Sep 23, 2023 10 min read Strategies

Strategy: Buy Top-Performing ETFs (Canadian Markets)

Click here for the version for US markets

Overview

This tactical asset allocation strategy switches between the best-performing assets among a basket of diversified passive index ETFs.
The goal is to outperform the market over the medium and long term by capturing most of the upside of the growth assets while significantly reducing drawdowns by switching to defensive assets in bear markets.
This type of strategy has been extensively studied and written about (see recommended reading). Still, it continues to be effective because it relies on market momentum (trend following), a persistent market feature.
Compared to buy & hold strategies, this strategy requires additional effort. It is tax-inefficient (if held outside a tax-sheltered account) but has much better returns and significantly smaller drawdowns in exchange.

General Rules

Select a diversified basket of ETFs (ideally with low correlation).
At month’s end, rank the performance of each ETF (including reinvested distributions) using the average of 3/6/12-month price change.
Buy (or continue holding) an equal dollar amount of the top-ranked ETFs (best performance).
Rebalance monthly (if large enough to matter compared to the associated transaction costs).

Global Diversified Portfolio

Assets

Symbol	Name	Inception Date	Daily Trading Volume ($ Millions)	Annualized Return
XEG	iShares S&P/TSX Energy	2001	$20.9	6.3%
XFN	iShares S&P/TSX Financials	2001	$9.8	9.0%
XGD	iShares S&P/TSX Gold	2001	$7.4	5.0%
XIT	iShares S&P/TSX Information Technology	2001	$0.6	7.3%
XRE	iShares S&P/TSX Real Estate	2002	$6.0	7.8%
XST	iShares S&P/TSX Consumer Staples	2012	$0.7	13.6%
XUT	iShares S&P/TSX Utilities	2012	$4.7	5.8%
XQQ	iShares NASDAQ 100 (CAD-Hedged)	2011	$6.3	16.1%
XSU	iShares U.S. Small Cap (CAD-Hedged)	2007	$3.0	4.8%
VEE	Vanguard FTSE Emerging Markets	2012	$1.0	4.4%
XBB	iShares Canadian Universe Bond	2000	$5.9	3.6%
XSB	iShares Canadian Short-Term Bond	2000	$1.7	2.9%
XCB	iShares Canadian Corporate Bond	2006	$1.5	3.5%

The average correlation between the daily price movements of these ETFs is low at 0.30 (0.11–0.42).

Performance (Apr 2002–Dec 2023)

	Strategy	Buy & Hold All ETFs	Benchmark (XIC)	Benchmark (60/40)
Starting with $10,000	$152,348	$55,050	$54,414	$38,328
Annualized Return	13.8%	8.5%	8.4%	6.6%
Maximum Drawdown (EOM)	-13.9%	-27.6%	-43.6%	-27.4%
Maximum Drawdown Length in Months	22 (2011–2012)	20 (2008–2010)	32 (2008–2011)	27 (2008–2010)
Best Month	21.7%	12.2%	11.2%	8.8%
Worst Month	-10.2%	-14.3%	-17.7%	-11.9%
Average Month	1.16%	0.72%	0.75%	0.57%
% Positive Months	67%	63%	64%	67%
Best Year	37.8%	28.2%	34.6%	22.4%
Worst Year	-8.0%	-20.6%	-33.3%	-19.0%
% Positive Years	90%	71%	71%	71%
Average # of Trades Per Year	15.7	0	0	1
Safe Withdrawal Rate (25 years)	[coming soon]

Buy Top ETFs vs. Benchmarks, logarithmic-scaling, growth of $10,000, $0 transaction fees

