SAT Residuals: Actual Minus Predicted and What the Sign Means

A residual measures how far an observed data value is from the value predicted by a model.

Use:

$$\boxed{\text{residual}=\text{actual}-\text{predicted}}$$

That subtraction order matters.

If a model predicts 80 but the actual value is 87:

$$87-80=7.$$

Residual:

$$\boxed{+7}.$$

The model underpredicted by 7.

If the actual value is 74:

$$74-80=-6.$$

Residual:

$$\boxed{-6}.$$

The model overpredicted by 6.

The SAT can test the arithmetic, the sign interpretation, or what a pattern of residuals says about model fit.

Why residuals are an SAT data skill

College Board's Problem-Solving and Data Analysis domain includes two-variable data, models, and scatterplots. The current framework expects students to fit and interpret models and make predictions from data.

Residuals are the natural language of prediction error.

For the broader topic, see SAT Scatterplots and Models and SAT Line of Best Fit.

The formula: actual minus predicted

Write it exactly this way:

$$r=y-\hat y$$

where:

• $y$ = observed/actual value;
• $\hat y$ = model-predicted value.

A positive residual means:

$$\text{actual}>\text{predicted}.$$

A negative residual means:

$$\text{actual}<\text{predicted}.$$

Residual sign table

Do not memorize "positive = good" or "negative = bad." Residual sign describes direction of error, not model quality.

Worked example 1: calculate from a prediction equation

A line of best fit is:

$$\hat y=4x+10.$$

For:

$$x=6,$$

the model predicts:

$$\hat y=4(6)+10=34.$$

Suppose the actual observed value is:

$$39.$$

Residual:

$$39-34=5.$$ $$\boxed{+5}$$

The point lies 5 y-units above the fitted line.

Worked example 2: negative residual

Model:

$$\hat y=3x+7.$$

At:

$$x=8,$$

predicted:

$$31.$$

Actual:

$$26.$$

Residual:

$$26-31=-5.$$ $$\boxed{-5}$$

The observed point lies 5 units below the fitted line.

Visual interpretation on a scatterplot

At a given x-value:

• point above the line → positive residual;
• point below the line → negative residual;
• point on the line → residual 0.

Residual is a vertical difference, because it compares observed and predicted y-values for the same x-value.

It is not horizontal distance from the line.

Worked example 3: read residual without full calculation

A scatterplot's line of best fit predicts:

$$y=50$$

at $x=12$.

The observed point is:

$$(12,43).$$

Because the point is below the line, the residual must be negative.

Magnitude:

$$50-43=7.$$

So:

$$\boxed{-7}.$$

The most common residual trap: reverse subtraction

Wrong:

$$\text{predicted}-\text{actual}.$$

Correct:

$$\boxed{\text{actual}-\text{predicted}}.$$

A memory aid:

A-P = Actual minus Predicted.

Residual of zero

If actual equals predicted:

$$y=\hat y,$$

then:

$$r=0.$$

That does not mean the entire model is perfect. It means the model predicts that specific observation exactly.

Worked example 4: find the actual value from a residual

A model predicts:

$$72.$$

Residual:

$$+9.$$

Use:

$$\text{actual}-72=9.$$

So:

actual=81

$$\boxed{81}$$

Worked example 5: find the prediction from a residual

Actual value:

$$44.$$

Residual:

$$-6.$$ $$44-\text{predicted}=-6.$$

Add predicted:

44+6=predicted

$$\boxed{50}.$$

A negative residual means the prediction was above the actual value.

Residuals in context need units

Suppose:

• x = advertising spending in thousands of dollars;
• y = sales in thousands of units.

A residual of:

$$+3.2$$

means actual sales were:

$$3.2\text{ thousand units}$$

above the model's prediction.

Residual units are the units of the response variable $y$.

Worked example 6: contextual interpretation

A model predicts a house price of:

$$\$410,000.$$

Actual sale price:

$$\$395,000.$$

Residual:

$$395000-410000=-15000.$$

Interpretation:

The model overpredicted the sale price by $15,000.

Residuals and model fit

A good fitting model should have residuals that are:

• relatively small compared with the scale of the response;
• scattered around 0;
• without a strong systematic pattern.

A model with residuals that curve or trend may be missing structure.

The SAT may ask you to reason qualitatively about a residual plot.

Residual plot

A residual plot places:

• x-values (or fitted values) horizontally;
• residuals vertically.

The horizontal reference line is:

$$r=0.$$

Good linear-model pattern

Residuals scattered randomly above and below 0.

Suspicious pattern

Residuals follow a U-shape or systematic curve.

That can indicate a nonlinear relationship that a straight line does not capture well.

Worked example 7: curved residual pattern

Suppose residuals are:

• positive for small x;
• negative for middle x;
• positive for large x.

That U-shaped pattern suggests the line of best fit systematically misses curvature.

A nonlinear model may be more appropriate.

