You can write an equation from a graph by reading two things off the grid: the y-intercept, where the line crosses the vertical axis, and the slope, the rise over run between any two clean points. For a straight line, those two numbers drop straight into y = mx + b. For a parabola, you read the vertex and one other point to build y = a(x β h)Β² + k.
Start With the Line Type Before You Read Anything
The first decision isn't a number β it's a shape. A straight line means you're building a linear equation. A smooth U-shaped or inverted-U curve is quadratic. Any curve that levels off into a flat horizontal ceiling or floor is likely exponential or a reciprocal, and those need a different template.
Graph-reading questions usually specify the family ("write the equation of the linear function shown") or supply a shape that makes it obvious. If the problem says nothing, look at the end behavior. Straight diagonal means linear; a curve that opens upward or downward with a turning point means quadratic; a curve hugging one axis means exponential.
Don't grab points before you've named the family. The template you choose decides which features matter, and reading a vertex off a line wastes time. A quick glance at the whole grid costs five seconds and saves a wrong answer.
If the graph is a function passing the vertical line test, one equation can describe it. If it fails β a vertical line at x = 4, for example β the equation is just x = 4, which is not a function at all.
Read the Y-Intercept First: The 5-Second Anchor
The y-intercept is where the line or curve crosses the y-axis, and it gives you b in slope-intercept form instantly. Find the vertical axis (the one running up and down) and trace the drawn line until it touches it. That y-value is your b.
Coordinates matter here. If the crossing sits two gridlines above the origin, and each line is worth one unit, b = 2. If the scale marks every 5 units, that same two-gridline crossing means b = 10. Read the axis labels, not the squares.
Write b down before hunting for slope. It's the least error-prone number on the whole graph because it sits at x = 0, so there's no arithmetic involved.
Watch for the trap: a graph whose y-intercept is negative reads "minus 3" not "3 below the line." A common student failure is dropping the sign because the eye registers "three squares down" as a magnitude. Keep the sign attached.
Measure Slope With Rise Over Run or the Formula
Slope is rise over run β how far the line climbs or drops vertically for each unit it moves right. Pick two points where the line crosses grid intersections exactly, count squares between them, and divide the vertical change by the horizontal change.
For points (1, 3) and (4, 9): rise = 9 β 3 = 6, run = 4 β 1 = 3, slope = 6/3 = 2. A descent works the same way. Between (0, 4) and (2, 0): rise = 0 β 4 = β4, run = 2, slope = β4/2 = β2. Rising lines have positive slope; falling lines have negative slope.
If the graph is flat (horizontal), slope is 0 and the equation is y = b. If the graph is vertical, slope is undefined and the equation is x = a β you cannot write it as y = mx + b, and that's a documented edge case worth flagging in any homework answer.
Verify your slope by checking a third pair of points. If the line passes through (1, 3), (4, 9), and (7, 15), the slope from the first pair and last pair should match. Consistency across three points confirms you read the scale correctly.
Assemble Slope-Intercept Form and Test It Back
Now plug m and b into y = mx + b. With slope 2 and y-intercept 3, the equation is y = 2x + 3. Rewrite with the actual coefficients, not the letters β the answer is the numbers in place.
Test the equation against a point you didn't use. If the graph also passed through (5, 13), check: y = 2(5) + 3 = 13. Match. If it doesn't match, one of your readings is wrong β usually the slope from counting too fast.
Slope-intercept form is the default for linear graphs. Point-slope form, y β yβ = m(x β xβ), works just as well and is often easier when the y-intercept isn't clean. Standard form, Ax + By = C, sometimes appears in textbooks; convert by rearranging.
| Form | Equation | Best when |
|---|---|---|
| Slope-intercept | y = mx + b | y-intercept lands on a gridline |
| Point-slope | y β yβ = m(x β xβ) | y-intercept is off-grid, another point is clean |
| Standard | Ax + By = C | Textbook specifies integer coefficients |
| Horizontal | y = b | Line is flat |
| Vertical | x = a | Line runs up-down |
A named failure example: students often report slope as run over rise and get 1/2 instead of 2. If you used the correct pair of points, the rise should be the larger number for a steep line. If your slope is less than 1 for a visibly steep line, you flipped the ratio.
