Is there any Blueshift goal to broaden the tradable universe from the S&P 500?
Yes we will broaden our universe to include US liquid ETFs for the US datasets and options for the Indian datasets. The ETA for this. however, is not clear. We are working on a few workstreams for some exciting feature release (coming soon!). We will pick up a host of data related activites following it, and this will be part of that.
thx for reply. I ask the question, because I'm comparing a quantopian model inside Blueshift, and find the results to be quite different. I assume universe selection plays a part (the first part worthy of investigation at least) and that is why I asked. QTradeable is quite different from S&P 500. if Blueshift only intends to add ETF's to the Blueshift Tradeable universe, that is still missing on equities from my perspective. How will quant enthusiasts do parity testing from one platform to another? Why wouldn't Blueshift pull in Russell 2000 for instance? only having 500 large cap equities, leaves a lot of testeabilty on the table, so to speak…
I see your point. I am sure we will add more on the equities side in the future on blueshift. However, at present we aim to be a tradeable platform, rather than a pure-play research platform like quantopian. We have noted strong demand for ETFs from our users. So ETFs will be our next pick. Putting in russell 2000 simply as is may not be a good idea as some of them may have some liquidity and quotes issues. Better to cast the stock universe through appropriate filter to create a universe which is amenable to quant approach. We are ok to put S&P 500 as is (i.e. without any filtering) as this cohort is expected to pass most basic tests (e.g. liquidity filter and quote availability) as is.
Prodipta, I would agree with Stephan, particularly as more and more come over from Quantopian. S&P500 is giving me much different results than QTradeable and even Q1500US. Hopefully your data related activities will get Blueshift into bigger equity universe. Our fingers are crossed. Blueshift is in such a great spot to take advantage of Quantopian's decisions to stop live, paper, and contests. And if Blueshift could get pipeline added for "Go Live" that would be fantastic…(at least your backtesting allows for pipelines via zipline). Cheers…
Absolutely. We would love to provide what the quant community longs for on blueshift. Obviously we have our roadmaps. But hearing from our users is most important. And I really appreciate these inputs from you and Stephan. We will, nevertheless, be cautious in dipping our toes in to lower caps worlds - as the last thing we want to provide data which are harmful! Small cap dataset will require quite a bit of scrubbing and processing, so I think we will have to wait till we get the team size large enough to do justice to it. At present we are a tiny team who code frantically all day long for feature developments! Pipeline on live, btw, is expected to come before russell 2000!
Great news for "Pipeline on live"…can't wait! I'd rather have that feature sooner rather than later. Thank you for listening.
thanks and noted! I'll make a point of testing s&p on Q for now then, and compare to Blueshift. I'm interested in comparing backtest results for now, and I'm curious to know if one is more 'optimistic' than the other…
That will be very interesting to know indeed!
I compared Blueshift result to Quantopian screening q's pipe with MarketCap.Top(500) etc…
Q's results are promising but Blueshift looks very pessimistic. I'll try to post screenshots.
At first I thought it must be QTradeable universe, and that must be the magic sauce (assuming backtesters were similar). But after this, I'd have to say your backtester is less forgiving? Your backtestes is way faster too.
Very interesting, screenshots will be insightful. If your strategy has a momentum or long bias, I would expect that btw. S&P is a different beast, given TSLA and SNAP and others! Backtests speed were comparable, but no faster I think after quantopian revamped their backtest display - I think a year back or so. Our quick backtest is streams minimal data for speed.
____________________
bummer, I was gonna post the blueshift code, but it exceeds limit. it's only 140 lines…
oh well, here is the Q post where I share the Q algorithm.
Porting it over to Blueshift was minimal.
https://www.quantopian.com/posts/dailyweeklymonthlyquarterlyyearly
Hi Stephan,
I'm interested in learning more on how you ported a Quantopian algo to Blueshift. I was wondering if you could provide some more insight on how you went about this?
Ross,
goto the quantopian link I posted above, and copy the algo I posted. there was minimal rework to get it running in blueshift. If you want the Blueshift code, tell me where to send it and I'd be happy to. Compare for yourself. I can't post the full script here, the chat has a line limit.