Transcription

• Speaker #0

  Hello, welcome back to Papers with Backtest podcast. Today, we dive into another Algo trading research paper.

• Speaker #1

  And today we're looking at adaptive asset allocation, a primer. It's by Adam Butler, Michael Philbrick, and Rodrigo Gordillo from Resolve Asset Management.

• Speaker #0

  Right, the 2015 revision specific.

• Speaker #1

  Well, that's the one. And it really dives into a core problem with, let's say, traditional investing ideas like modern portfolio theory, MPT.

• Speaker #0

  OK. MPT. What's the issue, Grace?

• Speaker #1

  Well, the paper argues that MPT often leans really heavily on these, you know, super long term average figures for returns, for risk.

• Speaker #0

  They're predicting decades out.

• Speaker #1

  Exactly. And the authors kind of ask, are those long term guesses really the best foundation for building a portfolio now?

• Speaker #0

  Uh-huh. And they bring up that acronym, right? GIGO.

• Speaker #1

  Yeah. Garbage in, garbage out. It's a classic. Basically, if the data you feed your model those long term return forecasts, volatility estimates, correlations.

• Speaker #0

  If that data is garbage or maybe just, you know, not very accurate for the current environment.

• Speaker #1

  Then the portfolio that comes out the other end, well, it might not be optimal. It could even be counterproductive for you as an investor.

• Speaker #0

  Makes sense. If you build based on flawed assumptions.

• Speaker #1

  Your results probably won't match your expectations. And they show this with figure two in the paper. Looking at stock versus bond excess returns. Right.

• Speaker #0

  That chart was interesting. It wasn't just a straight line up for stocks.

• Speaker #1

  Not at all. It shows how much that relationship fluctuates. There were actually pretty long periods where bonds did better than stocks, which kind of flies in the face of the simple stocks always outperform long term idea.

• Speaker #0

  So those long term averages we hear about aren't always reliable predictors for shorter or even medium time frames.

• Speaker #1

  Precisely. And then there's figure three looking at like. actual investor holding periods.

• Speaker #0

  People don't hold things for 30 years usually.

• Speaker #1

  Often not, no, maybe just a few years, typically, which again creates a mismatch if your whole strategy is built on multi-decade averages.

• Speaker #0

  So, okay, this sets the stage for adaptive allocation.

• Speaker #1

  Exactly. Instead of static long-term guesses, the paper explores using shorter-term observed market data, strategies that react.

• Speaker #0

  And for this deep dive, we're really going to focus on the specific trading rules they tested, and, crucially, The backtest results they found.

• Speaker #1

  Yep. Let's get into the mechanics and the numbers.

• Speaker #0

  Okay. First up, the baseline. The equal weight portfolio. How did that work?

• Speaker #1

  Pretty simple stuff. They took 10 global asset classes.

• Speaker #0

  Like stocks from the US, Europe, Japan, emerging markets.

• Speaker #1

  Right. Plus REITs, both US and international, different US treasuries, like intermediate and long-term, plus commodities and gold. A broad mix.

• Speaker #0

  And just split the money equally.

• Speaker #1

  10% in each, rebalanced every month. That's it. The sort of naive... approach, as they call it.

• Speaker #0

  And the back test. This is from 1995 to 2014.

• Speaker #1

  Over that period, it gave 8.1% compound annual return. Volatility was 11.2%.

• Speaker #0

  Okay, sharp ratio.

• Speaker #1

  0.72. And the maximum drawdown, the biggest peak to trough loss, was pretty significant, negative 39.2%.

• Speaker #0

  Ouch. Almost 40% down at its worst. So that's our starting point.

• Speaker #1

  That's the benchmark. Now, the first adaptive strategy they tested was volatility weighted.

• Speaker #0

  All right. How does that

• Speaker #1

  Instead of equal weight, The allocation changes monthly based on each asset's volatility over the previous 60 days.

• Speaker #0

  So inverse volatility. More volatile assets get less weight.

