Why don’t we see algorithmic systems that use volume? It seems that all of them use price, sometimes as spreads as in pairs trading, sometimes the term structure of futures, often Eurodollar rates or crude oil. But rarely does volume enter into the decision except when selecting the most liquid equities or futures contacts. Yet it offers valuable information on how the buyers and sellers are acting, the potential for diversification, and at the minimum, another piece of the puzzle.
Classic chart analysis says that increasing volume confirms the trend, while rising prices on declining volume is an opportunity to sell. That makes sense, but day-to-day volume is very erratic. Averaging it helps, but then it introduces lag, which makes it less timely. Volume in most markets declines in the summer, when many investors take off. It also varies during the day, highest on the open and close, lowest around midday when traders break for lunch. That can result in unreliable trading signals.
You get the point. Volume isn’t easy to use. But there are cases where we all seem to agree on its importance – when it spikes way above the normal level of trading. It consistently indicates a change of direction — trader exhaustion.
Basic Trading Rules
To implement this idea, we need the following:
A measure of extreme volume
- A way to decide the current direction of the price move
- A profit-taking rule because we don’t expect a reversal to be sustained
- The maximum number of days we’ll hold the trade
- A large set of markets to test the strategy
Let’s discuss each briefly.
Deciding When Volume is Extreme
Using equities, we will average the volume over a reasonably long period, 60 trading days (about 3 months), then find the spike ratio of today’s volume to the average volume.
Sometimes a volume spike is followed by another volume spike. If we use the 60-day average as of yesterday, then a number of spikes in a row will cause the average to increase sharply and we may not get a second volume spike signal. To avoid that, we lag the average volume, so today’s ratio is today’s volume divided by the average 30-days ago. You could also take the average over more days, but lagging it still helps.
How large should the spike be to generate a trade? Because the equities market is biased to the upside, the thresholds for buying and selling won’t be symmetric. We’ll buy when the price direction is down and the spike ratio is 1.5, 50% larger than normal. We’ll be more demanding for the short sale and require the spike ratio to be 2.0, twice as large. There’s a trade-off between lower ratios that produce less reliable signals and higher ratios that don’t generate enough signals.
Maximum Holding Time
We don’t think reversals indicated by volume spikes have a long memory, so we arbitrarily set the maximum holding time at 5 days. That gives us a day or two after entry for the price to move the wrong way, but if there is no recovery within 5 days, we’re out.
There is not going to be a stop-loss because this is mean reversion trade, that is, we are trading against the direction of the price move, so we fully expect to have a loss for the first day. If we wait for a reversal before entering, we can miss the biggest part of the profit. You may know that the profile of this type of trading is a lot of smaller profits with an occasional large loss.
We may not have a stop-loss, but we can capture profits and get out early. Again, we’ll want that to be asymmetric. It’s interesting that price moves in equities are very different for longs and short sales. Prices move higher slowly but fall quickly. So our profit-taking is only 75 basis points for long positions and 100 basis points (1%) for shorts.
We’re not looking for a long-term trend or an investment. If we can identify a spike based on a recent move, we can generate a lot of trades. We’ll chosen to use a 5-day moving average to determine the direction of prices. If today’s moving average value is greater than the previous day, the trend is up; if it’s lower the trend is down. There are many ways to determine the trend; this is one of the simplest.
Whenever we get a trading signal we enter on the next open. In some ways that reduces the loss if prices continue to move against us. It’s also practical because you may not have good volume numbers until after the close. For the exit, we either get out on profit-taking or day 5. Because we know it’s going to be day 5, we can get out on the close of the last day.
Putting It All Together
There is nothing complicated about any of this, nor does it use sophisticated math, although I’m sure that a fancy formula could have been impressive. You can probably get the signal off a chart that shows volume along the bottom. The only real problem is tracking enough markets to get diversification. In this example, we’ll test a broad set of 63 liquid ETFs:
- 4 U.S. major market equity ETFs
- 25 U.S. sector ETFs, including most SPDRs and many Vanguard ETFs
- 17 Country equity ETFs
- 10 assorted ETFs including bonds, precious metals, emerging markets
- 7 inverse ETFs, double and triple leverage
Here’s the interesting part. When we just list all the trades in the order they occur, without regard to how much money we have to invest, but with each trade assuming a $10,000 exposure, we get the results in Chart 1. Given that the 16 years of data covers some very extreme conditions, the results are good.
Chart 1. Cumulative PL of all trades based on extreme volume
The results show volatility in 2008 but it’s possible to reduce that using a filter based on annualized volatility of price, a standard way to control risk. Without that, the drawdown is not nearly as large as the one we experienced in the stock market and the recovery here is much faster.
Choosing ETFs for a Realistic Portfolio
We can’t trade all the signals, so some selection is necessary. It turns out that 66% of all trades are profitable, but selecting which ones will be the winners in advance is not easy. Normally, an equity portfolio is selected by the history of individual stock success, but here the trading signals may be few and far between, making that approach impractical.
Because of the high rate of success, we’ll construct a portfolio that takes only 6 trading signals at one time. An investment of $10,000 in each trade comes to a portfolio of $60,000. If we use less than 6 signals, we significantly reduce our diversification. If we want more than 6 signals we will find that our capital is sitting idle for most of the time. Of course, you might find that 5 or 7 works better for you.
If there are less than 6 active trades at one time, we take them all. We hold each trade for 5 days or until it reaches the profit target. If there are more than 6 trades we choose the ones we want randomly. That’s not choosing the ones starting with “A” one day, “B” the next, but we actually used a random number generator scaled to the number of new trading signals that day, and select one or more trades to add to the portfolio. The result is shown in Chart 2.
Chart 2. ETF portfolio based on volume, chosen randomly.
I need to emphasize how important it is to have trading strategies based on different concepts. It provides more diversification than trading different markets, which can all move together in a crisis.
You also need to implement and test this yourself. Although we’re sure this is correct for what it is, it may be missing some important risk controls, such as deleveraging during very volatile markets when many stocks and ETFs could have volume surges and all react the same way. You’ll also want to select your own set of ETFs and stocks. It’s much easier to follow a strategy if you’ve verified it yourself and this one is worth the trouble.
(c) Copyright 2015, KaufmanSignals.com. All rights reserved.