I’m collaborating with someone on strategy development. For backtesting, I’m using WL7, and he’s using Python. We’re planning to use Alpaca for brokerage. Between the two of us, we’ve tested data from three different sources: IQ Feeds, Alpaca, and IEX. We have observed some differences in the data.
For example, in my own tests with WL7, without changing anything about my strategy, two year backtest timeframe, or the symbols, there was an instance where the 5-minute scale data from IQ Feeds showed about ten times greater profit than the data from Alpaca.
In addition, when looking closely at the 5-minute scale data from Alpaca, there was a case on the morning of August 12 where three bars of data were missing on MSFT’s chart. Other data sources do not have the same problem with missing data for MSFT then.
Is the 5-minute scale Alpaca and IEX data of the same quality as the data from IQ Feeds? What data provider do you recommend for strategies that run on a 5-minute scale?
For example, in my own tests with WL7, without changing anything about my strategy, two year backtest timeframe, or the symbols, there was an instance where the 5-minute scale data from IQ Feeds showed about ten times greater profit than the data from Alpaca.
In addition, when looking closely at the 5-minute scale data from Alpaca, there was a case on the morning of August 12 where three bars of data were missing on MSFT’s chart. Other data sources do not have the same problem with missing data for MSFT then.
Is the 5-minute scale Alpaca and IEX data of the same quality as the data from IQ Feeds? What data provider do you recommend for strategies that run on a 5-minute scale?
Rename
QUOTE:
In addition, when looking closely at the 5-minute scale data from Alpaca, there was a case on the morning of August 12 where three bars of data were missing on MSFT’s chart.
You can try to reload the chart's data (make sure the source reads Alpaca at the bottom of the chart) by right clicking on it. Does it help?
QUOTE:
there was an instance where the 5-minute scale data from IQ Feeds showed about ten times greater profit than the data from Alpaca.
You have to explore the backtest trade by trade to find problematic ones.
Re: IQ Feeds. It's called IQFeed, I've renamed the topic title.
QUOTE:I'll ask my collaborator to give it a try -- he's the one who encountered the problem.
You can try to reload the chart's data (make sure the source reads Alpaca at the bottom of the chart) by right clicking on it. Does it help?
QUOTE:
You have to explore the backtest trade by trade to find problematic ones.
I'll take a look through the trades, then, to see if I can find the irregularities.
I would value any general input you may have to offer about the quality of the data from the three providers I listed above. Do you pay for IQ Feed data or use a free provider in your own backtesting and trading?
In my experience the Alpaca historical data has many anomalies and I'd never rely on it for backtesting. They just released a new data API (v2) but we haven't had the time to implement it for WL7 yet, so I can't speak to its quality.
Thanks, Glitch. I'll avoid using Alpaca's data then for the time being. Hopefully their new data API will be more reliable.
Do you know anything about the quality of the data from IEX?
Do you know anything about the quality of the data from IEX?
No, I haven't used it personally. I did read Flash Boys so I know about the origins of the IEX exchange but we haven't had sufficient demand yet to implement an interface to their data.
Isn't Tiingo providing intraday data by IEX ?
https://www.wealth-lab.com/extension/detail/DataExtensions
https://www.wealth-lab.com/extension/detail/DataExtensions
In general I would say it is very difficult to find a good and reliable ID data provider, regardless of the price you pay. Even if you take it straight from the exchange.
Take a look here: https://www.wealth-data.com
This is just showing the difference of EOD data. If you use ID data the difference will/could be similar on any bar of the day. On WD the intra day spikes are ignored and only the open, high, low and close are investigated. IT is a fairly complex formula to filter out spikes or funny opening prices.
That is why it my opinion that if you wish to trade on ID data you need to start cleaning the data first before backtesting. Then you need to run the same formula during the da before your system generates any signal.
Just trading a few highly liquid symbols might make the trading easier but it still requires to clean the historical data.
Take a look here: https://www.wealth-data.com
This is just showing the difference of EOD data. If you use ID data the difference will/could be similar on any bar of the day. On WD the intra day spikes are ignored and only the open, high, low and close are investigated. IT is a fairly complex formula to filter out spikes or funny opening prices.
That is why it my opinion that if you wish to trade on ID data you need to start cleaning the data first before backtesting. Then you need to run the same formula during the da before your system generates any signal.
Just trading a few highly liquid symbols might make the trading easier but it still requires to clean the historical data.
Your Response
Post
Edit Post
Login is required