I kept a relative strength screen running for a long time before I noticed it was quietly working against me. It ranked everything by current RS score and handed me the top twenty. Clean list, nice green numbers, and every name on it had already run. By the time an asset earns a top-decile RS score, most of the move that produced that score is behind it. I was buying leadership at the exact moment it was best known and most crowded.
The fix is not complicated once you say it out loud. Rank the change in relative strength, not the level. An asset sitting at the 55th percentile that was at the 35th a couple of weeks ago is doing something more interesting than one parked at the 95th that has been there for a month. The first one is accelerating. The second one is holding, which is fine, but holding is not where the next leg usually comes from.
Level versus slope
Relative strength is just an asset's performance measured against a benchmark or against its peer group. Most people compute it, sort descending, and stop there. That gives you the level. What you actually want alongside it is the slope, meaning how fast that RS reading is improving.
There are two ways to measure the slope and I use both because they catch different things.
- Rate-of-change in the RS line. Take the raw relative strength series, an asset divided by its benchmark, and measure its percentage change over a lookback. A rising RS line means the asset is outpacing the benchmark right now, regardless of where its absolute rank sits.
- Rank improvement. Compute the RS percentile for every asset in the universe today and again as of some prior date, then take the difference. A name that jumped from the 40th percentile to the 70th over two weeks moved thirty rank points. That delta is the signal.
The rate-of-change catches raw momentum in the ratio. Rank improvement catches relative momentum against the whole field, which matters because a rising tide can lift an asset's RS line while it still loses ground to its peers. When both agree, the asset is genuinely climbing through the pack. When only one fires, you look closer.
A watchlist you can actually build
Here is the version I run. It works on stocks, on crypto, on anything where you have a clean price history and a sensible benchmark to divide against.
- Pick a universe and a benchmark. For equities that might be a sector index or the broad market. For crypto I usually measure against a majors basket rather than a single coin, because one coin as your denominator drags its own noise into every reading.
- Build the RS line for each asset as asset price divided by benchmark price, and smooth it lightly. A short moving average on the ratio kills a lot of single-bar garbage without lagging you into next month.
- Compute two lookbacks per asset. A shorter one, roughly one to two weeks, and a longer one, roughly one to two months. The short window is your early warning. The long window tells you whether the improvement has any staying power or is just a bounce.
- Rank the whole universe by RS percentile today, and again as of the start of each lookback. Store the deltas.
- Sort your watchlist by rank improvement over the short window, then filter to names where the long window is also positive or at least not falling apart.
The output is a list of assets moving up through the middle of the distribution. Not the leaders, the challengers. Some of them will stall, plenty will, but the ones that keep climbing tend to show up on this list well before they print the new high that gets them onto everyone else's top-RS screen.
The failure modes, because there are a few
The first one is dead-cat improvement. An asset that fell hard shows a lovely positive rate-of-change simply because it stopped falling. Percentage change off a low base is flattering. The long-window filter catches most of this, and I also glance at the absolute trend. If the asset is still below where it was two or three months ago, a two-week RS pop is a bounce until proven otherwise.
The second is thin names gaming the ranking. In a wide universe, the biggest rank improvements often belong to small, illiquid things that jumped on one piece of news. A liquidity floor, some minimum on dollar volume, quietly removes a lot of false positives. If you cannot get into the trade at size without moving the price, the rank improvement is not tradeable anyway.
The third is over-tuning the lookbacks. It is tempting to hunt for the window lengths that would have caught last quarter's big winner. Do not. Pick two horizons that make sense for how long you actually hold, one short and one longer, and leave them alone. A watchlist that only works on one specific historical run is not a watchlist, it is a story you told yourself.
The fourth is treating the list as a buy list. It is not. It is a queue for attention. An asset surfacing here has earned a closer look, a check of the actual chart, the volume, whatever else you use to decide. Rank momentum tells you where to point your eyes, and the rest of your process still has to say yes.
Wiring it into a real routine
I run the ranking once a day after the close, which is plenty. Relative strength is a slow-moving thing and refreshing it every few minutes just adds noise and temptation. The daily job spits out three groups. New arrivals, names that were not on the list yesterday and just crossed my rank-improvement threshold. Persisters, names climbing for several days running, which I trust more than a one-day jump. And fallers, names whose rank improvement has rolled over, which is my cue to stop watching before I get attached.
The persisters group is the one that has earned its keep. A single day of rank improvement is often noise. Four or five days of steady climbing through the field is the thing that, more often than not, precedes the move that lands an asset on the ordinary top-RS screens. You are just seeing it a couple of weeks earlier.
The cross-asset angle is worth a line too. Because everything is normalized to a percentile within its own universe, you can run the same logic across stocks, crypto, and whatever else you track, then look at which universe is producing the most fresh rank-improvement across its members. When one asset class starts throwing off a cluster of climbers, that is usually the tape telling you where attention is rotating. I built a version of this into Blockcircle's scorecards for exactly that reason, so the same rank-momentum view runs across venues instead of one market at a time.
None of this predicts anything with certainty, and I am wary of anyone who says their screen does. What it changes is the timing of your attention. Instead of meeting leaders after they are leaders, you meet them while they are still arguing their way up from the middle. Some of those arguments fail. The ones that do not are the reason the list is worth running.