Convert Between Forms Without Losing the Graph
Point-slope is the safety net when b is unreadable β say the line crosses at y = 2.33 because the axis scale is coarse. Pick a clean point like (3, 7) with slope 4/3, and write y β 7 = (4/3)(x β 3). Expand to slope-intercept: y = (4/3)x + 3.
Standard form asks you to clear fractions and shuffle terms. Taking y = (4/3)x + 3, multiply through by 3: 3y = 4x + 9, then rearrange to 4x β 3y = β9. The coefficients become integers, which is the textbook expectation.
If the answer key or auto-grader specifies a form, match it. Most online homework (Khan Academy, IXL) accepts slope-intercept by default, but printed worksheets often say "in standard form." Confirm on the assignment sheet before submitting.
Round answers only as much as the graph allows. A slope that appears to be 0.67 on a coarse grid is probably exactly 2/3, so write the fraction.
When the Line Isn't Clean Enough: Edge and Failure Cases
Not every graph crosses nice integer points. Some problems place intercepts at fractional coordinates deliberately. If the y-intercept looks like "about 2.4," and your slope is 3/2, then the exact intercept is probably 5/2 = 2.5. Check whether a fractional value recreates the other points.
Undefined slope and zero slope are the two failure modes students miss. Zero slope means a horizontal line; undefined slope means a vertical line β not the same thing. Zero is a defined number; "undefined" means the expression has no value.
Scaling errors are the other big trap. A graph labeled 0, 10, 20, 30 on the axes means each gridline is worth 10, not 1. Reading slope as 1/2 when the actual is 5 is the classic result β the ratio is right, the scale is wrong.
If the graph passes through (2, 8) and (6, 20) with each gridline worth 1, slope = (20 β 8)/(6 β 2) = 12/4 = 3. A student reading without the axis labels might say slope is 3 squares per 4 squares and answer 3/4. Reading the labels is non-negotiable.
Graph a Parabola: Read Vertex and One Point
Quadratic graphs need vertex form, y = a(x β h)Β² + k, where (h, k) is the vertex β the turning point. Find the highest or lowest point on the U-curve and read its coordinates. That gives you h and k directly.
Then pick one more clean point and solve for a. With vertex (2, 1) and point (4, 9): 9 = a(4 β 2)Β² + 1, so 9 = 4a + 1, and a = 2. The equation is y = 2(x β 2)Β² + 1. A positive a means opens up; negative opens down.
Sign discipline matters for h. A vertex at x = β3 means the term is (x + 3)Β², not (x β 3)Β². The form is (x β h), so a negative h produces a plus sign inside the parentheses. Getting this sign wrong is the most common quadratic error.
Parabolas with a negative leading coefficient look identical but upside down. If the curve peaks rather than bottoms out, your a will come out negative if you consistently solve β don't panic and flip the sign manually.
Nonlinear Graphs Beyond Lines and Parabolas
Exponential graphs take the form y = aΒ·bΛ£, where a is the y-intercept and b is the growth or decay factor. Read a off the y-axis, then find another point to solve for b. From (0, 3) and (2, 27): 27 = 3Β·bΒ², so bΒ² = 9, and b = 3. Decay works with b between 0 and 1.
Sine and cosine graphs are read by amplitude, period, and midline. Amplitude is half the distance between max and min; period is the horizontal length of one full wave; midline is the average of max and min. A wave oscillating between y = 1 and y = 5 has amplitude 2 and midline 3.
Radical and reciprocal graphs follow similar "read the shape, solve a constant" logic. For y = a/x with a known point, a = xy. For y = a/xΒ², a = xΒ²y. The pattern is always: anchor one feature, solve one constant.
None of these nonlinear reads work without a clear point. On a poorly scaled graph, picking a point that's "close" is a guaranteed wrong answer β find a grid intersection or note that the graph is approximate.
Check Your Equation Against the Whole Graph
Plug a third point back into your final equation. If it satisfies, your work is verified. If not, one of slope, intercept, or the vertex is off. Backtrack to the last number you can defend.
Verify monotonicity: if the graph climbs left-to-right, your slope must be positive. A negative slope on a visibly climbing line means you swapped points in the subtraction, giving the wrong sign. Fix the order: slope = (yβ β yβ)/(xβ β xβ) is what most textbooks prescribe.
