Jul

23

For as long as I can remember I have spoken negatively about the use of volume data as input for trading. I never stated that volume was worthless, only that I could not discover any value added by using volume. I would now like to say mea culpa and illuminate the good and the bad.

In the past I had always researched volume data as something that should be multiplied against either price levels or changes. The logic is that a price move when no one placed bets isn't much of a price move, and price moves with lots of money being wagered are more significant. Several of the notables on this list have published books and my recollection is that they counsel in favor of weighting price action by volume. And although I could use volume data to generate profits, no prior efforts were as profitable as using price data without volume.

That is, trading markets according to price momentum can be profitable, but my efforts at weighting that momentum by volume brought down the profitability. Similar with price volatility; weighting it by volume only dropped my expected trading profitability. Hence my bias against volume.

So why would I keep looking? Well, philosophically speaking, volume data is important simply because it isn't price. That is, it offers a respite from possible multicolinearity of looking at price data over and over. However the idea that volume alone could be interesting just does not seem logical. There is a lot of mythology associated with the markets. For example the oft stated reason for the market going up "more buyers than sellers" is impossible, as there are never more buyers than sellers. I had come to believe that the recommended use of volume was such a myth.

A little while ago a fellow Speclist poster (for whom I have considerable respect) commented about certain data (not volume) that gave me the idea to see if volume could be used without weighting it against price. That is, could it be used alone, or relatively alone? "Relatively alone" means comparing volume on up days to volume on down days, or volume on up ticks to that on down ticks, but still not being weighted by price activity other than the sign (+/-). Despite years of research, I had not done this before.

When you look at volume data it just looks like static, whether it is just total volume or distinguished by sign. You have to smooth it. For most people that means moving averages. First tip: don't waste your time with them. They are great at smoothing data when that data has a periodicity that is known and predictable, which does not apply to volume. So instead of using moving averages, go a notch or two higher: exponentials or better yet, regression trends. And once you have that, look at the slopes.

Second tip: looking at equities, your lookback periods should be relatively long, by which I mean 6 months or longer. If you have a method of dynamically determining the best lookback period (without using look-ahead bias) then use it, as its profitability will exceed most others.

 Third tip: it appears best used as a discrimination tool (buy the stock vs. sell the stock) rather than as a ranking tool. However we have not yet exhausted all the possibilities of the latter. Timing decisions are less critical than using price momentum, probably because there are many followers of price momentum and the exits and entrances get too crowded at certain times.

Results: trading constituent stocks of the Russell 3000 on the basis of volume over the last 15 years has produced profitability comparable to that of using price momentum. Similar rates of return with better (i.e. less) drawdowns. That's about 8,000 stocks when you include the dead ones to eliminate survivor bias. Results are better with items that go into portfolios. That means individual stocks rather than indices. Unlevered ETFs of equities = yes; levered ETFs = no; ETFs of currencies and volatility = no. Volatility of the asset negatively impacts success, just as with momentum. We have not tried futures (as we do not trade them).

We are now testing the incorporation of volume data into our overall decision process. Will advise how it turns out.

Lesson learned: when something is making a move, the signals tend to be writ large across the landscape. Accordingly it is no surprise that volume data works. The problem was in my preliminary bias against it and the thought (and published material) that it had to be weighted against price. Resolved: just do the research and forget the preliminary bias.

Third tip: it appears best used as a discrimination tool (buy the stock vs. sell the stock) rather than as a ranking tool. However we have not yet exhausted all the possibilities of the latter. Timing decisions are less critical than using price momentum, probably because there are many followers of price momentum and the exits and entrances get too crowded at certain times.

Results: trading constituent stocks of the Russell 3000 on the basis of volume over the last 15 years has produced profitability comparable to that of using price momentum. Similar rates of return with better (i.e. less) drawdowns. That's about 8,000 stocks when you include the dead ones to eliminate survivor bias. Results are better with items that go into portfolios. That means individual stocks rather than indices. Unlevered ETFs of equities = yes; levered ETFs = no; ETFs of currencies and volatility = no. Volatility of the asset negatively impacts success, just as with momentum. We have not tried futures (as we do not trade them).

We are now testing the incorporation of volume data into our overall decision process. Will advise how it turns out.

Lesson learned: when something is making a move, the signals tend to be writ large across the landscape. Accordingly it is no surprise that volume data works. The problem was in my preliminary bias against it and the thought (and published material) that it had to be weighted against price. Resolved: just do the research and forget the preliminary bias.


Comments

WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '8521' AND comment_approved = '1' ORDER BY comment_date

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search