In a retro flashback to our earlier “zone” concept, it is important to understand how volume interacts with different basic market conditions. This is important because it shows you where you are likely to make errors in judgement if you were to simply use one variable on its own such as looking at today’s direction. This situational or conditional probability approach is used successfully by Rob Hanna of Quantifiable Edges http://quantifiableedges.blogspot.com/, and it is quite useful towards increasing the prediction accuracy of a generic condition such as daily follow-through. Another good blog which covers this approach for next day prediction of the S&P500 is run by Frank of Trading the Odds http://www.tradingtheodds.com/.
I would suggest that if you want to do this type of analysis you follow a similar format to what is presented below when doing this type of analysis and keep things simple. Both Rob and Frank are more experienced at combining multiple conditions and it is well-advised that you read through their archives and examples very carefully before attempting to do this on your own. That said, as a general piece of advice, the un-initiated are advised to try to minimize the number of subconditions and maximize the number of final observations in such a way that there is a noticeable difference across each box. In this case, there are well over 150 observations for each condition in the table below. All of the trend and trendless conditions are divided up using DVRAC- a DV Indicator that is an intermediate-term statistical trend filter. https://cssanalytics.wordpress.com/2009/10/09/dvrac-and-trading-in-the-zone/ The concept behind DVRAC is the use of multiple r-squared calculations using a regression, and values greater than 20 or less than -20 indicate a statistically significant trend. All values in between are considered statistically insignificant and therefore no strong trend currently exists in either direction. The table below investigates how various market conditions defined by DVRAC perform in conjunction with volume and up/down days. To represent our volume screen, we will use volume ROC(5) as in our previous posts. As a refresher, rising volume using this measure was positive, while falling volume was negative. To preserve space, and to save time I only included some simple statistics. Here are the results over the last 2000 bars on the SPY:
|avg daily return||w%||worst day|
|Trend UP||Today UP||Vol ROC(5) UP||0.07%||57.87%||-4.20%|
|Trend UP||Today DOWN||Vol ROC(5) UP||0.11%||62.34%||-3.90%|
|Trend UP||Today UP||Vol ROC(5) DOWN||0.04%||55.15%||-3.00%|
|Trend UP||Today DOWN||Vol ROC(5) DOWN||-0.01%||49.43%||-3.80%|
|No Trend||Today UP||Vol ROC(5) UP||-0.18%||50.91%||-6.41%|
|No Trend||Today DOWN||Vol ROC(5) UP||0.16%||57.26%||-4.40%|
|No Trend||Today UP||Vol ROC(5) DOWN||-0.19%||48.12%||-3.33%|
|No Trend||Today DOWN||Vol ROC(5) DOWN||0.23%||60.20%||-7.42%|
|Trend DOWN||Today UP||Vol ROC(5) UP||-0.26%||46.34%||-7.84%|
|Trend DOWN||Today DOWN||Vol ROC(5) UP||0.33%||56.43%||-6.99%|
|Trend DOWN||Today UP||Vol ROC(5) DOWN||-0.08%||52.30%||-8.85%|
|Trend DOWN||Today DOWN||Vol ROC(5) DOWN||-0.10%||52.63%||-9.85%|
As you can see, a strong up-trend with strong volume is positive regardless of whether today is an up or down day. This result is very important because it shows that you should not consider standard mean-reversion relevant when this condition is present–you are better off sticking with the trend, and in general it is slightly more profitable to buy on down days vs up days. When the trend was up but volume was falling, up days were more profitable than down days, and interestingly enough down days showed a slight negative average daily return. It appears that the market will move in the direction of today’s close–that is opposite to standard mean-reversion when volume is falling. This is another interesting observation, and also very consistent with the realities of today’s market. Broadly speaking, when looking across sub-conditions when the trend was up, mean-reversion using daily-follow through was not really a useful strategy. Furthermore, the greatest penalty using a mean-reversion strategy would have gone to the short side–ie shorting after up days which was dangerous across the board. Another final observation was that tail-risk, or the worst daily loss, was very low in this zone. Broadly speaking, traders are advised to go long or buy on pullbacks in this zone, and look to lighten up positions as the market weakens on falling volume and breaks key support.
Moving on to the trendless zone, we see in contrast that mean-reversion was in fact a highly profitable strategy regardless of volume condition. This makes sense, as the probability distribution of sideways movement should contain no particular trend bias, and tends to bounce off of support and resistance. There was a slight, though likely insignificant advantage towards fading on falling volume conditions vs rising volume conditions. This was particularly true when buying on down days– as they say “never short a dull market.” A more complex analysis may reveal some interesting setups based on volume reversals. Generally, tail risk is low making this a good area to execute mean-reversion. Trades can be exited in this area if the regime were to change, or support/resistance is broken.
Moving on to the down- trend zone, we see a very interesting result: mean-reversion strategies perform best when volume is rising. Apparently this pattern denotes capitulation or exhaustion– up days on rising volume were followed by strong downside reversals and down days on rising volume were followed by strong upside reversals. In contrast, when volume was falling, the downward force of the trend was far more prevalent. Both up and down days led to negative returns when those two sub-conditions were present. Tail risk in this situation was the highest of any other setup. The bottom line is for traders–look to fade high/rising volume conditions in a downtrend, and go short or to cash when volume is falling.