Portfolio123 vs QuantConnect: rented years of history, or code over all of it

Someone choosing between these two has already decided what they want to test: whether ranking stocks on fundamentals, and holding the top of the list, would have worked over a long stretch of history without seeing figures before they were published. Among the cards here they are the two hosted routes that include fundamentals in the subscription. QuantRocket sells Sharadar fundamentals as a separate purchase, and Composer tests rules on price data only. What is left is a decision about whether you will write code. Each product also meters the research in a different unit, and each falls short of point-in-time in a different way.

What the money meters

Portfolio123 prices years of history. Its plans give 5, 10, 15 or 20 years of point-in-time stock history at $300, $1,000, $1,500 and $2,400 a year on annual billing, or $35 to $278 a month month-to-month. Custom ranking systems start at the $1,000 Backtest plan, and full buy and sell rules at the $1,500 Portfolio plan. Each price covers one region, North America or Europe and the UK, and a second region costs 40% more. Compute is metered in resource units, which run from 400 to 10,000 by plan.

QuantConnect does not meter history at all. Its US Fundamental Data, from Morningstar, starts in January 1998, and its AlgoSeek US equity prices start in the same month, survivorship-bias-free. Both are mounted in the cloud on every tier, including the free one. What it meters is compute nodes, seats and orders. The free tier gives one backtest node and one research node, and caps a single backtest at 10,000 orders and insights. A Researcher seat raises that to 10 million. A 50-stock portfolio that replaces ten names a month places about 20 orders a month, so 28 years of it fits under the cap. Replace ten names every week instead and the same test needs about 29,000.

So the gap that matters is not $2,400 against $0. It is 20 years of history with a ranking interface against about 28 years with an empty code editor. Portfolio123 counts 4,300-plus factors and 430-plus functions in its formula language. Morningstar's dataset tracks about 900 fields. Those numbers are not the same unit either. On QuantConnect, any factor beyond a raw field is code you write yourself.

Two kinds of not quite point-in-time

Both vendors sell the history as point-in-time. Both say, in their own documentation, where it falls short.

Portfolio123's history keeps moving a little. FactSet supplies its fundamentals and estimates, and its data article says FactSet backfills figures released later, after an announcement. As a result, the same backtest can give different results when you run it on different dates. Portfolio123 tracks those backfills and has a setting to include or exclude preliminary data. The drift is ongoing and small, and you can choose how to handle it.

QuantConnect's history has just been replaced. Morningstar retired the feeds behind the US Fundamental Data dataset, and QuantConnect rebuilt the whole history, back to 1998, from the replacement feeds. Its migration page states the problem the old data had. Ratio, growth, per-share and average values were dated at the end of the reporting period, and the filing behind them came a median of 63 days later. Backtests on the old data could therefore read figures before they were public. Live algorithms never could. The new data is dated by the filing. About a third of the values the two versions share are different, usually by more than 20%. Coverage rises from 8,717 to 9,705 companies, which by itself changes a universe such as "the 500 most-traded stocks with fundamentals".

The dates matter if you are evaluating QuantConnect this month. The old data stays the default on LEAN's master branch until 10 October 2026. After that it sits on a separate branch until the end of October, and on 31 October it is retired for both backtesting and live trading. Any fundamental backtest you ran or read before the switch will not reproduce.

Where each one simply stops

Universe. Morningstar's US dataset excludes ETFs, ADRs and OTC stocks. A ranking that includes Canadian, UK or European listings, or that holds an ETF as a hedge, has no data on QuantConnect. Portfolio123 covers TSX and TSX Venture, the London Stock Exchange, Xetra, Euronext, SIX and the Nordic exchanges, and keeps ETF price history from 1999 on every plan. It stops at stocks and ETFs on daily bars. QuantConnect adds options, futures, forex, crypto and intraday data down to the tick, but second and tick resolution need a paid seat.

The data stays on the platform. Neither lets you take the history away. Portfolio123 truncates screen downloads to the first 500 tickers for recent dates, and leaves them blank ten days back, unless you hold your own FactSet or Compustat licence. QuantConnect includes the data for cloud backtests only. Downloading it to run LEAN locally is billed separately: the US Equity Security Master alone costs $600 a year on a Researcher seat.

Execution. Portfolio123 sends a rebalance as orders to Interactive Brokers or Tradier. At Interactive Brokers only individual PRO accounts whose number starts with U can be linked, and IB can take up to a week to process the link request. QuantConnect deploys the same algorithm live to more than a dozen brokers. That needs a paid seat, and the $84-a-month Researcher pack is the first one that includes a live node.

The recommendation

Portfolio123 if any stock you rank is listed outside the US, if the ranking needs to hold an ETF, or if you will not write code. The no-code ranking interface is what its price buys, and QuantConnect has no data at all for the non-US case.

QuantConnect once the universe is US common stock and you are willing to write Python or C#. For that job its free tier backtests over every year since 1998, which no Portfolio123 plan offers at any price. It is also the choice if you need a broker other than Interactive Brokers or Tradier. Treat any QuantConnect fundamental result from before October 2026 as unverified until you re-run it on the new Morningstar data.

The rest of the category, sorted by the job each engine is bought for, is on the backtesting frameworks page.