Coming Soon - L1 Equity Quote Data!

Very exciting and wonderful update guys!
The more data the merrier. Much preferred to be upset now then later!!

I totally agree with Ted. You guys are doing a great job!

Thanks Arthur! =)

Very interested to use this new data.

How will this effect the alpha markets existing algos? It's likley that algos that were previous thought to be suitable may no longer be?

I guess having an overall, higher quality/realistic algo offering would amount to a better QC reputation amongst funds.

Looking forward to the performance impact on my algo's, exciting! Short side question relating to this: what will be the impact of including this data on the Alpha slippage model, is it still relevant or will this be depricated?

best,

Dirk

The alphas in the market will be stopped and redeployed (with their history intact but a mark on curve noting the change in the data source). If they continue to perform well with the new slippage factored in they'll be left to run. 

Separately we're building in an alpha-system-wide, automatic "stop" mechanism for alphas who no longer perform according to backtest. This should stop any which fail due to the new slippage.

The AlphaStream slippage model will be reduced but not eliminated. Even after accounting for the bid-ask spread there is still slippage on the filling of a large order. This slippage should be made proportional to the size of the trade. 

Fantastic news. Overly optimistic fills was really one of the main time killers when starting out (before knowing better). Not to mention all other potential uses. :-)

Will it be possible to find e.g. average spread in universe selection to e.g. filter out securities with too wide spreads?

Cool idea Petter; once installed we can start exploring expanding coarse universe to factor in spreads.

Hey Jared Broad 
That is a very good enhancement! It is so important (at time for Alpha's authors and for QuantConnect!) to prove that the Alpha on the marketplace are not profitting from any data anomally to make the QC Alpha Streams marketplace the reference!

Looking forward to it

Hi Jared! Any news on the progress, or an ETA, I'm so freaking excited to see what it does to my backtests : )

Best,

Dirk

=D Thanks Dirk getting there. We hit a roadblock in the final deployment of the data and needed to rebuild the entire 40TB data set. It's now processed and we're re-running the production update process to ensure it is updated each day, on time for the universe selection in live trading.

We're optimistic it'll be ready this week. 

Hey Jared! 

Hope you are well, and that this improvement will soon be effective!

Do you know when we will be able to try and test this L1 Equity Quote Data?

Best regards, and keep up the good work!

Hey Laurent! Getting there! We've green-lit the pull request and are not making the tests to ensure the redundant data processing is online from day-1 as well. We currently have full redundancy in the data stack and would like to keep it that way.

-----

One interesting change was the new coarse universe data dollar volume ranks change due to the new rules of what ticks will be included in the final volume math. Previously FINRA ticks were excluded from bars as they are not tradable (FINRA feed is post-trade reporting). This generated thousands of support tickets and forum questions from people wondering why we didn't match Yahoo.

We've opted to include this volume now to quiet these questions; in the hope that the more complete representation of volumes will improve those strategies sensitive to dollar-volume rankings; and technical indicators that require trade volume. 

Since the new data starts in 2007 we'll also go back and reprocess all the 1998-2007 data with these new tick rules to match the SIP feed data. This will ensure consistent ranking behavior from 1998-present.

Important Note #3: Dollar volume rankings of stocks will change with this new data. The analysis of the rank changes shows smaller ETFs (e.g. Brazilian Stock Market ETF) which previously had high ranks are now displaced by stock counterparts. We can see the deviation of the two data sources from last week's coarse universe within the plots below:

For a single day the differences are more obvious:

Which if you look at in detail is mostly movements of obscure assets out of the top 100, and popular assets into the top 100. We think this is a good change and although a breaking change, makes the data better overall. 

The further you go back in time the differences become less apparent. Presumably due to lower dark-pool volume which is reported to FINRA; this is the same scatter from 2008:

-----------

Ultimately the most important thing is that our backtesting matches live-trading. So once the new live SIP feed is installed we'll also have the FINRA dark pool data there for algorithm consumption! We'll email all the impacted live users and recommend they re-backtest their strategies to confirm the coarse selection process still run as they expect or to take their strategies offline for review.

Hi Jared: I'm super excited about this addition. Is this data in full effect for backtesting and live trading now? If not, how far out is it expected to be?

Again, super exciting stuff! Thank you and best regards

Thanks David we're excited! The backtesting can be this week. The live feed paperwork is signed and the prototype live feed is in place. We're doing load tests on the live data feed now to ensure the new increases in tick volume won't overwhelm the running algorithms. I'd ideally like to hold off deploying the backtesting data until we can swap the live feed at the same time. This might make it after market close Friday.

Hello All! This was deployed today! Hopefully, you start seeing more accurate backtests =) If you're trading very liquid assets you won't notice much of a difference, but illiquid smaller assets will see a spread between the buy and sell price they achieve. 

If you see anything you think is a bug to support@quantconnect.com ideally with a code snippet so we can repeat the issue.  

Hi Jared,

This is good news, well done to the team!

I tried one simple backtest and my result actually got somewhat *better*. I am not complaining... unless this reveals a deficiency. Any idea what could cause results to improve (perhaps other changes in the same release)?

Also, I am attempting other backtests now but they get queued for a very long time. Last one queued for 29 minutes and counting. Could this be due to many people re-submitting their backtests, or is it just a problem on my side?

Thanks !

Sorry for the false report -- we discovered a number of issues and had to immediately revert back to the old data and old LEAN code. We're fixing the issues in both the data and the quote modeling and will try again at deploying at the end of the day. 

Hey Jared: Just wondering about the status of the redeployment. Is it back up and running or still needing a little more time to fix up? I'm excited to play around with it =)