• Speaker #1

  Exactly. The idea is to make each asset contribute roughly the same amount of risk or volatility to the total portfolio. With a cap, though, max 100% exposure.

• Speaker #0

  Interesting. So managing risk more actively, how did that perform?

• Speaker #1

  The compound return actually ticked up a bit to 8.5%. But the real story was the volatility. Lower. Yeah, significantly lower, down to 8.6%.

• Speaker #0

  Ah, so that boosts the risk-adjusted return.

• Speaker #1

  Right. The Sharpe ratio improved quite a bit to 0.99. And maybe most importantly for someone watching their account balance.

• Speaker #0

  The drawdown.

• Speaker #1

  Much better. Maximum drawdown was cut substantially down to Tenegr 24.2%.

• Speaker #0

  Okay, that's a big difference from nearly negative 40%. Shows the power of just managing risk based on recent volatility. See,

• Speaker #1

  definitely. It smooths the ride quite a bit.

• Speaker #0

  Next. they brought in momentum. Strategy three was top five equal weight by six month momentum. Yep.

• Speaker #1

  So here every month you rank all 10 assets based on their total return over the past six months.

• Speaker #0

  The trend following idea. What's been going up?

• Speaker #1

  Kind of. You pick the top five performers from that list.

• Speaker #0

  And just hold those five.

• Speaker #1

  Hold those five in equal weight. So 20% each for the next month. Then you re-rank and rebalance again.

• Speaker #0

  Focusing on the recent winners. How did that strategy do?

• Speaker #1

  This is where the returns really started to pick up. Compound return jumped to 13.0%. Wow,

• Speaker #0

  quite a leap from the 8ish percent range. What about volatility?

• Speaker #1

  Volatility was actually similar to the basic equal weight, around 11.0%.

• Speaker #0

  Okay, so higher returns for similar volator. That means a better sharp.

• Speaker #1

  Much better. Sharp ratio hit 1.17. And the max drawdown also improved further, down to Negness 21.7%.

• Speaker #0

  Interesting. So focusing on momentum gave more return and didn't really add much risk compared to the baseline, even slightly reducing the worst drawdown.

• Speaker #1

  Seems so. It suggests there's value in following recent trends, at least in this context.

• Speaker #0

  Now, the next logical step seems to be combining the previous two ideas, momentum and volatility weighting.

• Speaker #1

  Exactly. Strategy four, top five. by six-month momentum volatility weighted.

• Speaker #0

  So same first step, pick the top five based on six-month returns.

• Speaker #1

  Right. But then instead of weighting them equally, you apply that inverse volatility logic we talked about earlier to just those five assets. Ah,

• Speaker #0

  so you pick the current leaders, but then you scale their position size based on how bumpy their recent ride has been.

• Speaker #1

  Precisely. Aiming for equal volatility contribution among the chosen top five. It's a neat combination, trying to get the best of both worlds, capture momentum, but manage the associated risk.

• Speaker #0

  Okay, how did that combo work out in the backtest?

• Speaker #1

  Even better. Compound return nudged up again to 14.0%, and the volatility actually decreased this time, down to

• Speaker #0

  9.9%. Lower volatility and higher returns. Yep,

• Speaker #1

  which means the Sharpe ratio got another significant boost, up to 1.41.

• Speaker #0

  Very nice. And the maximum drawdown?

• Speaker #1

  Improved again, quite dramatically, actually, down to negative 14.8%.

• Speaker #0

  Under 15% drawdown. That's starting to look really compelling from a risk management perspective for investors.

• Speaker #1

  Absolutely. It really highlights how layering these different adaptive factors can potentially build on each other.

• Speaker #0

  Okay. That brings us to the final and perhaps most complex strategy they tested. Top five by six month momentum, minimum variance.

• Speaker #1

  Right. This one adds another layer of sophistication.

• Speaker #0

  Still starts with the top five momentum selection.

• Speaker #1

  Yes. Find the top five performers over the last six months. But then instead of just using inverse volatility weighting.

