Similar to that post here, Im analyzing the impact of the NET volume traded on the NYSE and how it is impacting day trading models. The Net volume on the NYSE is essentially all the USD Volume that is flowing into stocks that trade higher minus all the USD Volume of stocks that are trading lower.
I am using the same dummy day trading strategy as previously as a benchmark. The rules again are:
If the Low is lower than the lowest low of the last 5 bars then BUY SPY
Profit Target 100$
Stop Loss 100$
Maximum time in trade = 2 hours
Close all open trades at the end of the regular session
Symbol: SPY on 5-min
Symbol 2: NYSE NET Volume (Symbol at IQFeed: DINN.Z)
Period: 01.08.2015 to 01.08.2017 leaving the rest of data for Out-of-Sample testing
Amount per trade: 100’000$
No commission or Slippage
Max Bars back: 200
In case you are wondering why the benchmark results are not exactly the same, the reason is I have a different max bars back setting.
The most important performance figures to me are Profit factor, ROA and Avg. Trade.
Profit factor: 1.07
Avg. Trade: 3.45$
1.) Buy SPY on a pullback (Benchmark Strategy) when the NYSE Net Volume makes a new x bar high (Switch in last column below = 1) or low (Switch = 2)
Variation 17 is the benchmark strategy without the above filter. What can be immediately seen is that it is much better to buy the SPY when the NYSE net volume is making a higher high (Switch = 1 means buying on a higher high). Look on the last column to the right and see how many variations are above the benchmark strategy (I sorted it via the Profit Factor) with only 4 variations below it. It is also clear that you should not buy the SPY when the NYSE Net Volume is making a lower low. Basically it means it is much better to go long when there is more volume flowing into up stocks than down stocks and this makes intuitively sense.
2.) Buy SPY when the absolute NYSE Net Volume is above (Switch = 5) or below (Switch=6) x millions
Row 16 shows the performance when the NYSE Net Volume is > 0. Profit Factor, ROA and Avg. trade are all already higher then the benchmark strategy. What is also totally obvious is the fact that every single variation where NYSE Net Volume is below 0 is worse than the benchmark strategy and once Net volume gets below 7 million the performance is even negative. I really like the linear relationship showing the more negative the net volume the worse the performance. This could be a real edge but needs to be validated on out-of-sample data (see below).
3.) Buy SPY when the NYSE Net Volume makes a new high
I actually expected this to be more profitable than then benchmark too but its actually not the case. There had been 452 Trades with a total net Profit of only 705$. Profit Factor was 1.03, Avg. Trade $1.56 and ROA 34%. Pretty surprising this underperformance.
4.) Buy SPY when the NYSE Net Volume is above (Switch = 11) or below (Switch = 12) an x period bollinger band
Although it shows a similar behaviour that you are better of buying when the NYSE Net Volume is above its x period Bollinger band, only very few variations show an outperformance. Parameter 1 means the 5 period BB, 2=10 period, 3=15 etc.
Lets check the 4 models above on out-of-sample data from 01.08.2017 to 01.02.2019.
The benchmark strategy performance for the out-of-sample period:
Net Profit: 7471$
Profit Factor 1.04
Avg. Trade $1.94
1.) Out-of_sample: Buy SPY on a pullback (Benchmark Strategy) when the NYSE Net Volume makes a new x bar high (Switch in last column below = 1) or low (Switch = 2)
Variation 15 shows the benchmark strategy. Unfortunately it shows completely the opposite behaviour since now all variations perform better when the NYSE Net Volume makes new lows. This shows how important it is to do solid testing and not fool yourself. When I am developing fully automated models I do a lot more stringent testing but this would already be a model I wouldnt trade since it clearly failed out of sample.
2.) Out-of-Sample: Buy SPY when the absolute NYSE Net Volume is above (Switch = 5) or below (Switch=6) x millions
The best pattern from the in-sample test also help up out-of sample. Although it has less variations that outperform the benchmark, it certainly seems important to monitor the absolute value of the NYSE Net Volume. If it is above 15 million then it shows a very high edge although this only occured 28 times together with the Spy pullback “dummy pattern”. What I really like is the linerarity of this pattern. Net volume of 15 million is better than 14 million is better than 13 million and so on. All down to 7 million seems a good edge/filter for long trades.
3.) Out of sample: Buy SPY when the NYSE Net Volume makes a new high
The results were mixed and not really better. Same as in-sample.
4.) Out of sample: Buy SPY when the NYSE Net Volume is above (Switch = 11) or below (Switch = 12) an x period bollinger band
90% of all variations show an outperformance vs the Benchmark strategy with higher profit factors and higher avg. Trade. However there is no linearity and i prefer if the opposite signal is clearly negtive. Mixed bag here, if you ask me.
Although I am a bit surprised that Pattern 1 completely reversed, we still found something that can be of incredible use. Monitoring pattern 2 and where the NYSE Absolute volume trades seems important. When NYSE Net Volume gets above 7-10 million this is very supportive of bullish strategies and should be monitored.