Returns by Year

	Strategy	Buy & Hold All ETFs	Benchmark (XIC)	Benchmark (60/40)
2023	10.4%	13.5%	11.7%	9.8%
2022	12.7%	-9.1%	-5.8%	-8.0%
2021	26.3%	15.9%	23.4%	12.4%
2020	36.9%	11.4%	5.6%	7.4%
2019	28.3%	22.8%	22.8%	16.5%
2018	0.1%	-3.6%	-8.7%	-4.8%
2017	8.0%	9.1%	9.0%	6.4%
2016	4.1%	16.1%	21.0%	12.8%
2015	1.3%	-0.3%	-8.6%	-3.9%
2014	25.4%	11.6%	11.0%	9.9%
2013	32.8%	7.2%	12.5%	6.8%
2012	3.5%	8.2%	7.0%	5.6%
2011	-5.1%	-1.5%	-8.8%	-1.6%
2010	10.4%	13.2%	17.4%	13.0%
2009	26.1%	28.2%	34.6%	22.4%
2008	-8.0%	-20.6%	-33.3%	-19.0%
2007	3.3%	1.4%	9.6%	7.2%
2006	20.4%	15.7%	17.1%	11.5%
2005	37.8%	17.1%	26.9%	18.4%
2004	11.1%	12.9%	13.0%	9.8%
2003	21.3%	21.3%	24.5%	15.9%

Portfolio Considerations

Taxes — Very tax-inefficient. It is most suitable for tax-sheltered accounts unless you have limited taxable income, as it often generates capital gains, primarily as short-term gains.

Contributions — Are supported at any time. Since this strategy generates new target positions each month and will switch positions often throughout the year, it can absorb contributions quickly with minimal or no additional transactions.

Withdrawals — There will be many opportunities to withdraw cash without triggering additional capital gains throughout the year. Also, existing positions can be trimmed if necessary. Since drawdowns are much smaller than with buy & hold, there is significantly less timing risk for unplanned withdrawals.

Minimum Account Size

$5,000 — when using a zero-commission broker (less if they support fractional shares).
$50,000 — when transaction costs are $9.99 to get total annual transaction costs under 0.5%.

Maximum Account Size

$500,000 — if your account size is more extensive, you introduce liquidity and slippage risk because your transactions could represent a substantial portion of the daily volume. One option is to spread out your transactions over multiple days or run the strategy on additional days other than the last day of the month (e.g. 7th and 21st). The other option, and the one I would suggest, is to diversify into the US markets, which have substantially more daily trading volume.

Diversification — This strategy is diversified across market sectors and asset classes but mainly concentrated in North American assets and Canadian dollars. To effectively diversify globally, you’ll need to trade in US markets.

Note: there will be times when all the positions held by this strategy could be concentrated in higher volatility assets, increasing the chances of more significant drawdowns during market shocks.

Effort — Requires calculating the monthly ETF performance rankings and potentially executing time-sensitive transactions. You can do these calculations yourself or subscribe to my monthly newsletter.

Timeframe — Ideally, it should be at least two years, the shortest rolling period with positive returns 95% of the time.

FAQ

Why use the average of 3/6/12-month returns?

The goal is to calculate the direction of each asset’s trend (momentum). Broad-based ETFs move slowly enough that looking at their performance over the past year is a good gauge of overall direction. But if we only compared today’s price to a single historical price (e.g. 12 months ago), we wouldn’t know the shape of the price movement—is it trending up recently, or was all the price appreciation nine months ago? We want a chart that is trending up and to the right. Using the average 3/6/12-month performance is a simple way (but hardly the only way) to give us a sense of the overall direction.

Why does it work?

Ranking the assets based on their performance gives us a statistical edge in predicting the price performance for the next month—the strongest assets tend to outperform the weakest ones. Notice that the average monthly return and % positive months correlate roughly to rank.

Next Month’s Return vs. Performance Rank

Rank	Average	Maximum	Minimum	% Positive
Top 1	1.42%	39.6%	-19.5%	62%
Top 3	1.16%	39.6%	-19.5%	63%
Top 5	0.99%	39.6%	-19.5%	63%
Bottom 5	0.51%	33.4%	-46.7%	59%
Bottom 3	0.62%	33.4%	-46.7%	59%
Bottom 1	0.35%	33.4%	-46.7%	55%
All	0.69%	39.6%	-46.7%	60%

Why use the top 3 (rather than the top 1 or 5)?

The top-ranked asset has the highest average monthly return but a lower worst-case return and lower % positive than other top-ranked assets. As such, investing only in the top-ranked asset will typically have higher returns and disproportionately higher drawdowns. Buying just the top-ranked asset can be a reasonable place to start if you have a smaller account and a longer investment time horizon (e.g. more than five years).

The top 3 give us a better balance between returns and drawdowns.

