Incorporating time series smoothing

My value strategy focuses on long positions by attempting to optimize for signal to noise ratio and identifying the types of response curves seen in both past and current data.

The basic ideas stem from my experience with multivariate time series analysis from other disciplines. I apologize if I don't use the correct terminology. I may also have what seem like obvious omissions due to not knowing what the metrics for certain stats are called here, so I definitely welcome any corrections!

I have so far used a qualitative synthesis approach to evaluate trends with this order:

1) Basic visual inspection of the trading price. I personally prefer Heikin Ashi over standard candle sticks but I don't think it is a significant enough improvement to warrant switching over if you prefer candle sticks. I would love Moomoo to add violin plots as an option so you can see both the price range as well as the difference in trade volume across the range. (Not something I have seen finance use; probably will write my own dashboard eventually)

2) I largely ignore the moving averages and prefer exponential moving averages. I like to keep the MA values there to be able to see how far it diverges from the EMA (but can't figure out how to have the text there but remove the lines 😅). The main advantage of EMA is that it gives less weight to data points the older they are (relative to the set duration) so infrequent or one-off noise events (e.g. 0% interest rates) don't impact the current value as much.

3) I have found a couple of secondary metrics in the app that I find helpful. EMV/EMVA helps by getting to that question of incorporating both trading volume as well as price. I quickly learned that price itself doesn't matter much if you don't account for what's going to happen with volume (e.g. short sellers, trying to liquidate when there are no buyers, etc). I use it as a very rough equivalent to eigenvalue analysis, so not exactly the intention. My general rule of thumb:

- if the value of either are approaching 0 from a negative value it indicates positive trend.

- if the value of either are approaching 0 from a positive value, it indicates negative trend.

DMA/DDD/DDDMA I use roughly as an equivalent to derivative (as in calculus 😅) analysis of the change over time. The interpretation changes based on the intervals you select. With the default settings, I evaluate it as if the DDD line is above the DDDMA line - that indicates a positive trend. And if DDDMA is above DDD - that points to a negative trend.

AR is the latest matric I have started to use and so least familiar with the general content. I use it as a proxy for a 2nd derivative analysis to get an idea about the volatility (e.g. the change in the change in value over time). I know it doesn't exactly fit well into the framework, so happy to have suggestions for that! Mathematically, I don't see the BR offering a better result for extra steps and I operate with a K.I.S.S. mindset to avoid over fitting. Roughly, when the AR wobbles around 80-120 or so I take that to mean the price of the stock isn't going to change drastically. Theoretically, an AR > 150 can indicate a pending sell-off and an AR < 80 can indicate large buying volume but I haven't seen it empirically yet to attest to the reality of it.

I included the following data so you can decide for yourselves how well my qualitative interpretations fit the reality.

I know long trends aren't as fun as short swings, but hopefully the perspective of an outsider recently joining in the game can be helpful!