SWR using Momentum investing

Pat38493

For those who are interested in SWR research, this article on Early Retirement now seems to be generating quite a bit of interest:

https://earlyretirementnow.com/2025/11/12/momentum-trend-following-swr-series-part-63/

This guy has been posting updates to his SWR series for a few years, and this is the first time I've seen him say that a strategy he was investigating made a significant positive difference.  Usually when he investigates some new silver bullet he concludes that it is rubbish.

The idea is around changing your asset allocation dynamically (and quickly)  based on the strength of market trend signals.  In theory this would result in you selling a lot of your equities soon after the start of a bear market, and buyingn back soon after it starts going up again.

However there is a catch - it involves rebalancing your portfolio monthly in response to mathematical momentum  indicators, so anyone wanting to try this would need to assess whether his estimates of transactional charges based on US markets would be too low for the UK.

Secret2ndAccount

In the States there is a momentum index, and an ETF (MTUM) which tracks it. If you want the UCITS version for the UK, that's IUMO,  but the algorithm seems quite different from what you are describing. They look back 6-12 mths to determine the momentum. Flitting in and out of the market on a monthly basis sounds like a recipe to miss out on some big gains, which often occur shortly after big losses.
I've heard plenty of spiel from the guy pushing his quality/momentum ETF. It rebalances quarterly, based on momentum, but will only admit companies which consistently make a profit. Always sounds like double snake oil to me. He's underperformed the market by about 25% in the last 5 years, but of course collected a healthy pile of fees doing so.

Pat38493

Secret2ndAccount said:

In the States there is a momentum index, and an ETF (MTUM) which tracks it. If you want the UCITS version for the UK, that's IUMO,  but the algorithm seems quite different from what you are describing. They look back 6-12 mths to determine the momentum. Flitting in and out of the market on a monthly basis sounds like a recipe to miss out on some big gains, which often occur shortly after big losses.
I've heard plenty of spiel from the guy pushing his quality/momentum ETF. It rebalances quarterly, based on momentum, but will only admit companies which consistently make a profit. Always sounds like double snake oil to me. He's underperformed the market by about 25% in the last 5 years, but of course collected a healthy pile of fees doing so.

Yes - I think the ERN research though is mainly focussed on improving the minimum safe withdrawal rate over long retirement periods rather than beating the market.  It appears in the simulations for this approach, that there was a significant improvement from about 3.5 to 3.9%, which is a pretty big improvement compared to other approaches he has tested.  The simpler one which provided a smaller improvement was the glide path - starting retirement with lower (60% or 70%) equities but gradually increasing the equities over the first 10 years or so.

michaels

Interesting analysis although fitting clever strategies to historic data is to me always questionable as you will be looking at rules that worked and for any random path it is no doubt possible to find such rules....

QrizB

michaels said:

Interesting analysis although fitting clever strategies to historic data is to me always questionable as you will be looking at rules that worked and for any random path it is no doubt possible to find such rules....

For some extreme examples, see:

https://www.tylervigen.com/spurious-correlations

Linton

Another clever investment strategy I remember from years ago was only invest in companies with names starting A-L. The data showed that they were signicantly likely to perform better than those with names in the second half of the alphabet.

Pat38493

michaels said:

Interesting analysis although fitting clever strategies to historic data is to me always questionable as you will be looking at rules that worked and for any random path it is no doubt possible to find such rules....

To be fair to the ERN guy, in the article he admits that the formula he chose is somewhat arbitrary and it certainly can't be stated that it is the best one you could find.  He also acknowledges that there is a danger of overfitting which is why he didn't make the formula even more complicated.

It may seem ridiculous, but fundamentally it is trying to see what happens if rather than maintaining a fixed allocation or set of rules, or a time based one, you alter your allocation based on the trends that appear to be happening - e.g. if equities are starting to trend down, you gradually rebalance away from equities.

This might seem daft, but it's no more daft than holding x years of cash to cover you during a market downturn as you have no idea how long the downturn will last and where the bottom is.  If trend indicators are that the market is going down, it arguably makes sense to sell soon after the top than to sell at the bottom.

He also acknowledges that using this approach, there would have been a few false flags where you rebalanced but then the market moved back the other way again.  

For those who aren't familiar with ERN arbicles, he has a long history of being extremely sceptical about these kind of approaches so he seemed surprised himself that this showed a significant difference in the initial simulations.

Cus

I guess with any momentum type investment strategy it's critical you enter early and exit early otherwise you've been pumped in a way.

Secret2ndAccount

Yeah, I had to sell my Facebook shares when they changed their name to Meta

michaels

Pat38493 said:

michaels said:

Interesting analysis although fitting clever strategies to historic data is to me always questionable as you will be looking at rules that worked and for any random path it is no doubt possible to find such rules....

To be fair to the ERN guy, in the article he admits that the formula he chose is somewhat arbitrary and it certainly can't be stated that it is the best one you could find.  He also acknowledges that there is a danger of overfitting which is why he didn't make the formula even more complicated.

It may seem ridiculous, but fundamentally it is trying to see what happens if rather than maintaining a fixed allocation or set of rules, or a time based one, you alter your allocation based on the trends that appear to be happening - e.g. if equities are starting to trend down, you gradually rebalance away from equities.

This might seem daft, but it's no more daft than holding x years of cash to cover you during a market downturn as you have no idea how long the downturn will last and where the bottom is.  If trend indicators are that the market is going down, it arguably makes sense to sell soon after the top than to sell at the bottom.

He also acknowledges that using this approach, there would have been a few false flags where you rebalanced but then the market moved back the other way again.  

For those who aren't familiar with ERN arbicles, he has a long history of being extremely sceptical about these kind of approaches so he seemed surprised himself that this showed a significant difference in the initial simulations.

IF there was a simple rule that could beat the market consistently then all the massive amount of analysis that have been thrown at the market data would have arbitraged it out of existence (or will going forward if we assume that such analysis was not complete in the past).