A ranked leaderboard is a seductive thing. You sort a basket of assets by trailing return, the top row lights up green relative to the rest, and the natural move is to rotate into it. I have done this and been burned by it, so I want to be specific about why. The asset at the top of a relative strength ranking can be falling in absolute terms. It is just falling less than everything else. In a broad drawdown, the leaderboard hands you the least-bad thing and calls it a winner, and you end up long the best-looking loser on the way down.
The fix is not to throw out relative strength. It answers a real question. The fix is to understand that it answers a different question than absolute momentum, and to make both of them agree before you take the trade.
Two questions that sound like one
Relative strength is cross-sectional. You take a set of assets, measure each one's return over the same window, and rank them against each other. The output is ordinal. It tells you who is beating whom right now. It says nothing about whether any of them are going up. If you rank ten things and eight are bleeding, the top of that ranking is still a name that is bleeding, just slowest.
Absolute momentum is time-series. You take one asset and compare it against its own past, or against cash, or against a risk-free proxy. The output is directional. It tells you whether this thing is actually trending up on its own terms, independent of what its neighbors are doing. Historically this is the filter that keeps trend-following systems out of extended bear markets, because when everything correlates to one and rolls over together, absolute momentum goes flat or negative across the board and tells you to stand down.
The trap lives in the gap between ordinal and directional. A ranking will always produce a number one. It has to. Someone is always in the lead. That is a property of sorting, not a property of the market being healthy. When you treat the top rank as a buy signal without checking direction, you have quietly assumed that the leader is rising, and the ranking never made that promise.
How the failure actually shows up
Picture a risk-off stretch where a whole sector or a whole asset class is sliding. Your rotation model still needs to be invested, so it does its job and rotates you into whatever is falling slowest. Each rebalance you hold the name that dropped the least since the last one. On a relative basis you are winning. On your equity curve you are down, month after month, holding a rotating cast of assets that are all individually in downtrends. The model is doing exactly what you told it. You just never told it to check whether the leader was above water.
I have seen the same shape in cross-asset dashboards. A commodity or a beaten-down alt tops the momentum board during a general risk-off move because its decline lagged the others by a week. The ranking screams strength. The price chart shows a lower high and a lower low. Both are true at once, and the ranking is the one that gets people into the position.
The reverse failure exists too and matters less but is worth naming. Pure absolute momentum with no relative filter will keep you in things that are grinding up slowly while a much stronger cohort runs away from you. You are trending, technically, and leaving a lot on the table. That is an opportunity cost, not a hole in your account, which is why I weight the absolute filter as the safety and the relative filter as the optimizer.
Stacking the two filters
The workflow I keep coming back to is a gate followed by a rank. Absolute momentum is the gate. Relative strength is the rank. Order matters, because the gate has veto power.
- Compute absolute momentum for every candidate. A simple version is trailing return over your window minus the risk-free rate, or price above a long moving average. Anything that answers up or down on the asset's own terms.
- Drop everything where absolute momentum is not positive. This is the veto. A falling asset does not get to be a winner no matter how it ranks.
- Rank whatever survived by relative strength. Now the leaderboard means something, because every name on it is already trending up in absolute terms.
- Allocate to the top of that filtered ranking. If nothing survives the gate, you hold cash. That is the whole point. There are stretches where the correct exposure is none.
The failure mode this specifically prevents is the empty-gate case getting silently overridden. If your relative rank runs first and always outputs a top pick, you will always be fully invested, including through the drawdowns where you should not be. Running the absolute gate first lets the model return nothing, and being allowed to hold cash is most of the edge here. A rotation system that cannot go to cash is a system that is long the least-bad loser at the worst possible time.
Practical notes and a couple of traps
Use the same lookback for both filters when you start, then tune separately. It is tempting to run a fast relative window and a slow absolute window, and there are reasons to, but you introduce a subtle bias where the two filters are effectively measuring different regimes and disagreeing for reasons that have nothing to do with the asset. Get them agreeing on a shared window first, understand the behavior, then let them drift apart on purpose.
Watch correlation during risk-off. Relative strength quietly assumes the assets in your basket are somewhat independent, so that a leader is meaningfully different from a laggard. When correlations spike toward one, that assumption breaks and the ranking becomes noise. The leader and the laggard are the same trade with a few days of lag between them. This is exactly the environment where the absolute gate earns its keep, because it does not care about the basket, only about direction.
Beware very short windows on the relative side. A one-week relative ranking will happily promote a dead-cat bounce to the top of the board. It is not strength, it is a violent retrace inside a downtrend, and a longer absolute filter usually catches it and vetoes the entry.
When I am checking a candidate by hand, I ask two separate questions and refuse to collapse them. Is this asset beating its peers, and is this asset actually going up. Cross-asset scorecards on Blockcircle let me pull both reads side by side, but you can do the same with two columns in a spreadsheet. The discipline is keeping them as two columns. The moment they become one blended score, you lose the ability to see that your leader is a name that is falling, and that is the exact thing that puts you long the wrong asset in a down market.