Method — Data Sources and Limits

Price data

Daily closes come from Yahoo Finance. We use adjusted closes, so dividends and splits are included.

Including dividends is not optional. Leaving them out understates returns for bonds and high-yield assets and reorders the ranking. We checked the 12-month total returns of US long treasuries, investment grade corporates and high yield corporates against the figures published by the fund manager; the largest gap was under 0.01 percentage points.

The price history is stored with the code for this site. If a provider goes down the site keeps working, and the record of delisted assets survives.

How month-end values are made

We do not request monthly bars. Instead we take the close of the last US trading day of each month from daily data. Monthly bars would let an unfinished month arrive as a partial aggregate, so a month still in progress would be treated as though it had ended.

A month that has not ended has no month-end close. For that month we use the most recent trading day's close instead, which is why its scores and ranking are provisional. Nothing is invented — it is a real close, just not the month's last one.

Months with no trading days are not invented. Filling a gap with the previous value would create a return that never happened.

Period returns are computed once from two month-end closes, not by chaining monthly returns together.

Confirmed and provisional

The month we print is not the month the numbers came from — it is the month you hold the result for. The September signal is computed from the last trading day close in August, is final once August ends, and stays put through September. Someone looking for the September signal in September should not land on an August page.

So a month being in progress does not make its signal provisional. What is not settled is next month's signal: until this month ends we compute it early from the latest close and label it provisional, shown differently from a confirmed one.

The asset strength ranking always uses the running month's data, so it is always provisional. After the US market closes we pull that day's close and recalculate, and scores and ranks can differ from one day to the next within the same month.

Whether a month counts as confirmed is decided by the data rather than the calendar: we check whether the prior month's last US trading day close is in hand. Holidays, time zones and daylight saving do not move the answer.

Once a month is confirmed we store it and never recalculate it. If dividends are later restated and past values shift slightly, the signal we published does not change quietly underneath.

Momentum scores

Dual Momentum uses a single 12-month total return. VAA and DAA use the 13612W composite — the 1, 3, 6 and 12-month returns weighted 12, 4, 2 and 1, divided by four.

13612W is defined in the Keller and Keuning papers; it is not a blend of ours, and we print the source next to it. Because it is a weighted average rather than a return, it never carries a percent sign.

Assets without enough history to score are excluded from the ranking rather than filled with zero. Missing is not the same as zero.

Calculating in Korean won

The ranking can be switched between US dollar and Korean won. The won view multiplies each asset's dollar price by the exchange rate at the same point in time.

The currency genuinely reorders the ranking in some months: an asset that rose in dollars can have fallen in won. Strategy signals themselves are always calculated in dollars, as the papers define them.

Known limits

Trading costs and taxes are not included. These rules rebalance monthly, so following them in practice costs money.

The hurdle in Dual Momentum is meant to be the T-bill rate; we approximate it with a short-term treasury ETF. The difference in maturity shifts the hurdle slightly.

The world ex-US equity ETF listed in 2007, so we do not calculate earlier periods for it. We do not splice in a similar asset to extend the history.

For commodities held by a strategy we use the ETF the paper specifies, which diverges from spot prices because of the cost of rolling futures contracts. We follow the construction in the paper knowingly.

Crude oil in the ranking goes the other way: we use the futures price, not an ETF. No paper fixes that choice for us, and oil ETFs drift further from the spot price as rolling costs accumulate.

We do not publish backtests of our own. Small changes in assumptions move those numbers, and using such a number to suggest an action would stop this from being information.