Do not simply say "some residuals are positive and some negative, so the model is good." Pattern matters.

Mean of residuals

For ordinary least-squares regression with an intercept, residuals balance around zero mathematically.

For SAT purposes, the practical takeaway is simpler:

A useful fit should not systematically overpredict or underpredict across the data.

If residuals are mostly positive in one region and negative in another, investigate pattern rather than only average sign.

Residual magnitude

Residual:

$$-20$$

has a larger absolute error than:

$$+5.$$

Magnitude:

$$|r|.$$

When comparing which observation is predicted most accurately, compare absolute residuals unless the question specifically asks about direction.

Worked example 8: which prediction is closest?

Residuals:

A: $+3$
B: $-1$
C: $+5$
D: $-4$

Closest prediction corresponds to smallest absolute residual:

$$|-1|=1.$$

B

Residual vs vertical distance

Residual is signed vertical difference.

Vertical distance is:

$$|r|.$$

So a point 8 units below a line has:

• residual: $-8;$
• vertical distance: $8.$

Pay attention to wording.

Residuals and interpolation

A prediction inside the observed x-range is an interpolation.

A residual can be calculated only when you also know the actual observed value.

A future predicted point with no actual measurement yet has no observed residual.

For prediction strategy, see SAT Line of Best Fit.

Residuals and correlation

A strong correlation can still have nonzero residuals.

Correlation summarizes strength/direction of linear association.

Residuals show observation-by-observation prediction errors.

They answer different questions.

Common trap 1: reverse the formula

Always actual minus predicted.

Common trap 2: positive residual means model overpredicted

Opposite.

Positive means actual exceeded predicted; model underpredicted.

Common trap 3: negative residual means bad observation

Residual sign is not a quality rating.

Common trap 4: compare signed values instead of absolute error

For prediction accuracy, compare $|r|$.

Common trap 5: forget response-variable units

Residual units match y-units.

Common trap 6: think a zero residual proves perfect model fit

It proves only that one point lies on the fitted line.

Common trap 7: ignore residual patterns

Small-looking individual errors can still show systematic curvature.

A fast residual routine

1. Find predicted y-value.
2. Find actual y-value.
3. Calculate: $\text{actual}-\text{predicted}.$
4. Interpret sign:
   • • → prediction too low;
   • • → prediction too high.
5. Attach y-units.
6. If comparing fit, inspect absolute size and pattern.

SAT-style practice

Question 1

Predicted 52, actual 60.

$$60-52=8.$$ $$\boxed{+8}$$

Question 2

Predicted 47, actual 41.

$$41-47=-6.$$ $$\boxed{-6}$$

Question 3

Residual $+12$, prediction 90. Actual?

$$90+12=102.$$ $$\boxed{102}$$

Question 4

Actual 75, residual $-5$. Prediction?

$$75-p=-5.$$ $$\boxed{p=80}$$

Question 5

Which is predicted most accurately: residuals $+4,-2,+1,-7$?

Smallest absolute residual:

$$\boxed{+1}$$

FAQ: SAT residuals

What is a residual?

The difference between an observed value and the model's prediction.

What is the residual formula?

$$\text{actual}-\text{predicted}.$$

What does a positive residual mean?

The actual value is above the predicted value; the model underpredicted.

What does a negative residual mean?

The actual value is below the predicted value; the model overpredicted.

What does residual 0 mean?

The prediction matches that observation exactly.

How do I identify a residual on a graph?

It is the signed vertical gap between the observed point and the fitted line at the same x-value.

What makes a good residual plot for a linear model?

Residuals should be scattered around zero without an obvious systematic pattern.

How to practice this on MastaPrep

MastaPrep starts with a short placement check, then builds a study plan around the SAT Math skills where you are losing the most points. If this one keeps costing you questions, it shows up in your plan rather than getting lost in a random practice set. Every question has a worked explanation, and the ones you miss come back for review a few days later, which is when you find out whether the fix has actually stuck.

See how MastaPrep works or try free SAT practice questions. The free plan needs no credit card.

Related MastaPrep guides

• SAT Scatterplots and Models
• SAT Line of Best Fit
• SAT Regression: Slope and Intercept
• SAT Linear Models in Word Problems
• SAT Probability and Statistics

Final takeaway

Residuals are prediction errors with direction:

$$\boxed{r=\text{actual}-\text{predicted}}$$

Positive residual:

model predicted too low.

Negative residual:

model predicted too high.

Then go beyond the sign: use residual magnitude to judge individual prediction error and residual patterns to judge whether the model may be missing structure.

---

Official SAT references:
https://satsuite.collegeboard.org/sat/whats-on-the-test/math/types/problem-solving
https://satsuite.collegeboard.org/practice/student-question-bank/math

Trademark note: SAT® is a registered trademark of College Board. MastaPrep is independent and is not affiliated with or endorsed by College Board.