Stock screeners

AI & Machine learning
1

Hi, there:

                   There are many scanners/screeners out there (FINVIZ, Investing.com screener, Stockfetcher and others)  But not of them seem to contain some variables relevant to quant strategies.



Does anyone here know about screeners that actually contain quantitative variables as input criteria for scan stocks??  for example,  Is there any screener out there that can be used for Hurst Exponent of the last 9 days close to one (for momentum strategies) or close to 0 (for mean reversion strategies)  ??



What about Augmented Dickey Fuller test for mean reverting  strategies??   is there any screener out there that have this varaibles as inputs??



Thanks

2

I doubt you will find a readymade screener available with the inputs you have in mind. I would say most screener apps will tend to cover the common aspects that most people are looking for. A highly customized screener cannot not work with pre-computed data and requires much elaborate system, like blueshift. On blueshift, you can use the pipeline APIs to screen assets based on any function you want. For more on pipeline, look here. Here is an example where it is used as a screener. The example uses some technical indicator functions. Replace them with your own.

3

Hi Prodipta, wanted to check with you if a slightly more customized screener (such as the one described below) is implementable in blueshift:-


  1. read 1 minute live bars from specified broker (eg: IB) on nifty500 stocks simultaneously
  2. filter population for those stocks which have:-

        a. gapped up or down on market open by more than 5%

        b. crossed 20 EMA anytime over the last 10 mins

        c. marketcap of atleast 5000cr



        i.e. ideally a combination of technical and fundamental indicators. 


  3. update the filtered population list every 1 min( i.e. each time updated market data is received)
4

The answer is kind of yes and no. Blueshift does not have fundamental data yet, so filtering on market cap is not possible now. But this is not the real problem here - we will add fundamental data soon. The rest of the processing like EMAs and gaps are straight forward as well. The real issue here is the data part.



Tracking a large universe in real-time is quite demanding. Most brokers and data sources will have severe restrictions on the max number of tickers which you can subscribe to for realtime data simultaneously. That ranges from 30 to sometimes 200. But in any case this number is usually far from the universe of tradeable assets or even benchmarks like NIFTY 500. If you have a data source that supports it, blueshift as such do not have any restrictions. But for all practical purposes, it is not possible.



Tracking a large universe in real time is a highly unsual workflow (and very demanding in terms of computational resources). The standard workflow is to track a large universe on a lower frequency (e.g. daily) and screen it to a smaller set for real-time (say 30 or even 100). That is why the pipeline APIs on blueshift are implemented on daily frequency.



Some sites and apps do offer realtime screeners (although as mentioned, it is not a standard quant workflow). Technically, if one is offering a few metrics of pre-computed values, the data problem can be addressed by dividing subscriptions to n different data sources. Here the computation is done by the app and hence this is possible. But if I am offering fully customiatizable and programmable screeners, you either need an unique data source that can conver the universe or do the subscription multiplexing yourself. Here the computation is done by the user, and hence is challenging.