Really nice and simple. Start your own ETF!

Hi, not bad, you chose well. Could you please post a short list of these hedge fund-style ETFs that dont't work just to make a comparison?

I got the list of hedge fund-style ETFs from here:
http://etfdb.com/etfdb-category/hedge-fund/
Here's a list of actively managed bond ETFs (I used a list I found somewhere else, but this one is more complete):
https://www.thebalance.com/actively-managed-bond-etfs-the-complete-updated-list-416942

So I just threw this together as a proof of concept. It definitely needs improvements before live trading.

One problem is that I manually weeded out the ETFs that had negative alpha or super high beta. I think ideally you'd have them all in there and have the algorithm dynamically weed out the ones that don't fit criteria. Since so many of these ETFs were only recently released, there wasn't enough out-of-sample data to do that properly. So even though I did use a small time sample (a few months in 2017) to do this weeding out, I introduced quite a bit of lookahead bias by doing that. The algo performs best during and around my in-sample, and the results are terrible during 2015 -- so a 1/3rd of the time it's failing. Not so good. A back test all the way to 2007 did produce 1.0+ sharpe, but as I pointed out, most of these ETFs are so recent, that historical data is irrelevant. It'll be fun to see the next 6 months to a year if it fails or not. Due to the mean-reversion weighting it could start putting an emphasis on perpetual losers if their behavior changes and they start consistently failing. Equal weighting might be safer even though it won't deliver as great results.

Again, it needs to be a bit more rigorously reconfigured before it's ready for live trading, but as a starting point I think it's nonetheless already safer than investing in any single of the underlying ETFs from that list.

Hi Viridian Hawk,

That is a very interested strategy especially given that fact that you have near zero beta with a long only portfolio. Are some of the ETFs negatively correlated?

Best regards,
Aqua Rooster

Aqua,

So yeah, basically the algorithm is going long a bunch of ETFs that internally employ long/short strategies. The hedge fund ETFs I selected exhibit between 0 and 0.3 beta, and the actively managed bond ETFs probably have negative beta (I assumed but didn't check), and then I threw in TLT as well, which is also negative beta. So that's how it evens out to 0 beta.

TLT is maybe a problem since it's so volatile, and people are speculating about a bond bubble.

I thought on a whim that this strategy might be useful to somebody who's really lazy -- you know, just take some Wall Street alpha instead of finding your own. But the expense ratios on these types of ETFs are terrible, especially considering their performance stats. So that eats away at your profit vs capital risked. Puts you at a disadvantage. But I guess you have to look at whether the expense ratios are better or worse than the transaction costs you'd be paying as a retail investor at IB with your own long-short strategy -- and that's even if you have a good long-short strategy that's dependable. I suspect a lot of people struggle finding something that continues to perform reliably out of sample.

Looking at the returns graph you can see that despite the bond hedge and 0 beta it was still hit pretty hard during the Oct 14, 2014. crash, so I'm super skeptical about how resilient these hedge fund-style ETFs are.

Swapped out SMA for TRIMA with a 50 day window.
Returns Increased 1.1
Drawdowns Increased 0.7
Alpha Increase 0.01
Beta Increase 0.01

I dont normally care for these types of algorithms but the low beta and simplicity of it all was interesting.

Is there any issue with partial fills due to the low volume?

Yeah -- these ETFs are super low volume. Some have huge bid-ask spreads -- so rebalances are going to cost a lot each time.

Hi Viridian,

Thanks for posting. There is one caveat to your algo that I see. Mind you, I haven't checked the turnover rates, just the fact that it is called each day. Some of the chosen ETF have very, very low volume and purchase cost would be prohibitive due to slip.

True. ETFs shouldn't slip too far from NAV, but I guess it can happen. However, you can rebalance monthly or quarterly and get similar results.

However, I do not recommend this strategy. I think the hedge fund-style and alternative ETFs available on the market are complete rubbish. So it doesn't surprise me that out-of-sample this algorithm has produced 0 alpha. Also, my methodology of hand-selecting the ETFs that appeared to deliver value simply introduced a look-ahead bias to the backtest. The weighting scheme makes no effort to neutralize beta, balance risk, or account for the nature of the various instruments -- it was arbitrarily chosen and tweaked (overfit) to deliver a nice result.

In short -- a terrible methodology making use of terrible instruments. Maybe a springboard for better ideas, but definitely not something anybody should trade.