Using the top 5 usually doesn’t help reduce drawdowns and typically lowers returns.

Why these specific ETFs?

My selection criteria were:

It doesn’t represent the overall Canadian market (i.e. TSX 60 or TSX Composite).
It has sufficient daily volume to supply enough liquidity to support many people trading the same strategy.
Has at least 10 years of historical data (ideally 20+ years).
It is not highly correlated to another asset with more liquidity.

Using these criteria leads to an unfortunately short list of possibilities. So, unlike the US version of this strategy, there is no practical way to create multiple asset baskets from which to choose.

There were only two actual decision points. The first was between XSP (S&P 500) and XQQ (NASDAQ 100). The strategy’s historical performance changes very little regardless of which is used, which makes sense since these two ETFs are highly correlated (0.87). I chose the higher growth potential of the tech-heavy NASDAQ.

The second decision point was between XIN and VEE for emerging markets exposure. XIN is CAD$-hedged and has more history available, but I chose VEE because it currently has more daily volume and lower management fees. These two ETFs are highly correlated (0.73).

Don’t many of these funds have higher management fees?

Yes, specialized sector funds tend to have higher MERs than funds tracking the overall market (0.6% vs. <0.1%). But that is the cost of getting more concentrated funds. I would not want to buy & hold these ETFs for precisely that reason, but they are well suited for this strategy.

I chose (almost) all iShares funds because they have been trading the longest and thus give us more data to validate the strategy. In the future, I will review if there are newer ETFs that have higher daily trading volume or lower management fees that may be worth substituting.

How important is the timing of executing trades?

This is an important question since the historical performance shown above is idealized in that it’s based on calculating the rankings using the closing price (i.e. after the markets close) but assumes that we executed trades as the markets closed. So, what’s the best way to implement this system?

The first (and easiest) choice is, in the evening after the markets close, enter a limit order at the close price (maybe with a little buffer) that will remain open for a few days and expect that it will get filled (which it usually will). On the rare occasion when it doesn’t, hold your nose, accept the slippage, and buy at the current market price.

The second choice is to execute the trades at the market open the next day (either in real-time or with market orders). If prices move overnight, usually everything will move together, so, for example, you might receive a worse fill on your sell but a better fill on your buy.

The third choice is to calculate the target positions just before the market closes (e.g. 15 minutes), execute a market-on-close (if your broker supports them), and assume that the position signals won’t change in the last few minutes of the market session (which they rarely will). This method will get us closest to the idealized performance but has the disadvantage of being very time-sensitive and not knowing precisely what price your orders will be filled at.

Overall, it is advantageous to implement the new positions at month-end because there is a natural performance benefit to the end-of-month price specifically. But, the price traded will not significantly affect your overall returns. This strategy vastly outperforms buy & hold. Therefore, we won’t be undone by a bit of slippage. Pick the execution method that works for you, keep it simple, and don’t stress over it. This strategy only takes a few minutes a month to implement and can be managed entirely outside open market hours.

For reference, here is a breakdown of the historical performance impact of different trade times.

Trade Price	Return Difference	Drawdown Difference
Signal day closing price	(baseline)	(baseline)
Next day opening price	+0.07%	–1.48%
Next day mid-price	–0.43%	–2.28%
Next day closing price	–0.49%	–2.81%

Why start in April 2002, and how do you handle assets that began trading later?

I wanted to start the analysis as early as possible, so I chose a threshold of six available assets as the minimum (twice as many as held each month). The remaining assets became available to the system once 12 months of history were available (to calculate the 3/6/12-month performance).

Having all assets available rather than six improves diversification and performance and reduces drawdowns. But even using only the original six assets throughout the test, it still handily outperforms buy & hold.

How are the number of trades calculated?

The average number of trades per year (16.0) counts two trades each time an asset drops out of the top 3 and is replaced by another (i.e. one buy and one sell). It does not count trades to rebalance since I don’t know the size of your portfolio and how often it will be cost-effective for you to rebalance. In ~30% of months, the monthly performance difference between the best and worst performing asset held is more than 10%, so there will be at least a few times per year when the positions become obviously unbalanced.

📬

Explore the features of our newsletter, which delivers updates on each of our strategies every month.