DRAG
F2 MATH · MATHEMATICAL THINKING IN EXCEL

Math you can
actually use.
In 2 hours.

The mathematical thinking behind data science, finance, CFA, and quant — built by hand in Excel.

Become the one they ask when the numbers get serious. Five formulas hide under every field. Most read them. Few build them — every field reads clearer.

One Excel file. F2 opens every formula across cells. Two hours. $15.

What the 99% skip. What the 1% build.

Get D1: Foundation · $15 See what's inside

Excel .xlsx · instant download · yours forever

PNAS, 225 studies: hands-on +6%, failure −55%.

THIS IS A Z-SCORE
Look easy. Easy to miss.
Two listings. Which one is the real deal?
Listing A · Brooklyn apartment
$750k
$620k
$750k vs $620k = +$130k above area
Listing B · Manhattan apartment
$1.20M
$1.50M
$1.20M vs $1.50M = −$300k below area
+21%
Listing A
−20%
Listing B
Listing B is the real deal.
B looks expensive ($1.20M vs $750k). But it's −20% below Manhattan; A is +21% above Brooklyn. The price tag isn't the deal.
That's a z-score — how far from the local norm, not the raw number. Same intuition reads home prices, salaries, KPI deltas, A/B tests, poll margins, lab readings. The basics that look small are the rare ones. Build them, every field reads clearer.
225studies behind the method
8sheets
500+live formulas
0hardcoded values
HOW IT WORKS

Two keys. That's it.

Don't memorize. Interact.

F2 traces every Excel formula across cells. ESC backs out safely. Together they turn variance, standard deviation, z-score, and regression from black-box functions into arithmetic you can see step by step.

esc
F1
F2
F3
F4
F5
F6
F7
F8
F9
F10
F11
F12
CLOSE
Back to result
OPEN
See the recipe

Press F2 to look.   Press ESC to come back.

Repeat as often as you like — nothing ever changes.

BEFORE F2
Just a number.
C
D (Sales)
9
A. Johnson
88
10
B. Smith
72
11
C. Williams
95
…
…
…
28
T. Scott
85
40
Mean →
77.85

20 employees, one Sales column. Mean = 77.85

You see a result.
Where does 77.85 come from?

AFTER F2
The recipe opens.
C
D (Sales)
9
A. Johnson
88
10
B. Smith
72
11
C. Williams
95
…
…
…
28
T. Scott
85
40
Mean →
=SUM(D9:D28)/COUNT(D9:D28)
= (88 + 72 + 95 + … + 85) / 20
= 1557 / 20
= 77.85
Check: =AVERAGE(D9:D28) → 77.85 ✓

20 numbers, sum them, divide by 20.
No black-box function. Just arithmetic you can see.

PRESS F2 ANYWHERE

What "trace by hand" actually looks like.

A harder concept — variance. The formula SQRT(VAR(C9:C28)) usually hides the answer. F2 opens it. Each step is one cell, pointing to the one before.

F2 →
D29  = SQRT(D28)
← standard deviation = 8.59
opens
D28  = D27 / (20-1)
variance — why minus 1?
opens
D27  = SUM(E9:E28)
add up the squared distances
opens
E9  = (C9 - $C$29)^2
squared distance from mean
leaf
C9  = 88
raw value — your input, change it

Five cells. Five steps. One press at a time. Most courses jump from "variance is spread" to SQRT(VAR(…)). F2 Math sits in between: every step is its own cell, and F2 walks you back to the raw number you can change.

REMEMBER
  • F2 only looks. It never changes anything. Safe to press anytime.
  • ESC always brings you back. Nothing is ever lost.
  • Yellow cells are yours to change. Watch the chart react instantly.
CURRICULUM

Five formulas. In the right order.

D1: one variable. D2: two variables. Same dataset across both. Same five formulas under data science, finance, CFA, quant, AI.

D1 covers mean, variance (with N−1 explained), the normal distribution, and standardization (z-score). D2 covers covariance, correlation, simple linear regression, and R² — everything a junior data analyst or CFA Level 1 candidate is asked to explain.

D1 · ONE VARIABLE · $15

When you look at a single column of numbers, four questions arise naturally.

