6-month ahead, a medium horizon: US Equity
We recently added 6-month ahead, a medium horizon. If one searches for the three US stock index ETFs “QQQ.US;SPY.US;DIA.US”,
the common causally predictive factors for their values 6-months out, as of Mon 7 Aug are computed to be
Altria Group MO (consumer, tobacco, linked to inflation)
Bristol-Myers Squibb Co. BMY (pharmaceutical, linked to inflation)
Nvidia NVDA (AI, semiconductor)
SPDR S&P 500 ETF SPY (the SPY itself, therefore SPY’s own trend)
International Business Machines Corp. IBM (technology, AI)
The net majority of predictions, from this horizon, for the coming day Tue 8 Aug are
QQQ 54% of models predict up (57% predict up, 3% predict down)
SPY 72% of models predict up (78% predict up, 6% predict down)
DIA 97% of models predict up (98% predict up, 1% predict down)
Understandable, in the high interest rate environment, stocks which attracted high enthusiasm usually are more volatile. To put these net predictions in one perspective, over the past decade, the returns were
QQQ 5x
SPY 2.7x
DIA 2.3x
So if the current environment persists, and one long hypothetical 54% max shares of QQQ, 72% max shares of SPY, 97% max shares of DIA, the returns, using historical figures (just for the current exercise), would be more in line with each other,
QQQ 5 x 0.54 = 2.7x
SPY 2.7 x 0.72 = 1.9x
DIA 2.3 x 0.97 = 2.23x
As another example, if one searches for “Apple; Amazon; Nvidia” or “AAPL; AMZN; NVDA”,
the computed common causally predictive factors included technology, defence, USD currency, etc., based on which,
the net predictions are for the next day
AAPL 67% of models predict up
AMZN 54% of models predict up
NVDA 26% of models predict up
Since the pandemic, both AMZN and NVDA experienced sudden pull-up for various reasons, only AAPL was always on a steady course. It happens that the net prediction of models is most favourable for AAPL among the three.
Under this medium horizon, the effective (averaged) net prediction for the upcoming day is always averaged using net predictions made over the past 6 months. Why the averaging is further explained in the white paper (section “Backtest”). What’s being displayed on tsterm are all averaged net prediction as illustrated: the one for the upcoming day would be quite smooth.
In Part 2, we’ll talk about interest rates and currency.
In the future, we may still keep and display causal graph for a short horizon, likely days or weeks ahead, but it is possible we’ll begin to compute on intra-day data for it, which seems informative for short horizons. Stay tuned.