Back to Community
Multi Factor Smart Beta Strategy You Can Actually Use with Confidence

I'm not a big fan of wall street.

I think too many of them routinely take every opportunity to profit off the average retail investor. And I get so frustrated when there are blatant attempts to overcomplicate investing and scare folks into overpaying for complex strategies that can be replicated with far simpler and cheaper strategies. What pisses me off the most is when they have outrageous minimums that prohibit an average investor from taking part in the "fun."

Take this interview with Cliff Asness, CIO at AQR. This is a quant firm that develops and uses algorithms to manage their investment products. As I watch this interview I'm liking everything he's saying: that markets are relatively efficient but that there are some tried-and-true factors that exist which can be exploited with quantitive analysis and decision making. The interview ends, and curiosity gets the best of me and I check out their products.

One of their products is a multi-factor (value, momentum, and quality) mutual fund, QSMNX that I rather like the strategy of. But then I go check what the fee and minimum is... they require a $1M minimum investment and charge a 0.97% expense fee, are you freaking kidding me! Not only is the minimum asinine, the expense fee is so stupid when you take a moment and realize they have an algorithm running it!

I want to prove to wall street that their arrogance will eventually get the best of them. I want to prove that there are non-professionals (like all of us) who know enough about investing to develop strategies that can compete with theirs; and that some of these people genuinely want to better the financial lives of others while asking for nothing in return.

So here's a smart beta strategy that can work in a Robinhood account (no trading fees) for a balance as low as $10K. Here's what I do to build an index of 20 stocks, rebalancing monthly:

  • Take the smallest 50% based on market cap, and lowest 50% of stocks based on P-E per sector
  • Of the remaining companies, filter away the lowest 50% based on dividend yield
  • Of the remaining companies, filter away the highest 50% based on price volatility
  • Take the largest 2 stocks left (tendency for mid-cap)
  • Combine all 20 stocks and determine weighting using mean reversion (companies that have made it through our filters aren't typically exhibiting momentum phenomena, they are typically undervalued companies that are oversold)

Is this the sexiest strategy in the world? No. Does this remove or minimize market risk? No. But does it work? Yes.

This is a strategy that you can feel comfortable running on your real money. It doesn't need a lot of capital. It is diversified across all sectors. And best of all this lets you feel the excitement many of us seek with investing. You get to own individual securities and you can feel engaged with your investments.

Let's show wall street that we don't want to keep overpaying them for products we can replicate ourselves for less money and for more enjoyment. Oh, and we'll probably beat those twits in the process.

Clone Algorithm
148
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 58ab8ca8d446705dfac75b31
There was a runtime error.
12 responses

Stephen -

You might have a look at the attached tear sheet. If I'm reading things correctly, the algo takes huge positions in individual stocks, which is a problem if one is putting a significant chunk of net worth into this thing (glancing at QSMNX, it looks like they limit exposure to ~1% in any given stock). Also, it seems pretty darn volatile, so it is not clear what it buys; the overall return is higher than SPY, but it seems that the risk is higher, too.

Also, I would suggest picking an appropriate benchmark ETF. You are referencing SPY, but it sounds like this is a small-cap strategy, right? Is the idea to "beat" the Russell 2000? Or something else?

Loading notebook preview...
Notebook previews are currently unavailable.

The high individual stock weights are due to the mean reversion weighting which I softened below. I will say that the mean reversion weighting scheme is effective and overweights stocks that have significantly dropped the month before. It also rebalances monthly so over the course of a full year your company specific risk is reduced. If you're a contrarian investor like myself, you may enjoy buying stocks everyone else is selling. But I understand I'm a little weird and most people couldn't stomach that.

Below, I also added in 40% allocation to intermediate treasuries for more risk reduction. Now the max holding (other than the intermediate treasuries ETF) is 7%; I'll attach the tear sheet immediately after.

For benchmarking, I should actually use mid-cap value for apples-to-apples. In the below algorithm though the appropriate benchmark is Vanguard's balanced index fund (VBINX). I still used SPY for appropriate alpha and beta calculations but here's a plot that shows the hypothetical growth of $10K invested in one of the following options for the same time period (1/1/2007 to 1/31/2017):

  • $19.7K VFINX: Vanguard's S&P 500 Index Fund
  • $21.1K VMVIX: Vanguard's Mid Cap Value Fund
  • $14.8K FGOVX: Fidelity's Government Income Fund
  • $18.6K VBINX: Vanguard's Balanced Index Fund

The growth of your $10K investment with this algo/index would be to $24.7K.

Clone Algorithm
148
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 58ac2a44d446705dfac76552
There was a runtime error.

