Why a 2024 song and a 1996 song with the same play count are not the same size
Streaming counts are the best single measure of how big a song is. They are also badly misleading the moment you put two eras next to each other.
The rating pipeline behind #1 Debut leans hardest on verified streams. A track with around 500 million plays lands in the mid-to-high 80s; a track past a billion opens up the low 90s. Those bands are calibrated against known anchors and applied across the whole catalog, which is what stops the ratings drifting into taste. Two songs with comparable evidence get comparable numbers whether or not anyone on the team likes them.
The problem is that the counter did not start running at the same time for everybody.
The asymmetry
Spotify launched in 2008 and did not reach anything like its current scale until the middle of the following decade. A song released in 2024 has spent its entire commercial life in a world where streaming is essentially how music is consumed, and its play count is a fairly direct measure of how popular it is right now.
A song released in 1996 spent its commercial life somewhere else. It was sold as a CD single, played on the radio, taped off the radio, and put on mix discs, and none of that produced a number anyone can query today. Its stream count is not a record of how big it was. It is a record of how many people went looking for it afterwards, which is a different and much smaller quantity.
Feed both into the same ladder and the ladder quietly becomes a recency machine. Songs that were enormous before 2010 get filed alongside modest recent hits, and a two-year-old single that is still riding a playlist outranks records that shaped the decade they came out in. Anyone who knows the music can feel that the list is wrong, even if they cannot immediately say why.
What I tried first
The obvious fix is a cap by era: decide that an album from a given period tops out at a given rating and clamp everything underneath it. I ran that, and it does suppress the recency problem, but it introduces something worse. It hands a song its ceiling because of when it was released rather than because of anything the song did. A forgettable album track from a celebrated year inherits a ceiling it has not earned, and a genuinely enormous recent single gets held down by a rule it has no way to answer.
So the rule now is that album leaders have to earn the era cap from the evidence rather than inherit it. The era logic moved somewhere less destructive: instead of capping the output, it adjusts the input.
Age-adjusted streams
Before a track's play count is looked up on the ladder, it is discounted according to how recently it was released. The discount runs on a straight line: a song out this year has roughly 39% shaved off its raw plays, a two-year-old release loses about 32%, a six-year-old release about 14%, and anything eight years old or more is looked up on its raw number with no adjustment at all, on the reasoning that by then a song has had a full streaming lifetime to accumulate whatever it was going to accumulate.
Eight years is a chosen cutoff, not a derived one. I picked it because it comfortably covers how long it takes a genuine hit's streaming numbers to stabilize, and adjusted it by ear against catalog I know well. It has not been fit to an external benchmark, and there is nothing sacred about the number eight. If it is wrong, it is wrong by degree rather than by direction: the case for discounting new releases at all is stronger than the case for exactly this curve.
Kanye West and Ty Dolla Sign's Carnival is the clearest worked example I have. Its 2024 numbers are genuinely huge, and on raw streams it came out at 90, sitting in the same band as songs that have held rotation for twenty years. Age-adjusted, it lands at 89.
One point sounds like nothing. At the top of a 55-to-99 scale it is not nothing, because that is roughly where the boundary sits between a very large recent hit and a song that has already demonstrated it lasts. The adjustment is not trying to punish new music. It is trying to stop the pipeline from treating "popular right now" and "durably enormous" as the same claim, when only one of them has actually been tested.
The constraint I would not break
Whatever the adjustment does, it is not allowed to reorder two songs from the same year relative to each other. It is a transform applied per age, applied identically to everything of that age, and it stays monotonic. If two 2019 singles ever swapped places because of it, that would be a bug and not a judgement call.
That requirement is also why the decay curve is smooth rather than bucketed. I learned that the hard way, on an afternoon when three Red Hot Chili Peppers songs all came out to exactly 94, but that is its own story.
What this does not fix
It is worth being straight about the limits, because there are several and they are not small.
Pre-streaming catalog is still under-measured. Age adjustment stops discounting after eight years. It never goes the other way and credits a 1979 single for the decade of radio play nobody logged. That weight has to be carried by chart peaks and certifications instead, which are curated by hand, which means coverage is uneven and always will be.
It cannot distinguish durable from merely old. Nothing in the adjustment knows the difference between a catalog song people still choose and a 2003 album track that has been quietly ignored for two decades. Both are simply past the decay window.
It sidesteps the real question. The top of the scale is ordered mostly by streams and views, and both of those measure popularity. Critical stature is a separate thing, curated on a few dozen tracks, which means very-popular-but-not-canonical songs currently sit beside canon in the low-to-mid 90s. Age adjustment makes that band more defensible without answering whether the number is supposed to mean "biggest" or "greatest". I have not settled that, and it is the argument I return to most often.
What it means when you are drafting
Practically: recent smashes are a little cheaper than their raw play counts suggest, and older songs with real chart history are a little better than theirs imply. If a 2024 single and a late-nineties single feel about equally famous to you, the older one is usually the stronger pick, because it is carrying weight the streaming number never captured.
Play #1 Debut