1
CENTER

Why the mean
isn't enough

Sum and count split into three steps. See the moments where the mean misleads.

1_Mean
WHERE YOU'LL USE THIS
Reports: KPI averages · Students: GPA · Work: team metrics
2
SPREAD

Same mean,
different distributions

Deviation → square → sum → ÷(N−1) → √. A degrees-of-freedom experiment cell answers "why N−1?" by hand.

2_Variance
WHERE YOU'LL USE THIS
Finance: volatility · Data Science: feature spread · Quality: process control
3
SHAPE

A score of 60 —
top what percent?

Measure where 68-95-99.7 actually comes from. The math behind "2σ outlier."

3_NormalDist
WHERE YOU'LL USE THIS
Students: percentile rank · Reports: outlier detection · Polls: margin of error
4
COMPARE

Math 90 vs
English 80?

Put both on the same scale with z = (x−μ)/σ. The math behind StandardScaler.

4_Standardize
WHERE YOU'LL USE THIS
Data Science: StandardScaler · Finance: cross-asset comparison · Students: subject-fair grading

Convinced D1 covers what you need? Skip ahead.

Get D1: Foundation — $15

Or keep reading — D2, fields, FAQ below.

Once one variable makes sense, two-variable relationships follow naturally
↓
D2 · TWO VARIABLES · $19

When two columns sit next to each other, five new questions emerge — ending at the bridge to machine learning.

1
RELATE

Do two things
move together?

Covariance: deviation product summed. Sign tells direction, magnitude tells closeness.

1_Covariance
WHERE YOU'LL USE THIS
Finance: asset relationships · Reports: ad spend vs sales
2
MEASURE

How strongly?
Numerically.

Pearson r — covariance scaled to −1 to +1. Compare relationships across any two variables.

2_Correlation
WHERE YOU'LL USE THIS
Data Science: feature selection · Students: study-time vs grade
3
PREDICT

Draw the
best line

Slope = Cov/Var. Intercept = mean shift. The same line behind every regression model.

3_Regression
WHERE YOU'LL USE THIS
Reports: revenue forecasts · Finance: factor models
4
EVALUATE

How well
does it fit?

R² from SST decomposition. Residuals show what the line misses, not just the score.

4_Residual_R2
WHERE YOU'LL USE THIS
Data Science: model evaluation · Research: statistical reports
5
BRIDGE TO ML

From a line
to a neural net

The leap is smaller than textbooks say. Loss functions and gradient descent grow out of regression.

5_DeepLearning_Bridge
WHERE YOU'LL USE THIS
AI/ML: gradient descent intuition · Coding: PyTorch / scikit-learn

Ready for D2? Or grab both — D1 first, D2 follows naturally.

Get D1: Foundation — $15 Get D2: Regression — $19

Same 20-person dataset across both. D2 builds on D1.

WHY THIS ORDER

It's the order you naturally ask. "What's the center? How spread? What shape? How to compare?" — then with two columns: "Do they relate? How strongly? Can I predict? How well does it fit?"

These nine steps are the skeleton of statistics, finance, machine learning, and every quantitative field. Once you have the foundation, the rest stops feeling like memorization.

Plus onboarding (0_Guide, 0_LearningPath), deep dive (Appendix_NORM), and wrap-up (Summary) sheets in each file. D1: 8 sheets · D2: 8 sheets · 16 total.

One file. $15. Done.

Excel statistics workbooks — no subscription, no video course, no Coursera-style 8-week commitment. About two hours of focused work, yours to keep.

D1: FOUNDATION · ONE VARIABLE

One-variable analysis

Mean · variance (degrees of freedom) · normal distribution · standardization

WHAT YOU'LL BE ABLE TO DO
  • → Spot the traps the average hides (work KPIs)
  • → Explain why N−1 in 30 seconds (stats interviews)
  • → Read "top 7%" with confidence (grading, polls)
  • → Compare across different scales (StandardScaler, finance)
WHERE IT SHOWS UP
Data Science: StandardScaler, outlier detection
Finance / CFA: volatility (σ), VaR, normality checks
Work Reports: reading KPI averages honestly
Clinical Trials: baseline distributions
$15
Yours forever · free updates
Get D1: Foundation — $15
D2: REGRESSION · TWO VARIABLES

