Datasets

Data issues can occur in both historical and live data providers. Some common examples of data issues include the following:

• Missing or incorrect values
• Splits and dividends
• Listings and delistings
• Ticker changes

Some data that looks wrong is correct. Before you report an issue, rule out the following causes. A quick search for news about the asset on the dates in question usually explains them.

Missing Intraday Data

Illiquid assets can go minutes or hours without a trade, so there is no trade bar to deliver. Enable fill-forward or check the quote bars for the same period. Some contracts, like out-of-the-money Option strikes and back-month Futures contracts, rarely trade at all.

Missing Dates

Trading halts, temporary suspensions, and delistings stop the data. Data also doesn't exist before an asset's IPO date or on exchange holidays. Early closes shorten the trading day, so check the market hours of the asset before you report a missing afternoon.

Price Jumps

Splits, reverse splits, and dividends create discontinuities in the raw price series. If you compare adjusted prices with raw prices from another source, the values differ by the adjustment factor. Adjusted prices also change retroactively every time a new corporate action occurs. For more information, see Data Normalization and Adjusted Prices.

Ticker Changes

When a company changes its ticker, the historical data stays under the ticker the asset traded under at the time. If a security looks like it has no history, check for a symbol change.

Differences From Other Platforms

Prices, volumes, and timestamps can legitimately differ from other platforms because of the tape coverage, the time zone, and the time the bar is stamped. For more information, see Cross-Platform Discrepancies, Timezone Differences, and Misaligned Timestamps.

Outliers in Tick and Second Data

Single trades far from the prevailing price are often real. Odd-lot prints, flash crashes, and opening and closing auctions can all produce them.

To view the list of current data issues, log in to the Algorithm Lab and then, in the left navigation bar, click Datasets > Data Issues. Before you report a new issue, review the list of current issues to ensure that the issue is not already reported. The number of open data issues can sometimes be large, but our Data Team works on resolving them as quickly as possible while prioritizing the most important ones.

When all of the QuantConnect members report the data issues that they find, we can ensure the datasets are high quality for everyone. The easier it is for our Data Team to detect and reproduce the issues you report, the faster we can resolve them. If you find an issue with the historical data of a dataset, follow these steps to report it:

1. Log in to the Algorithm Lab.
2. Open the Data Issues page.
3. Click the Report Issue tab or the + Report Issue button.
4. Fill out the form. Follow the best practices to help our Data Team reproduce the issue.
5. Follow these steps to attach a backtest or notebook that demonstrates the issue:
1. Click Attach Backtest.
2. Click the Project field and then select the project from the drop-down menu.
3. Click the Backtest field and then select the backtest from the drop-down menu.
6. Click Publish.

History requests in live algorithms still use the historical datasets, so report issues with them as described above. The real-time stream comes from the data provider you select when you deploy, so the Data Issues page doesn't track it. If you encounter an issue with real-time data, first confirm that the data provider supports the asset. We recommend the QuantConnect data provider as the first choice. If the issue persists, open a support ticket and attach the project and the live deployment.

The easier it is for our Data Team to reproduce an issue, the faster we can resolve it. Follow these best practices when you report a data issue:

Rule Out False Positives

Review the common causes of data that only looks wrong before you report it. A search for news about the asset on the dates in question explains most of them.

Compare With Other Sources

Check the same asset and period on another data provider or charting website and state what it shows. Keep in mind that other platforms can legitimately differ from our data.

Plot the Data Instead of Logging It

Don't log prices. Logs are hard to inspect and count against your log quota. Plot the affected series instead, so the gap or spike is visible at a glance.

Attach a Screenshot

Attach a screenshot of the chart or notebook output that shows the issue. Mark the dates and values you consider wrong.

Attach a Backtest or Notebook

Attach a backtest or notebook that reproduces the issue with the smallest algorithm possible. Subscribe to the affected asset at the affected resolution, over the affected dates only. You can also paste the code in the body of the report.