3 new Free Smart Alpha models

I just opened three free Smart Alpha models that have been in incubation. They are . . .

  1. LOW VOLATILITY SELECT - SP 500

https://www.portfolio123.com/app/r2g/summary?id=1379348
Momentum: 72, Value 41, Quality 91

This is a so-far successful effort to control volatility, not through statistical measures but based on company fundamentals. The market has given it a good test lately, and to this point, it’s doing what I expected it to do.

Notice that the Value score is low. Those who’ve been following my on-line seminars (all are welcome, one can start any time), know that the association of value with conservatism is not necessarily accurate. You have to pay up for protection. That’s so when you buy insurance, when you invest and use hedge vehicles, or whether you buy lower-risk companies.

  1. CURVE FITTING FOR FUN AND MAYBE PROFIT

https://www.portfolio123.com/app/r2g/summary?id=1382192
Momentum: 91, Value 69, Quality 65

There’s been a ton of talk in the forums about curve fitting, data mining, and so forth. How people concoct those rocket-ship 90%-plus alphas is really no mystery at all. Since they are predicting the past, the factors they use are known with 100% certainty and easily discoverable. I constructed a ranking system using many of these and decided, for fun, to see if I could make sense of them work out of sample.

There may be something. Besides using lower-is-better rank factors for size and price (the easiest way to pump[ simulated performance), I notice a lot of things that people say are Value or Quality but are, in truth, momentum (as can be seen by the way my supposedly quality-value ranking system produced a portfolio that actually matches up highly with classic momentum and not so much with the other things). But it’s not price momentum. It’s “fundamental momentum.” Is that a legitimate factor? That’s what I’m trying to figure out. I also matched that ranking system with a screen that hopefully pre-qualifies the universe to reduce the probability of dumpster fire companies reaching the rank process.

So far, in a bad market, one that has hammered many high-momentum models whose that did not use timing-relatred rules to get positions down to zero, this momentum monstrosity has matched the Russell 2000 ETF, which is not bad considering that my naive expectation for a style profile like that called for much worse. It should, at the very least, be interesting to watch.

  1. SMART ALPHA EQUITY INCOME
    https://www.portfolio123.com/app/r2g/summary?id=1389937
    Momentum: 68, Value 54, Quality 83 Yield: 4.4%

This is a different approach to dividend security. Rather than relying on historic payout ratios.I rely on the full constellation of company quality and Street sentiment to incorporate a greater variety of factors.

Total return since launch, during this bad market period, has been marginally better than SPY. Relative to major equity-income ETFs, it has outperformed those that aim to capture some equity-like returns but underperformed those designed more with a view toward imitating the bond market. Looking ahead, if rates rise, I believe the latter could suffer since the top line of the yield fraction is fixed. I like having bond exposure but prefer to do that elsewhere (through the still-in=incubation Guggenheim bond ladders). So in terms of what I want, yield plays that preserve some equity-lie qualities, not hot now but which could be good for a longer horizon, this model is doing what I expected it to do.

Marc, thanks a lot for sharing your knowledge and your models.

Thanks a lot - do you care to share the backtest charts as a picture in advance?

Just include the models in a Book simulation to see the historic simulation results.

Walter

Cool! Thank you Marc!

Normally, as many here know :-), I’m leery of over-reliance on backtests (I use them only as feedback to see if I effectively articulated an idea in language a server can understand and process; I never use a test to tell me if the idea is worthwhile or not.)

That said, if you’re asking about the Curve Fitting model, I suppose i can whip up a pdf since that’s a special case, a for-fun model designed to explore regularly data mined factors.

I am well aware of that. You can’t hide it :wink:

But you provide an answer for my curiosity yourself:

However, if you have articulated your idea well enough and if it is based on a sound strategy, there should be at least a slim chance that it has outperformed in the past as well - hence, my interest in the results. As an indication (not proof) of validity. I find it even more interesting to look at a backtest result after a strategy has been designed without any/much curve-fitting to start with.

I would be even more curious about the factors you used than the backtest, but I find it more understandable if you don’t provide those.

Anyway, the discussions here are what matter, not a display of more flashy backtests.

Actually, that’s not necessarily so. What I look for in sim is that the strategy performed as I expected it would. For example, a while back I created what the sim suggested was a really good REIT strategy. But I never released it because it only matched the major REIT ETFs. so I figured that my model added no extra value. The Low volatility Select model wasn’t special in terms of returns relative to the SP 500, but it succeeded in cutting the volatility; there are times when I’d expect it to underperform. I test income models against major equity income ETFs and focus on specific periods, such as the 2013 taper tantrum and the recent months.

I’ll get to a pdf on the curve fitting model when I get a chance.