing in its place. Signals arrive late by construction. A symbol exists only after the full box has been covered, and a new column exists only after three boxes have been given back, so there is no such thing as a forming signal to read intent from the way you can watch a live candle develop.
And the quiet one: the chart is a function of two numbers you chose. Box size and reversal amount are not properties of the market, they are settings, and changing either rebuilds the entire chart. A breakout that is clearly visible at a 50-cent box may not exist at all at a $1 box. This is why the chart type pairs badly with an over-stuffed screen and reasonably well with a small number of inputs that answer different questions. Point and figure is bare price action with no built-in indicators, and if you want a trend-strength or momentum read alongside it you are computing it from the underlying data and reading it elsewhere, since the point and figure chart itself will not display it. The argument for picking a short list of indicators that genuinely disagree with each other matters more here than usual, because the chart has already removed two of the things you would normally cross-check against.
Where an AI Chart Grader Fits, and Where It Does Not
Straight version, since this site sells a tool. SnapPChart does not construct point and figure charts. There is no box-size setting, no reversal-amount setting, and nothing anywhere in the product that tracks X or O column state, because it never touches a raw price series in the first place. It reads the pixels of a chart screenshot you upload. Nothing here should be taken as a claim that the analysis understands point and figure construction or treats a P&F screenshot as a distinct chart type. It does not.
The narrower and more useful claim is this. If you have already switched your platform to a point and figure chart before taking the screenshot, the read works on the trend structure, the levels and the price geometry visible in that image the same way it works on any other chart, because that is what it was built to do. What it cannot do is tell you what the boxes removed, and on this chart type that is a long list: the pullback that stopped one box short of a reversal, the volume that was or was not behind the breakout, the four hours of nothing sitting invisibly between two symbols in the same column. A grade computed from a pre-filtered chart is a grade of the filtered chart.
If you want a second opinion on a point and figure setup, my suggestion is to take the screenshot from the standard candlestick view of the same symbol and period, run the read on that, and keep the point and figure chart for what it is genuinely good at, which is refusing to react to pullbacks you should have ignored anyway. The mechanics of how a single-image read works are covered in the guide to AI trading tools, and a neutral description of what one chart read includes sits on the AI chart analysis page.
X's go up, O's go down, no column mixes them, and nothing prints until price covers a full box. A new column needs box size times the reversal amount against the current run, conventionally three boxes, which is why the chart ignores ordinary pullbacks. Column length is a persistence read and a break past the prior same-direction column is the trigger. Count targets vertically where you can and horizontally only off a clean base. Time and volume are gone, and the two numbers you set are doing more work than the market is.
Frequently Asked Questions
Is there a free point and figure chart tool, or do you need paid software?
Free options exist, but the chart type is far less universally supported than candlesticks, and that is the practical problem rather than price. Some free web charting platforms include it as a chart type alongside line, bar and candle. Others do not offer it at all, and a few offer it with the box size locked to a preset you cannot change, which is close to useless because box size is the setting that determines what the chart says. Before you commit to a platform, check three things: whether you can set the box size yourself, whether you can set the reversal amount yourself, and what underlying data the symbols are computed from. A tool that gives you the chart but hides both parameters is showing you somebody else's filter, not yours.
Do point and figure charts work on intraday timeframes?
They can, but the format did not grow up there and it shows. Point and figure spent most of its history as an end-of-day tool, drawn by hand from a price list, which is exactly the use case it is shaped for. Pushing it intraday means shrinking the box until symbols print often enough to be useful inside a single session, and a small box reintroduces the noise the chart exists to remove. It also makes the underlying data resolution matter a great deal: a box computed from 5-minute closes will miss a move that touched the level and came back, while the same box computed from tick data will catch it. If you want the noise filter on an intraday chart, the honest answer is that it works, that you will be tuning the box constantly, and that you should confirm what data your platform is feeding it before you trust a breakout.
What box size should a day trader start with?
There is no universal number, because the right box depends on how much of the day's range you are trying to make visible. A percentage of price is the usual first guess, and an ATR-derived box is the usual automated answer, but both are starting points rather than settings. The test that actually works is retrospective: pick a candidate box, apply it to a session you already lived through, and compare the columns against what you remember happening. If a day you remember as one clean trend produced six columns, the box is too small and you are charting noise. If a day with two genuine reversals produced a single unbroken column, you have filtered the session into a straight line. Tune until the column count matches your memory of the tape, then leave it alone.
Do point and figure price targets actually get hit?
Often not, and the arithmetic explains why. A vertical count multiplies the column length by the box size and again by the reversal amount, so a three-box reversal setting triples whatever the column measured. That produces targets a long way from the current price, which is fine as a rough sense of how much room a move might have and poor as a level to place an order at. Treat a count as a scale estimate rather than a prediction. The useful question it answers is whether the potential move is large enough to be worth the risk you would take to get it, and that question is answerable even when the target itself never prints.
Does a point and figure chart repaint or change after it prints?
An individual symbol does not move once it exists, and neither does a completed column. Two things can still change under you. The first is the live column, which keeps extending as long as price keeps covering boxes in the same direction, so the column you are looking at during the session is not finished. The second is bigger: change the box size or the reversal amount and the entire chart is rebuilt from scratch, because every symbol on it was a function of those two numbers. The same stock over the same period genuinely looks like a different instrument at a different box size, which is the strongest argument for picking your parameters once, writing them down, and not adjusting them because you do not like what the current chart is telling you.
This article is for educational and informational purposes only and is not investment, financial or trading advice. The point and figure construction rules described here, the box-size and reversal-amount conventions, the three-box reversal standard and the older one-box method, the vertical and horizontal count arithmetic, the 45-degree trendline convention, the double top, double bottom and catapult pattern descriptions, and the attribution of the method to the 1898 book published under the name Hoyle, to Victor De Villiers' 1933 manual and to A.W. Cohen's work at Chartcraft are the conventional published descriptions reproduced by charting platforms and reference sources; implementations and definitions differ between platforms and between authors, so confirm the behaviour of your own tool before trading it. The prices, box sizes and count figures used as examples are illustrative numbers, not a trade record. Nothing here is backtested performance and no method described is claimed to be profitable. Day trading and active trading carry a substantial risk of loss and are not suitable for every investor. SnapPChart grades a static chart screenshot you upload and returns a setup grade, entry, stop, targets and reasoning for that single image; it does not construct or compute point and figure charts, does not set box sizes or reversal amounts, does not scan the market, and does not track your account, positions or P&L. It can only account for what is visibly drawn on the image you upload. Do your own analysis, size positions so that being wrong is survivable, and consider speaking to a licensed financial professional about your own circumstances before trading.
Writes about AI-assisted day trading, technical analysis, and the systems traders actually use to stay disciplined.
A point and figure chart is a filter with two settings. It will show you a clean breakout on a stock that did not deserve one.
The filter is the feature right up until it hides the reason a setup was bad. Screenshot the chart you are actually trading from and SnapPChart reads that one image against a fixed rubric, then returns a setup grade, an entry, a stop with the reasoning behind the level, targets, and the reward-to-risk those levels imply.