Check the shape against the leading coefficient for parabolas and the base for exponentials. A U-curve with a negative a will graph as an inverted U β a visible mismatch. An exponential graph with b > 1 but drawn as decay means b was inverted.
Cross-check by trying x = 0. Every slope-intercept equation must produce b when x = 0, and every quadratic must produce a(hΒ²) + k when x = 0. If your x = 0 result doesn't match the visual y-intercept, keep working.
Plot It With a Graphing Calculator Instead of Guessing
If you have access to Desmos (free, web-based), paste your candidate equation and compare the plotted line to the original image. This costs about 60 seconds and catches sign errors the human eye misses. It's the rate-your-own-work method teachers seldom mention.
For a line through (0, -1) and (3, 5), type y = 2x β 1 into Desmos and zoom to match the original grid. If your line sits one unit above the printed graph, b should be -2 instead of -1. If the tilted line is mirrored, m has the wrong sign.
Physical graphing calculators (TI-84 Plus family) do the same job via the Y= menu, then GRAPH. A common workflow is entering two candidate equations in Y1 and Y2, checking which overlaps the printed original, and submitting the closer pick. Confirm on your specific model's menu path if uncertain.
Table-based verification beats all of it. Enter 3β4 x-values in a table, compute y from your equation, and see if the (x, y) pairs sit on the printed graph. Any mismatch flags the wrong reading.
Key Form: Two Points With No Axis Crossing
Some graphs never touch the y-axis cleanly; others give you exactly two labeled points and nothing else. Use the two-point formula: slope = (yβ β yβ)/(xβ β xβ), then plug one point into point-slope. From (β2, 4) and (3, β6): slope = (β6 β 4)/(3 β (β2)) = β10/5 = β2.
Substitute into y β yβ = m(x β xβ): y β 4 = β2(x + 2), which expands to y = β2x. The y-intercept comes out to zero even though no crossing is labeled β the algebra finds it.
When the graph gives three or more labeled points, pick the two safest ones β no decimals, both grid intersections. Using a labeled midpoint with a fraction is possible but risks rounding errors.
If the "graph" is really a data scatter with a line of best fit drawn through it, read the line, not the dots. Scatter plots are governed by the trend line equation, and dots above or below the line are expected.
FAQ
Does it matter which point I use for slope?
No, and that's a diagnostic. Slope is constant along any straight line, so two different pairs should produce the same quotient. If your slope comes out different for pair A versus pair B, you either misread the scale or picked a point the line doesn't actually pass through. Retest against a third intersection.
What's the most common mistake in this process?
Dropping the sign on a negative intercept or slope. A line crossing at y = β2 is often answered as y = 2, and a falling line read as m = 3 instead of m = β3 flips the entire graph's direction. Every negative result needs to survive a quick re-look at "does this graph climb or fall?"
Can I skip finding the y-intercept?
Yes, but only if you go through point-slope form. With slope and any point, y β yβ = m(x β xβ) writes the equation directly, and the intercept appears on expansion. This is the standard workaround when the y-intercept sits on a fractional coordinate and isn't readable to a clean number.
How do I handle a vertical line?
A vertical line has an undefined slope and cannot be written as y = mx + b. Its equation is x = a, where a is the constant x-value every point shares. This is a legitimate, expected answer in textbooks, and no amount of algebra will convert it to slope-intercept form.
What if the graph is curved but not parabola-shaped?
Classify by end behavior. If one side levels off into a flat horizontal asymptote, the form is exponential, y = aΒ·bΛ£, or a reciprocal, y = a/x. If it oscillates between two values in a smooth wave, it's sine or cosine. Each family has its own anchor feature (y-intercept for exponential, midline and amplitude for trig), and you solve one constant from any other clean point.
Do graphing calculators accept point-slope form?
Desmos does, as long as the syntax is correct (y β 4 = β2(x + 2)). TI-84 calculators require solving for y before entry in the Y= menu β enter y = β2x β 4, not the point-slope version. The output is the same plotted line, but the entry field differs. Confirm your specific model's manual if the equation won't plot.
Is there a quick way to test my answer without full re-derivation?
Enter 2β3 x-values from your chosen points, compute y from your candidate equation, and compare to the printed graph coordinates. Three successful matches (with at least one match on a point you didn't use during derivation) are strong confirmation. Every mismatch is a clue about which number β slope, intercept, or a quadratic constant β is wrong.