Here's the tear sheet.

Loading notebook preview...
Notebook previews are currently unavailable.

Stephen,

If you will allocate 50% to long treasuries the backtest results will be even better in any risk metrics.

Total Returns
155.5%
Benchmark Returns
96.6%
Alpha
0.07
Beta
0.28
Sharpe
0.99
Sortino
1.44
Volatility
0.10
Max Drawdown
-23.9%

Total Returns
1.47
Benchmark Returns
0.97
Alpha
0.05
Beta
0.48
Sharpe
0.85
Sortino
1.23
Volatility
0.11
Max Drawdown
-0.30

Clone Algorithm
33
Loading...
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 58ac6baa696e475df57e573b
There was a runtime error.

Thanks Vladimir,

Long term government bonds have definitely been historically a good compliment to a portfolio of equities. I was being a little speculative (I know I shouldn't do that) and theorizing that long term government bonds won't fair as well in the coming 20 years compared to intermediate bonds.

Looking at the plot of yields over time, the last 15 years of this bond market probably won't repeat itself moving forward. Intermediate bonds have done better in periods where yields remained flat or trended upward (corresponding downward move in price).

But either option is a good one and a nice compliment to equities!

Stephen -

You might also consider if the same kinda performance could be achieved with a suitable combination of ETFs, rebalanced "smartly" per an algo. In the end, I think you can get much better diversification.

Also, it seems that your second algo above mostly gets its boost from avoiding the Great Recession, and not from the beating the market after that. You also end up with ~50% in a single stock per the tear sheet. Kinda scary.

Grant, that "~50% in a single stock" is IEF, the iShares intermediate treasury ETF, not a single stock and not very scary. That algorithm has 40% in treasuries (IEF) yet is generating equal returns compared to the market (look at the beta and volatility). If you want to beat it, adjust line 28 to be 1.0 (full on equities/stocks).

The point I'm trying to make is that this is a good starter algorithm to use if you want to own individual securities. This lets someone have a bit of "fun" yet still satisfy traditional investment best-practices (diversification across sectors, diversification with stocks/bonds, under valued companies paying good dividends etc.). I'm not trying to suggest this is the best algorithm out there, its not. But its simple, stable, and its worked.

Sorry...missed that IEF is an ETF. No criticism here, really. Just a thought that if the same thing can be accomplished with ETFs (I don't know), then it might be preferred from a simplicity and diversification standpoint.

No worries, I do have you to thank in the first place for highlighting the high single stock allocation in the original version (I was amplifying the mean reversion piece by an order of magnitude)!

It's a valid point in regard to ETFs, and there probably is a sector or asset class rotation strategy that has offered similar performance while providing more "diversification." But there is such a thing as too much diversification in that at some point, you'd be better off just owning the whole dang market with one index fund.

What this (or a similar algorithm) provides is that entertainment and engagement piece that comes with owning individual securities. I've been calling my strategies or "engineered indexes" a way to make boring investing fun. Boring in the sense that its buying undervalued, underloved, and relatively un-sexy companies. But its fun in that you get to feel a bit of excitement of owning a small enough number of companies that you can remember their names. You get to actually feel like you are investing in a piece of something, and I've been surprised how much of a vested interest I have on the companies in my indexes. And of course its fun when it seems to work (at least historically).

Interesting work in pipeline. Unfortunately with default commissions, the second algo at least, fails to beat the benchmark (in part due to unaccounted-for negative cash).

2017-01-31 13:00 pvr:243 INFO PvR 0.0344 %/day cagr 0.1 Portfolio value 20053
2017-01-31 13:00 pvr:244 INFO Profited 10053 on 11512 activated/transacted for PvR of 87.3%
2017-01-31 13:00 pvr:245 INFO QRet 100.53 PvR 87.32 CshLw -1512 MxLv 1.10 RskHi 11512
2017-01-31 13:00 pvr:333 INFO 2007-01-03 to 2017-01-31 $10000 2017-02-22 02:30 US/Eastern
Runtime 0 hr 49.0 min

Don't forget about Robinhood! But yes, if you want to use something similar in an IB account it'll be more effective with more capital than $10K to overcome transaction costs.

Also, remember though that the above algo has a 40% (or Vladimir's 50%) allocation to bonds. If you get rid of the bond component it wil beat the benchmark even with transaction costs and an initial balance of $10K.

93% without the bond component, better. Still below the benchmark at 97%. (Second algo)
If you remove the ordering of .safe, returns will appear to be only 66%, that's because it doesn't use all cash. Returns on the amount invested: 93%.