What Moneyball Teaches Us: Don't Buy Players, Buy Wins — and Don't Count Questions, Count Points

A Poor Team's Problem

First, the story. The 2002 Oakland Athletics faced a rigged game: their payroll was roughly a third of the Yankees', and the previous season's three star players had all been lured away by richer clubs. General manager Billy Beane understood the math: if a poor team plays by the same rules as the rich ones — scouts' intuition, star signings, bidding wars — it doesn't even get a seat at the table.

The film is adapted from Michael Lewis's 2003 book of the same name, and it documents how that team survived on a different set of rules: not intuition, but data.

(English subtitles are on by default; if the embed won't play, watch it on YouTube.)

---

Don't Buy Players, Buy Wins

The turning point arrives with a young Yale economics graduate, Peter Brand (a fictionalized version of the real-world Paul DePodesta), who tells Billy the most important line in the film:

"Your goal shouldn't be to buy players. Your goal should be to buy wins."

What makes the line powerful is that it lifts the goal up one level. A team never actually wanted "players"; it wanted wins. Wins come from runs — and runs, it turned out, come from a number the entire baseball establishment was underpricing at the time: on-base percentage. The traditional market priced players by batting average, RBIs, looks, and the elegance of a swing, which meant the unglamorous players who simply kept getting on base were massively undervalued. What the A's did was buy "getting on base" in bulk, at a discount — "He gets on base" became the only sentence that mattered in the draft room.

That season, the team with a bottom-of-the-league payroll ran off a 20-game winning streak, an American League record at the time.

(Subtitles on by default here too; backup link here.)

---

Test Prep Has Its Own Batting-Average Myth

Now swap the stadium for TOEIC.

Exam prep also has a traditional pricing system: hours studied, workbooks finished, words memorized, streak days logged. None of these numbers is bad — but they are all "players," not "wins." You can accumulate a beautiful pile of study hours the way a team can accumulate star signings, and still lose.

Translate Brand's line into exam-prep language:

Your goal shouldn't be to do more questions. Your goal should be to buy more points.

Points (wins) come from a rising percentage of correct answers (runs), and that rise comes from investing your practice where the expected value is highest (on-base percentage). It's the same one-level lift of the goal — and, as in baseball, you can't do it by feel. You have to measure.

---

We've Been Buying You Points with Engineering All Along

If you've been following this blog, you've probably noticed: this entire app is built with exactly this kind of discipline.

• To estimate your score, we think like roboticists. Your "true ability" is a hidden state observed only through noise — the same class of problem as a robot not knowing where it is. Self-driving cars solve it with a Kalman filter; for right/wrong answer data, the natural member of that same Bayesian-estimation family is a decayed Beta–Bernoulli posterior — and that's what we run. See How Handy 990 estimates your TOEIC score.
• To teach you vocabulary, we use memory science. Frequency ranking decides which words deserve your time, the FSRS memory model decides when to review, and a binary-search checkup decides where you start. See How Handy 990 helps you learn vocabulary.
• To label difficulty, we use re-runnable rules, not gut feeling. Every question's difficulty is measured from structural features and calibrated Part by Part against a commercial mock-exam set. See the Difficulty Overview.
• Even drilling itself — we worked out the "why" first. The substance of language is convention, and conventions can only be absorbed through mass exposure — a position backed by four lines of testimony: philosophy, linguistics, memory science, and large language models. See In Defense of Drilling.

Analytics and engineering worked on a ball club and they work on self-driving cars; they work just as well on pushing a score upward. Our entire focus is on buying you the next 50 points.

---

Where Your On-Base Percentage Hides: the Sweet Trap of Part 1

Here's a real example from our data.

Part 1 is the most comfortable stretch of the whole exam: look at a photo, listen to short sentences. The questions are friendly, the feedback is instant, and drilling it feels great. A few of our users have actually completed our entire Part 1 question bank — we salute the persistence, sincerely.

But here's the cruel number: Part 1 has 6 questions out of 200 — 3% of the exam. It is the player with the gorgeous swing: delightful to practice, limited contribution to wins. The real on-base percentage hides on the other side: Parts 3, 4, 6, and 7 add up to 139 questions — nearly 70% of the exam. They're harder, longer, and less fun to drill — which is exactly why they're undervalued, just like the unglamorous players who simply kept getting on base.

So when you open Smart practice, you'll find we often don't leave you in cozy Part 1; we push you toward Part 3 conversations, Part 4 talks, and the long passages of Parts 6 and 7. It's not that we don't know how to keep you comfortable. It's that we're buying you points.

---

Outsource "What Do I Practice Today" to Us

The division of labor in Moneyball is actually very clean: players play; the analytics department decides where the effort should go. Both are full-time jobs, and nobody should be doing both at once.

Same with test prep. Your job is already hard enough — actually getting the English in: catching a conversation over the rattle of a commute, finishing a passage on whatever focus is left after work. The other job — "What should I practice today?" "How do I spend these 15 minutes on the bus for maximum expected value?" "Which Part is actually my weakness?" — hand that whole job to us.

We track every answer you give, measure your accuracy and speed on every Part and question type, and point your next practice session at the spot most likely to get you on base. Handy 990 wants to be your advisor and your friend: the friend practices with you; the advisor runs the numbers. That is the most valuable thing this app provides — not the nine-thousand-plus questions, but the brain that decides which question comes next.

There is no shortage of free question banks out there. But look at how they charge: quite a few products lock the very act of choosing which Part to practice behind the paywall — on the free tier, you practice whatever you're fed. That leaves you two endings: either you do all the thinking about what to practice yourself, or you diligently grind the questions with the worst return on investment — and "diligently doing the wrong thing" is exactly the tragedy Moneyball is about. We run it the other way around: all six Parts are always yours to practice, and choice is never what's locked. What we charge for is unlimited planning and the full reports — the answer to the hardest question, "where should I practice?" Others sell access to the roster; what we want to sell you is wins.

And, as always, the honest ending: Moneyball isn't a fairy tale — the A's still didn't win the championship that year, because data can't swing the bat for the players. Same here: no algorithm can learn English for you. What it can guarantee is that every swing you take is taken at the highest-expected-value pitch.

The rest is yours. Don't count how many questions you did — count how many points you bought.

---

Footnote: Six Words from the Clips

If a few words in the clips tripped you up — good, that's convention-exposure happening in real time. Take these with you:

Word	As used in the clips
misjudge (v.)	to get someone or something's value wrong — mis- (wrongly) + judge; the whole industry kept misjudging its players.
mismanagement (n.)	running something badly — same mis- family: mis- + management.
medieval (adj.)	literally "of the Middle Ages"; figuratively, hopelessly outdated — the film uses it for baseball thinking.
ostracized (v., passive)	excluded, shut out of the group. From ancient Athens' ostracism, where citizens voted to banish someone from the city.
leper (n.)	a person with leprosy; figuratively, someone everyone avoids — "treated like a leper."
aggregate (n./adj.)	the total, combined amount — in the aggregate; stats price a player's aggregate contribution, not one heroic night. TOEIC reading loves this word too.