Most financial data is either sold as Aspiration or Justification. You frequently see the words "industry standard," or "professional-grade." The implication is that all the real professionals have this data, and if you don't, then you're not a real professional. And whether for the sake of your self-image (aspiration), or for the sake of your clients (justification), you don't want to be considered "an amateur."

This is a very powerful sales pitch. It works on high-AUM money managers just as well as it works on retail punters. Yet the data itself hardly ever gets used to do anything except confirm the user's priors, or to justify a decision that was going to be made regardless. The data enables the user to say, "I am a professional, I have the data, and I did my due diligence."

And so, to be quite clear, that's not what this stuff is. All the numbers and charts on this website are for amateurs. Rather than satisfying Aspiration, or providing Justification, this data is meant to actually give you Perspective. It will disagree with you more often than it agrees with you, and it will probably annoy you sometimes, but if you allow it, it will provide an entirely new angle from which to view the way a stock trades.

In other words, it does what data is supposed to do.


And so, with that disclaimer, let's give you a sense of what we mean. This is what our data looks like:

These are "weather maps." They describe how each of the coordinate pairs of our data have historically influenced subsequent 1-week returns in any given security, and then tells us (by drawing a yellow circle) what the coordinates are right now. The above maps are for Apple (AAPL).

At a glance, you can see that positive (blue) and negative (red) returns tend to cluster. For example, in the price, volatility relationship (top-left panel), flat price movement (x=0.0) and rising volatility (y=0.5) predict poor 1-week returns going forward (red blob).

Conversely, in the gamma, dark relationship (bottom-right panel), we don't see reliable predictions for negative returns, but we do see extremely positive returns (blue blob) in the top/top-right, which indicates high call gamma combined with high dark pool short volume is exceptionally bullish.

Viewing the data pair-wise and in this weather map format is essential to understanding how a security is impacted by a combination of factors. For example, if you'd run a linear regression, or even a fancy quantile regression, on the relationship between AAPL's price path and subsequent returns, you would never find a relationship between flat price and subsequent returns. But if you were able to add the "volatility" dimension, you would see very clearly that volatility is the determining factor in whether recent 0% returns are bullish or bearish going forward.

If price and returns are length and height, then volatility is depth.


This multidimensional perspective simply cannot be reflected in any two-dimensional way. That's a big deal. And to the extent that we think this seems a valuable representation of financial data, common statistical methods are obviously the wrong tool for the job—because they completely lack "depth perception." Consider another similar example:

These are the weather maps for the SPY ETF. Look at the volatility, dark map (middle one on the bottom). Observe how the upper third of the map has a bold blotch of both blue and red, one to the left, and one to the right.

This means that when dark pool short volume is high, this can either be really bullish, or really bearish—depending entirely on whether volatility has been rising or falling recently. This is interesting, because it means that if you were to analyze the dark pool data alone, you would find very little predictive information; but by simply adding the "volatility" dimension and visualizing it this way, you suddenly see two distinct, and powerful, historical signals in the SPY ETF. Depth!

Nor is it difficult to understand why rising and falling volatility would impact the meaning of the other predictors in the case of both AAPL and SPY. The motivation of market participants, and the mechanical reaction of the market, depends greatly on the trend of volatility. None of these data points exist in isolation, so... let's not treat them as if they do. Perspective!

Robot Jim

Last step: Let's feed all of these coordinates into a machine (affectionately named Jim) and derive the full historical distribution of returns implied by the data. Since the machine can "see" in multiple dimensions, it can search for only those historical returns that are nearest, in every dimension (including time), to the current data coordinates.

The result is a distribution of returns that takes into account not only each data point, but also all of their complex interrelationships.

And from this distribution of returns, we can derive a mean and median expected return, as well as a volatility forecast. Very tidy.

Pack all this up into a few non-industry-standard charts, PDFs, and spreadsheets, and that's what you're buying here.

Tell me more...

What's it cost?