Jane Street Programs & Competitions

Overview

Jane Street is the world's most elite quantitative trading firm. They don't run traditional "competitions" — instead they run selective programs that function as competitions. Getting into these programs is equivalent to winning a competition.

---

1. Monthly Puzzles — janestreet.com/puzzles/current-puzzle/

What it is: Monthly math/logic puzzles published publicly. Anyone can solve and submit. Prize: Top solvers listed on the leaderboard. No cash — but huge prestige. Difficulty: Very hard. Requires combinatorics, probability, optimization. Why enter: Jane Street recruiters explicitly watch the leaderboard. Consistent solving = cold outreach from them. Archive: janestreet.com/puzzles/archive/index.html — practice on past puzzles NOW.

---

2. FTTP — First Year Trading & Technology Program

What it is: 4-day immersive program for first-year undergrads. Mock trading + classes on markets + probability. 2026 HK dates: March 29 – April 1, 2026 Eligibility: First-year undergrad, STEM focus Cost: Free (Jane Street covers costs) No sponsorship required How to apply: janestreet.com/join-jane-street/programs-and-events/fttp/ Why it matters: Direct pipeline into Jane Street internship consideration

---

3. AMP — Academy of Math and Programming

What it is: 5-week summer program for exceptional high school graduates 2026 dates: June 29 – July 31, 2026 Eligibility: Graduating high school 2026, or graduated 2025 and not yet started undergrad Cost: Free — Jane Street covers ALL living costs Focus: Advanced math, probability, programming How to apply: janestreet.com/join-jane-street/programs-and-events/amp/

---

4. IN FOCUS Programme — Jane Street IN FOCUS 2026

What it is: Diversity-focused insight program in New York Eligibility: Underrepresented students in trading/quant URL: Search "Jane Street IN FOCUS 2026"

---

Skills Jane Street Looks For

• Probability: Expected value, conditional probability, Bayes' theorem — deeply intuitive
• Mental math: Fast arithmetic, estimation
• Logic puzzles: Deductive reasoning under uncertainty
• Programming: Python, clean algorithmic thinking
• Trading intuition: Market making, adverse selection, bid-ask dynamics

How to Prepare

1. Solve Jane Street puzzles monthly (archive has years of problems)
2. Read: "A Mind for Numbers" + "Thinking in Bets"
3. Practice: mental math apps (e.g. Math Workout, Zetamac)
4. Study: "Heard on the Street" (quant interview bible)
5. Practice: Probability brainteasers from "50 Challenging Problems in Probability"

---

Electronic Trading Challenge (ETC) — Specific Winning Strategies

What It Is

A day-long programming contest. Teams connect to a simulated exchange (TCP/JSON protocol) and trade instruments against each other algorithmically in real time.

Confirmed Winning Approaches (Documented, 2nd place from 44 teams)

Infrastructure first:

• Language: Python (readability beats performance in 1-day format)
• Set up Git aliases before the competition — every second matters
• Add 0.01-second delays between consecutive trades — prevents server rejection ("connection refused by peer")
• Use randomized order intervals, not constant — prevents "broken pipe" errors
• Test locally before connecting to live exchange

Bonds (fair value = 1,000 fixed):

• Buy any bond offered below 1,000. Sell at 1,001.
• Simple, nearly risk-free, earns immediately but edges compress as all teams discover it.

ADRs (American Depository Receipts):

• Do NOT use book offers to estimate fair value — book prices are stale/manipulated
• Use the most recent trade price of the liquid ADR as your benchmark
• Apply ±10 USD margins around that trade price
• Hedge: after every 10 units, convert between liquid and illiquid form
• This fix alone moved one team from ~$0 to ~$2,000 per round profit

ETFs:

• Value the ETF as: weighted sum of bond + stocks (using known bundle composition)
• Update component valuations from most recent trade prices (never book prices)
• Apply ±30 USD margins (higher volatility than ADRs/bonds)
• Result: ~$7,000 per round — highest-margin product in documented competitions

The master principle: Use trade prices, never book prices, to estimate fair value for any instrument. Book prices are too noisy and often stale.

Kaggle: Jane Street Market Prediction

• Neural networks (autoencoders + MLP ensembles) dominated the leaderboard
• Core challenge: feature-response relationship changes over time — temporal overfitting is #1 cause of failure
• Fold strategy: split by day (not random shuffle) — prevents date-information leakage
• Imputation: use rolling mean of prior 100 observations for same feature_0 group
• XGBoost + custom imputer performed strongly in 2024 edition