The TFSA book
For a TFSA, the recommendation is XAW 48%, XIC 32%, VDY 20% — three funds, adding to 100%.
It is the cell labelled “40/60 · VDY tilt” in the table below. That label is a ratio, not an allocation: the tilt is carved out of the whole, not added on top of it, so the equity pair is funded pro rata and the three weights sum to one hundred. The weights here are printed from the declaration that walked, never retyped.
It was the strongest of the 4 blends that passed every clause — the largest return edge over XEQT at +1.87 percentage points per five-year window, on 22 of 24 windows, with the second-best drawdown at +1.14 points.
What it gives up, stated. The 40/60 · XDV tilt blend shelters 0.30 points more in a downturn — and charges 1.49 points of return for it. The rule this test applies is that protection has to be worth more than it costs, and that shelter is not. The cheaper sleeve buys return, not shelter. Both blends are in the table below; neither is hidden.
Declared after round two closed. The nine-cell grid and its first tilt were written down before any of them ran. This blend was not — it became declarable only when its fund's fee document arrived, and its expected advantage was written into the record before that document existed, so a good result could not be read as a discovery. It is marked as such wherever it appears.
This is a TFSA book and nothing else, and the test models no tax at all. Not for a TFSA either — the previous wording named withholding tax only when explaining the RRSP, which could be read as saying the figures above account for tax in a TFSA. They do not, for any account. What a TFSA changes is how much that omission distorts: distributions here are measured as they were actually paid, already net of any tax withheld inside a fund, and no registered account recovers that. An RRSP's treatment of foreign dividends differs in ways this test does not carry, so no RRSP book has been ruled. Nothing here is advice, and no blend on this page is bought, held or offered.
The blend question, and the answer
11 declared blends of Canadian and world equity, with and without a bond damper, each run through 24 rolling five-year windows on the same prices and the same deposit schedule as the funds, and compared against XEQT window by window. 4 were adopted. 7 were refused, and the refusals are on this page because they are the reason to believe the rest.
The grid was written down before any of it ran, in a record that cannot be revised afterwards. What follows is every cell of it, plus one candidate declared after it closed and marked as such — its expected advantage was written into the record before the document that made it declarable even existed, so a comfortable result could not be read as a discovery.
| Blend | Return vs XEQT median · won |
Drawdown vs XEQT median · won |
Return vs XGRO median · won | |
|---|---|---|---|---|
| 25/75 · no damperXAW 75% · XIC 25% | +0.79pp 21 of 24 | -0.39pp 5 of 24 | +11.42pp 24 of 24 | refused protection |
| 25/75 · damper 10%XAW 67.5% · XIC 22.5% · ZAG 10% | -4.85pp 0 of 24 | +1.26pp 24 of 24 | +5.88pp 24 of 24 | refused protection |
| 25/75 · damper 20%XAW 60% · XIC 20% · ZAG 20% | -9.69pp 0 of 24 | +2.61pp 24 of 24 | +1.08pp 24 of 24 | refused return allowance/protection |
| 40/60 · no damperXAW 60% · XIC 40% | +1.03pp 22 of 24 | +0.31pp 20 of 24 | +11.68pp 24 of 24 | adopted |
| 40/60 · damper 10%XAW 54% · XIC 36% · ZAG 10% | -4.40pp 0 of 24 | +1.72pp 22 of 24 | +6.30pp 24 of 24 | refused protection |
| 40/60 · damper 20%XAW 48% · XIC 32% · ZAG 20% | -9.00pp 0 of 24 | +2.71pp 24 of 24 | +1.49pp 23 of 24 | refused return allowance/protection |
| 50/50 · no damperXIC 50% · XAW 50% | +1.36pp 18 of 24 | +0.68pp 17 of 24 | +11.82pp 24 of 24 | adopted |
| 50/50 · damper 10%XIC 45% · XAW 45% · ZAG 10% | -3.81pp 1 of 24 | +1.79pp 17 of 24 | +6.73pp 24 of 24 | refused protection |
| 50/50 · damper 20%XIC 40% · XAW 40% · ZAG 20% | -8.92pp 0 of 24 | +2.78pp 17 of 24 | +2.40pp 17 of 24 | refused return allowance/protection |
| 40/60 · XDV tiltXAW 48% · XIC 32% · XDV 20% | +0.38pp 16 of 24 | +1.44pp 19 of 24 | +10.97pp 24 of 24 | adopted |
| 40/60 · VDY tiltdeclared after round two closedXAW 48% · XIC 32% · VDY 20% | +1.87pp 22 of 24 | +1.14pp 19 of 24 | +12.49pp 24 of 24 | adopted |
The margins are small and are printed at their true size. The 4 adopted blends beat XEQT by a median of +0.38 to +1.87 percentage points per five-year window. That is a direction, not a margin anyone should act on alone.
A second floor was applied and never bound: every blend also had to beat XGRO, the off-the-shelf 80/20 portfolio, and all 11 did — by +1.08 to +12.49 points — the column is in the table above, per cell. It is on the record because a test only counts if it could have failed.
24 overlapping five-year windows are one path chopped 24 ways, not 24 experiments. Adjacent windows share almost all of their days. The window counts describe stability across start dates; they are not significance, they do not become significance by being counted, and nothing here should be read as saying they have. That sentence entered this project's record when the store supported 32 windows; the count changes and the rule does not.
What the refusals say
The bond damper never paid. Every blend carrying bonds was refused. At 10% the damper cost between 3.8 and 4.9 percentage points of return to buy between 1.3 and 1.8 points of shallower drawdown; at 20% it cost 8.9 to 9.7 to buy 2.6 to 2.8. The rule this test was written to apply says protection must exceed what it costs, and it never did.
One blend won 21 of its 24 windows on return and was refused anyway, because its drawdowns were deeper than the benchmark's. Winning on return is not the test.
What this does not settle
Eight of the store's 32 windows are not in this comparison. XEQT began trading on 14 August 2019 and did not exist for them; a comparison against a benchmark that did not exist is not a comparison, and including them made every blend look better than it is.
These are results over one stretch of one market, from January 2019 forward. More Canada happened to be better on both counts over it, and that is a statement about this stretch and not about Canada. Nothing here is advice, and no blend on this page is bought, held or offered.