Summarize VPIN

[rcyeh]

Start with the Wikipedia article on VPIN, and also check out the first reference in vpin/README.txt. The other articles may offer helpful supplementary data or alternative perspectives.

[rcyeh]

Is Friday, 7/26 a reasonable deadline for an initial read of these articles and references? One person, assign yourself to this issue, and then write a summary. Everyone else who has time, please also comment on the summary if you disagree or want to expand.

Next week, we will (#3) plan the steps we shall take to evaluate the paper.

[zc238]

7/26 should be all right.

[jf434]

Completed Summary under vpin/VPIN Summary

[rcyeh]

Great. Now that Jiongjia has summarized the paper, can we have another volunteer read the Andersen and Bondarenko paper (at the end of vpin/README.txt)?

[zc238]

I just read this paper this afternoon, I'll provide a summary shortly. Do you want the rest of us to each assign themselves one remaining paper?

[zc238]

By the way for the VPIN summary, I think it is very well summarized. Two things I want to clarify: 1) informed trader is a very general term here, and does not necessarily imply insider information. 2) the mechanism behind volume bucketing may be worthy of more notes. So to expand a little bit, within each volume bucket, a single unit of interest actually combines all tick information within a 1 minute interval. And when calculating order imbalance (OI), the weights on these units depend on the number of ticks happening in between the 1 minute interval. One motivation behind aggregation pointed out by the author is to prevent the noise caused by splitting orders, which is the norm in HF world.

[rcyeh]

The goal of this step is to know enough about the technique that we can make a realistic plan in #3. In particular, I think the method used to classify buys and sells, and group trades into volume buckets, is important to our evaluation. My impression is that because there is no precise specification, the group in the third paper (who I would expect to be friends of author de Prado) ran many different variations.

[zc238]

Anderson paper summary submitted.

[rcyeh]

I can see that the summary is detailed to the level that I could outline a list of tasks I would want to do to replicate the results. That's good. Let's proceed: For issue #3, let's create a plan to see whether VPIN is useful and what we can do with it, using US equity data.

When I write these plans, I structure them as action items or questions. For example:

Goal: Evaluate VPIN

1. Select evaluation criteria a. Does [a VPIN-like indicator] predict future prices? Over what time frame? b. Does [a VPIN-like indicator] predict future volatility? Over what time frame?
2. Prepare data a. Identify and learn about tick data files. b. Produce a time series of observed returns or volatilities. c. Assign Buy/Sell label to each tick or bin. d. Aggregate and fit. ....

Would you like me to make a first template for this outline, or does one of you want to do that?

[zc238]

Sorry about the delayed response. We shall be able to do it.