• Speaker #0

  It does something else.

• Speaker #1

  It uses a minimum variance optimization. It looks at those five chosen assets and calculates the specific weights for each one that would have resulted in the lowest possible overall portfolio volatility. Ah,

• Speaker #0

  so it considers not just individual asset volatility, but also how they move together their correlations.

• Speaker #1

  Exactly. That's the key difference. Minimum variance takes those correlations into account to try and build the smoothest possible portfolio from those top momentum assets.

• Speaker #0

  Trying to maximize diversification benefits within the selected group.

• Speaker #1

  You got it. It's a more holistic view of... They did a few things. First, they looked at returns above the risk-free rate, subtracting the 10-year treasury yield to see the actual risk premium earned.

• Speaker #0

  Makes sense. You want to know what you're getting for taking the risk.

• Speaker #1

  Right. Then they factored in estimated trading frictions, things like commissions and potential slippage when you buy or sell. They assumed higher costs earlier in the period, scaling down to about 0.5%. percent more recently.

• Speaker #0

  Realistic. Trading isn't free.

• Speaker #1

  And finally, they also deducted a hypothetical 1% annual management fee.

• Speaker #0

  Okay, so looking at that top strategy momentum plus minimum variance, how did it hold up after subtracting the risk-free rate, trading costs, and fees?

• Speaker #1

  So the nominal return was 15.1%. Subtracting the 10-year treasury yield brought that down to 10.2% annualized excess return.

• Speaker #0

  Still pretty solid.

• Speaker #1

  Definitely. Then, subtracting the yield and those estimated frictions and fees, the annualized excess return came down to 7.7%.

• Speaker #0

  7.7% above the risk-free rate after costs and volatility.

• Speaker #1

  The volatility stayed the same in their calculation across these adjustments at 9.4%.

• Speaker #0

  So how did the risk-adjusted performance, the Sharpe ratio, look after these deductions?

• Speaker #1

  The Sharpe was 1.61 on the nominal return. It dropped to 1.09 after subtracting the yield. And then. to 0.82 after subtracting the yield and the costs and fees.

• Speaker #0

  OK, 0.82, still a respectable sharp ratio, especially considering it's after costs and represents return over treasuries.

• Speaker #1

  Exactly. And the maximum drawdown figures also shifted slightly. It was anise at 8.8 percent nominal, then negus at 10.2 percent after subtracting the yield, and a negus at 11.3 percent after all deductions.

• Speaker #0

  So even after applying these real world adjustments, the drawdown remains quite contained around ending at 11 percent. And the risk-adjusted excess returns are still compelling.

• Speaker #1

  That's the conclusion, really. Even when you account for the practicalities, the adaptive approach, especially that final strategy, still looks very strong compared to the passive baseline. And remember, that 7.7% is the return over what you'd get from treasuries. The paper noted the 10-year yield was about 2.2% then, so your total nominal return could still be substantial.

• Speaker #0

  Right. So wrapping this up, what's the main message for someone listening thinking about their own allocation?

• Speaker #1

  I think the key takeaway is that this paper really demonstrates quite clearly through backtesting how adaptive strategies, ones that react to market conditions using things like momentum and volatility, have the potential to significantly improve risk adjusted returns.

• Speaker #0

  Compared to just holding a static mix.

• Speaker #1

  Yes, especially strategies that combine factors like using momentum to select assets and then using volatility or minimum variance techniques to construct the portfolio. They seem to offer better returns and substantially reduced drawdowns in the periods studied.

• Speaker #0

  A compelling argument for not just setting and forgetting, but adapting based on what the market is actually doing.

• Speaker #1

  Precisely. It's about dynamically managing both risk and return potential.

• Speaker #0

  Thank you for tuning in to Papers with Backtest podcast. We hope today's episode gave you useful insights. Join us next time as we break down more research. And for more papers and backtests, find us at https.paperswithbacktest.com. Happy trading.