Two-variable relationships

Covariance · correlation · regression · R² · bridge to ML

WHAT YOU'LL BE ABLE TO DO
  • → Tell whether two variables actually move together (asset correlation)
  • → Read R²=0.85 and know what it really claims (regression diagnostics)
  • → Defend correlation vs. causation in interviews (data interview)
  • → See the bridge from arithmetic to ML (linear regression → ML)
WHERE IT SHOWS UP
Data Science / ML: feature correlation, regression baseline
Finance / CFA: factor models, beta, asset correlation
Quant: pairs trading, portfolio diversification
Work Reports: ad spend vs revenue, KPI relationships
$19
Yours forever · free updates
Get D2: Regression — $19
NO REFUNDS

The price is low and refunds aren't offered. Read this page carefully — the method, the questions it answers, what's inside D1 and D2 — and decide before you buy.

Secure checkout via LemonSqueezy · instant download after purchase

WHAT THIS COSTS ELSEWHERE

Stats textbook
$40+
OpenIntro · Khan
Coursera
$49/mo
Google Data Analytics
Online tutor
$40/hr
Wyzant · Preply
F2 Math
From $15 · forever
Start with D1
WHERE OTHER COURSES STOP

Five questions lectures skip.

Every stats course covers them in one line. F2 Math answers them in cells you can trace.

These are the staple confusions in introductory statistics — the same ones that show up in data analyst interviews, CFA Level 1 quantitative methods, and the first weeks of any machine learning course.

Q1

"Why divide variance by N−1, not N?"

A staple data interview question. Most courses dispatch it in one line — "because it's a sample."

D1 Foundation
MOST COURSES

"It's a sample, so we divide by N−1."
"It's called Bessel's correction."
"We need an unbiased estimator."

→ "But why subtract 1? Where does the 1 come from?" Still unanswered.

F2 MATH'S ANSWER

Say 4 values average to 10. You can pick the first three freely (e.g., 8, 9, 12). Then the fourth is forced — it has to be 11.

The mean used up one degree of freedom. Variance measures distance from the mean, but the mean is already "spent." That leaves N−1 free.

D1 Foundation's 2_Variance includes a degrees-of-freedom experiment cell. Enter 3 values; watch the 4th lock in. Once you do this by hand, you don't forget it.

Q2

"Is R² = 0.85 a good model?"

"86%, sounds good" — vague, and not enough to make a real call on.

D2 Regression
MOST COURSES

"R² is the coefficient of determination. Closer to 1 is better."

→ So is 0.85 good? What about 0.6? Does it depend on the field? No answer.

F2 MATH'S ANSWER

R² = 0.85 means the regression explains 85% of the total variation in y (SST). Decompose SST = SSR + SSE, and the meaning becomes visible in the cells.

In physics, R² = 0.99 is the standard. In marketing, 0.4 is solid. In behavioral science, 0.2 can be meaningful. Without the context of the field, R² alone tells you nothing.

D2 Regression's 4_Residual_R² decomposes SST · SSR · SSE in cells. The real meaning of R², in one pass.

Q3

"What does sklearn's StandardScaler actually do?"

Every ML tutorial says "use it." Few explain what it does or why.

D1 Foundation
MOST ML TUTORIALS

"scaler = StandardScaler()"
"X_scaled = scaler.fit_transform(X)"
"Done!"

→ A black box. Whatever happens inside stays inside.

F2 MATH'S ANSWER

All it does is z = (x − mean) / standard deviation. That's the whole thing.

Why use it? Height (170 cm) and weight (60 kg) are on different units, so they can't be compared directly. Convert both to z and they're on the same scale — which is what most ML models assume.

D1 Foundation's 4_Standardize: z-score in 3 steps, with scatter plots side by side. Original vs. standardized — StandardScaler in plain view.

Q4

"Correlation 0.9 — does that mean causation?"

Mixing up correlation and causation happens in reports, news, and papers every day.

D2 Regression
MOST COURSES

"Correlation ≠ causation." End of explanation.

→ But what does the correlation coefficient actually measure, and why isn't it causation? Not addressed.

