Dynamic-DataSet-for-stock-scanning-extendable-for-custom-criteria

Hi,

Sorry, it's impossible. We're linking up to a fixed list of criteria already available at the source which we can choose from. Moreover, a decent replacement would be hard to find should the current source go down.

Ultimately I want to build a new stock screener that will be powered by our Building Block Conditions. Did it previously for Quantacula so the Proof of Concept is there. Would be very nice to scan the whole US market and return the stocks that match your specific Building Block criteria!

QUOTE:
return the stocks that match your specific ... criteria!

The problem with all stock screeners is setting up the criteria so it picks stocks that perform well (i.e. positive sloping equity curve) with my two strategies. Just because a stock follows a particular criteria doesn't mean it will trade well with a given strategy. What kind of tool can you (or me) develop to bridge this functional gap? We need a strategy-->criteria generator tool.

Also, I usually start with a strategy then try to devise a selection criteria for it. Is that backwards thinking? (Or am I asking the wrong question and need to approach this problem from a different angle? Maybe a big-data angle.)

Maybe we can create an Indicator that takes a Strategy as an input, runs the Strategy on the source BarHistory, and then the result of the equity curve. Now, couple that with a new Transformer Indicator, say IndOnInd, that lets you apply one Indicator onto another, so you can arive at the slope, ROC, etc, of the equity curve, and you'd have a way to use the Condition Screener to screen for stocks that do well with your 2 Strategies.

QUOTE:
create an Indicator that takes a Strategy as an input, runs the Strategy on the source BarHistory, and then the result of the equity curve.

And that's the other problem. Most of the ScoreCard metrics are not a function of time; e.g. Sharpe Ratio (which is a Z-score) is not time dependent. And a ScoreCard merit metric has to measure a study, positive-going slope with time on the equity curve to be a "bonafide" merit metric for matching stocks to strategies.

And then how should you fit the positive-going slope on the equity curve? Is it a time series fit or a regression fit? I would say it's some of both (but we could argue that). If it's both, then it has to be fit as a mixed model. Should one fit the coefficients of this mixed model simultaneously or sequentially? That's an interesting question....

So you kind of open a Pandora's box of multivariant statistics when you go down this path. Maybe that's why nobody has addressed this problem rigorously.

But I agree. We need to "attempt" something even if it raises implementation questions for the next 15-20 years. At least it will get the literature talking.

Just completed one piece of the puzzle for WL8, a new Transformer Indicator IndOnInd that lets you apply one indicator onto another. An RSI of an SMA, or a BBand of a CMO, for example. This opens up even more flexibility for Building Block Strategies and the eventual Stock Screener.

Which providers does the current Dynamic DataSet for stock scanning use?

It does not use providers i.e. it does not use the data you may have downloaded in your WL7. It pulls the results directly from a web feed, operating on the entire US stock universe (or just on exchanges you select). The current name of the feed can be found in the built-in Help.