# What to use instead of Quantopian

Quantopian took the platform down on 14 November 2020. The engine survived as an open-source fork; the free data and the fund paying for it did not.

*https://stockmarketstack.com/alternatives/quantopian*

## The subject

| Field | Value |
| --- | --- |
| Product | Quantopian |
| Status | retired |
| Stopped serving | 2020-11-14 |
| Website | https://www.quantopian.com |
| Category | Backtesting Frameworks & Algo Trading Libraries |

A free hosted platform for writing and backtesting Python strategies against bundled US equity data, paid for by a crowd-sourced hedge fund that licensed the best algorithms from its own community.

## Replacements

*Ordered by how much of the original’s job each one covers, closest first. Paid placement does not affect this order.*

1. [QuantConnect](https://stockmarketstack.com/tools/quantconnect.md) — The only thing here that replaces the platform rather than a part of it — hosted IDE, bundled data, backtest, live brokers and a forum, with a free tier.
2. [Zipline-reloaded](https://stockmarketstack.com/tools/zipline-reloaded.md) — Quantopian's own engine, maintained. Same Apache-2.0 licence and the same zipline import path, so an algorithm's imports survive the move.
3. [Blueshift](https://stockmarketstack.com/tools/blueshift.md) — Free hosted Python backtesting with minute data already in it — the closest echo of what Quantopian was to somebody starting out.
4. [QuantRocket](https://stockmarketstack.com/tools/quantrocket.md) — The same engine on your own hardware, with its own maintained Zipline fork and Moonshot beside it, wired to Interactive Brokers.
5. [NautilusTrader](https://stockmarketstack.com/tools/nautilus-trader.md) — For the half of Quantopian that was live trading. One strategy that backtests and trades unchanged, rather than two code paths.
6. [VectorBT](https://stockmarketstack.com/tools/vectorbt.md) — If what you actually used the platform for was sweeping thousands of parameter combinations, which Quantopian's event loop was always slow at.

## The product

Quantopian was a free, hosted place to write a Python trading strategy, backtest it against US
equity minute data that was already there, and — for the strategies that worked — have the company
license it for its own crowd-sourced hedge fund. Founded in 2011 by John Fawcett and Jean
Bredeche, it ran for nine years and taught a large share of a generation of retail quants.

The fund is what broke. In February 2020 Quantopian returned outside investors' money after its
strategies underperformed. In October 2020 Fawcett announced the free community offering would be
wound down, and on 14 November 2020 the platform was taken down; the team subsequently went to
Robinhood. The open-source libraries — Zipline, Alphalens, Pyfolio, Empyrical, Trading Calendars —
were left on GitHub, which is the only reason this page has anything to recommend.

## What to use instead

Nothing replaces it whole, because Quantopian was four things at once: an engine, a data feed, a
notebook and a buyer. Three of the four are replaceable.

If you want **the platform** — somewhere to write, run and deploy without assembling anything —
[QuantConnect](https://stockmarketstack.com/tools/quantconnect) is the closest by a distance, and it is where most of the
community went in 2020. Its LEAN engine is open source, its cloud is not free past a point, and
its API is nothing like Zipline's, so treat this as a rewrite rather than a port.

If you want **the engine** — the Pipeline API, the event loop, the code you already wrote —
[Zipline-reloaded](https://stockmarketstack.com/tools/zipline-reloaded) is Quantopian's own Zipline, maintained, with the
import path intact. What it does not have is data, and that gap is the whole of the work.

If you want **what Quantopian felt like** — a free notebook with market data already in it —
[Blueshift](https://stockmarketstack.com/tools/blueshift) is the nearest thing running today. Its API grew out of Zipline, so
`initialize`, scheduled handlers, the `data` object and Pipeline all read as familiar.

If you want **it on your own hardware**, [QuantRocket](https://stockmarketstack.com/tools/quantrocket) ships its own
maintained Zipline fork alongside Moonshot in Docker, wired to Interactive Brokers.

## What you give up

**The buyer.** Quantopian did not sell you a backtester — it paid for good strategies. Nothing on
this list does that, and no retail-scale market for licensed alpha has replaced it. If that was
why you were there, the honest answer is that the thing you want no longer exists.

**Data in the box.** The platform shipped US equity minute data, free, with survivorship handled
for you. [Zipline-reloaded](https://stockmarketstack.com/tools/zipline-reloaded) ships almost none by design; QuantConnect and
Blueshift bundle data on their own terms and their own tiers. Whatever you pick, the data is now a
line item and a decision, and it is usually the larger half of the bill — see
[market data APIs](https://stockmarketstack.com/categories/market-data-apis).

**Nine years of forum.** The threads are archived and read-only, and the people are dispersed.
Every remaining platform has a community; none of them has that one.

## Migration notes

**The import path is the good news and the limit of it.** Zipline-reloaded kept `zipline` as the
package name, so engine-level code moves with almost no edits. Anything importing from the
`quantopian` namespace — the research notebook's pricing and fundamentals helpers above all —
has no counterpart anywhere, because those functions were the data, not the engine.

**Build the bundle before you port the strategy.** The usual order of work is backwards here: a
ported algorithm cannot be tested until data exists behind it, and assembling a bundle takes
longer than the port. Decide where the bars come from first, and where a year of them lives —
[tick data storage](https://stockmarketstack.com/categories/tick-data-storage) if you are going below daily.

**Check what your new bundle does with delistings.** A price history that silently drops companies
that went to zero makes every long-horizon backtest look better than it was, and free end-of-day
sources usually drop them. [Norgate Data](https://stockmarketstack.com/tools/norgate-data) is in the catalogue specifically
for keeping them.

**The tearsheets survive.** `pyfolio-reloaded` and `alphalens-reloaded` come from the same
maintainer as Zipline-reloaded, so performance attribution and factor analysis do not have to be
rebuilt — which, after the data, is the second-largest thing people assume they have lost.

## FAQ

### When did Quantopian shut down?

John Fawcett announced the wind-down in October 2020 and the community platform came down on 14 November 2020, after nine years. Users had until that date to pull their work out through a Download Code button in account settings. Research and backtesting stopped with it, and no successor product was offered.

### Is Zipline dead?

The original is; the engine is not. Quantopian's repository stopped at a last commit on 14 October 2020 and a final PyPI release of zipline 1.4.1 on 5 October 2020. Zipline-reloaded has carried it since, under the same Apache-2.0 licence and the same import path, updated for current Python, pandas and NumPy. The same maintainer publishes pyfolio-reloaded and alphalens-reloaded, so the tearsheet and factor-analysis halves survive too.

### Why did it close?

The platform was free because the fund paid for it. In February 2020 Quantopian returned outside investors' money after its strategies underperformed, and the free community offering was wound down that October; the team then went to Robinhood. That is worth holding on to when picking a replacement — ask what pays for the free tier you are about to build on.

### Can I still run my old Quantopian algorithm?

Parts of it. On Zipline-reloaded the import path is unchanged, so the engine-level code largely still reads. What does not survive is anything under the quantopian namespace, the research notebook environment, and the bundled data every one of those functions assumed. Blueshift keeps the closest hosted echo of the API. Budget for supplying data rather than for a port.