F2 MATH'S ANSWER

r measures how much two variables move together. Causation asks whether A produces B. Two different questions.

Ice cream sales and drownings show r = 0.8. Cause? Hot weather drives both — the classic third-variable trap.

D2 Regression's 1_Covariance · 2_Correlation breaks down the path from products of deviations to r — cell by cell — so you see exactly what r measures.

Q5

"Covariance is 720 — is that big or small?"

Change the units and the number changes by 1,000×. Compared to what?

D2 Regression
MOST COURSES

"Covariance is how much two variables move together."

→ So is 720 big or small? No way to tell.

F2 MATH'S ANSWER

You can't tell. Covariance carries the units — sales ($) × projects (count) gives "$·count." Switch sales to thousands of dollars and 720 turns into 0.72.

That's why dividing by both standard deviations strips the units away — what's left is r. r = 0.95 is strong, 0.3 is weak. Comparable across any setting.

D2 Regression's 2_Correlation has a unit-conversion experiment cell and the full derivation of r. See for yourself why r is unit-free.

Which of the five blocked you most? Here's where each is taken apart.

D1 FOUNDATION · $15
One-variable analysis

Mean · variance (degrees of freedom) · normal distribution · standardization (z-score)
Answers Q1 and Q3

D2 REGRESSION · $19
Two-variable relationships

Covariance · correlation · regression · R² · bridge to ML
Answers Q2, Q4, and Q5

Questions.

Common asks about prerequisites, time commitment, data analyst interview prep, and how F2 Math compares to a Coursera or DataCamp course.

Do I need to be good at Excel?
No. The first sheet (0_Guide) covers the basics — arithmetic operators, cell references (relative and absolute), and how to use F2 to trace. If you can type =A1+B1, you're ready.
How long does it take?
Depends on what you're after. Path A (30 min) is enough to tell if the file fits you. Path B (2 hours) covers the three staple interview questions. Path C (~6 hours, including Appendix_NORM) is the full pass. The 0_LearningPath sheet walks through all three.
How much math do I need going in?
Arithmetic (+, −, ×, ÷), exponents (^), and square roots (√). That's enough to take every formula apart. No need to remember stats or linear algebra — that's what the file teaches.
Will this actually help in data interviews?
"What is variance?" "Why N−1?" "What is standardization?" are staples of data analyst and data scientist interviews. The file decomposes each one in cells so a clean 30-second answer comes naturally. The point isn't to memorize a script — people who've traced it with F2 explain it on their own.
How is this different from a textbook?
Textbooks show formulas. F2 Math shows every intermediate step as a live Excel formula. Change an input and the whole chain updates. Where most textbooks dispatch "why N−1?" or "why square?" in a single line, this file answers with experiment cells and side-by-side measurements.
Does it work in Google Sheets?
Most formulas work. But the core learning experience — tracing formulas with F2 — works best in desktop Excel (2016 or later). Mobile Excel and LibreOffice handle parts of it, but desktop Excel is recommended.
How do I receive the file after buying?
Right after checkout, an email arrives with the download link. One Excel file plus three PDFs (F2 trace checklist, 30-second answer scripts, learning-path guide). Checkout is by card via LemonSqueezy.
Do I get future updates?
Yes. The latest version shows up automatically in your LemonSqueezy account. Free updates for life — typo fixes, expanded notes, anything — buyers always have the current version.
Are refunds available?
No refunds. The trade-off is the low price (D1 $15, D2 $19). Read this page carefully — the method, what's inside, and the questions it answers — and decide before you buy.
Will other topics come out? (regression, matrices, etc.)
The series starts with two files. D1 Foundation (one-variable: mean, variance, normal distribution, standardization) and D2 Regression (two-variable: covariance, correlation, regression, R²). Both share the same 20-person dataset, so D1 leads naturally into D2. Further topics (inference, time series, etc.) depend on how the first two are received — no promises.

Build the math. Once. By hand.

Arithmetic is enough. Two hours. The same five formulas under data, finance, CFA, quant, AI.
D1: $15. D2: $19.

Get D1: Foundation — $15 See both

Yours forever · instant download · no refunds