📈 Build Your Own: Algotrading System
The most demanding data system one person can build alone — a market-data store that does not lie to you, a backtester that cannot see the future, execution against a broker that disagrees with your records, and the monitoring that fails closed. It teaches the engineering, never a strategy: what you leave with is the ability to look at any backtest and say what is wrong with it.
No lessons published yet — this is the roadmap, so you can see exactly what the course will cover before it lands.
Before you start: ETL & Data Pipelines — Series 1. The market-data layer in Series 1 of this course is an ETL pipeline with unusually unforgiving requirements — idempotent loads, gap detection and a vendor who restates yesterday's data — and this track applies those ideas rather than teaching them.
Before you start: Python for Data — Series 1. Every build in this track is code, and the ecosystem a reader will meet — dataframes, numerical libraries, notebooks — is Python's. The engineering transfers to any language; the examples do not.
Short on time? The 20 episodes marked core are the shortest path to being productive at work. Start there and pick up the rest whenever a job asks for them.
Series 1 · Market Data Is the Whole Problem
everything that makes historical prices quietly lie to you
- 01
Four Subsystems — and What This Course Will Not Give You
coreintermediate#foundationalcoming soon - 02
Bid and Ask — the Bar That Cannot Tell You What Happened First
coreintermediatecoming soon - 03
Splits and Dividends — the 50% Loss That Never Happened
coreintermediatecoming soon - 04
The Universe Has a Date — Survivorship, Restatements and Buying the Future
coreintermediatecoming soon - 05
Parquet, DuckDB or ClickHouse — Choosing on the Access Pattern You Actually Have
intermediatecoming soon - 06
Loading It — Idempotent, Gap-Checked, and Loud About Corrections
coreintermediatecoming soon - 07
Project — a Gap Report, an As-Of Universe, and a Split That Does Not Lie
intermediatecoming soon
Series 2 · The Backtester, and Why Most Are Wrong
a simulation that cannot see the future, and a result you can defend
- 01
Two Backtester Designs — and Why the Slow One Is the One You Trade
coreintermediatecoming soon - 02
Look-Ahead — Four Ways the Future Gets In, and the Object That Prevents It
coreintermediatecoming soon - 03
Costs First — Spread, Slippage, and the Turnover Arithmetic That Ends Strategies
coreintermediatecoming soon - 04
From Signal to Quantity — Why the Same Signal at Two Sizes Is Two Strategies
coreintermediatecoming soon - 05
The Metrics, and How Each One Lies — Including Time Under Water
coreintermediatecoming soon - 06
You Are the Search Algorithm — Why Most Backtested Edges Are Noise
coreintermediate#lopez-de-prado#foundationalcoming soon - 07
Project — Backtest It, Then Attack It Until You Know What It Proves
intermediatecoming soon
Series 3 · Execution — Where It Becomes Real
orders, fills, reconciliation, and a switch a human can throw
- 01
The Order State Machine — Where the Venue Changes the State, Not You
coreintermediatecoming soon - 02
The Request That Timed Out — Client Order IDs and Writing It Down First
coreintermediatecoming soon - 03
Reconcile Before You Trade — the Broker Is Right and You Are Wrong
coreintermediatecoming soon - 04
Paper, Then Small — and Feeding Real Slippage Back Into the Backtest
coreintermediatecoming soon - 05
Fail Closed — Limits Outside the Strategy, and a Switch That Cancels Everything
coreintermediatecoming soon - 06
What Runs When — and Why a Late Job Must Skip Rather Than Catch Up
intermediatecoming soon - 07
Project — Run It Unattended, Break It on Purpose, and Explain Every Difference
intermediatecoming soon
Series 4 · Living With It
monitoring, reproducibility, and deciding in advance how it ends
- 01
Alert on the Silence — Stale Data, a Position You Did Not Intend, a Job That Never Ran
coreintermediatecoming soon - 02
The Research Log — Write Down What You Expect, and Count Every Trial
coreintermediatecoming soon - 03
Decay — Deciding the Exit Rule While You Still Have Nothing at Stake
coreintermediatecoming soon - 04
Your Code, Your Responsibility — the Questions to Ask Before You Start
intermediatecoming soon - 05
Do You Need to Build This — and the Skill That Is Worth More Than a Strategy
coreintermediatecoming soon - 06
Project — the System, and an Honest Document Saying What It Proves
intermediatecoming soon
Not here, on purpose
- What to trade, which strategies work, and any claim about returns — not covered, deliberately and permanently. In this course, Series 2 · 25 teaches you to judge a strategy's evidence, which is the opposite service and the only honest one
- Financial, investment or tax advice — not covered. In this course, Series 4 · 15 says plainly that the rules differ by jurisdiction and that record-keeping and compliance are the reader's own responsibility
- Machine learning models, feature engineering and training loops — covered in AI & Data. In this course, the modelling step is deliberately a black box with a signal coming out of it, because every failure this track teaches happens on either side of that box and not inside it
- Time-series storage engines in depth — covered in the engine tracks — ClickHouse and DuckDB have their own courses. In this course, Series 1 · 20 chooses between them for this workload and explains the criteria, then stops
- Running the jobs, alerting and on-call — covered in Data Ops, Series 3. In this course, Series 3 · 25 and Series 4 · 00 cover only what is specific to a system that spends money while it is broken
- Low-latency and high-frequency engineering — colocation, kernel bypass, microseconds — not covered. That is a hardware and networking discipline for funded teams, and a course that gestured at it would teach nobody anything. This track builds systems that hold positions for minutes or longer, and says so
- Order book microstructure and market making — not covered as a subject. In this course, Series 1 · 05 explains bid, ask and spread because a reader cannot model costs without them, and goes no further