Altcoin season index

The index is a counting exercise, not a model. Take the largest assets by market capitalisation, remove the stablecoins, remove Bitcoin because it cannot outperform itself, and ask one question of each survivor. Did it return more than Bitcoin over the same lookback. The yes answers are added up and divided by the number of assets that had a usable price change, and multiplying by 100 gives the reading.

Two thresholds carry the names. A reading of 75 or above is called altcoin season, because three quarters of the field is clearing the benchmark. A reading of 25 or below is called bitcoin season. Everything between is neither, and the number usually sits in that middle band, which is the part the headlines skip.

On the fetch behind this build, the top 50 over 30 days reads 50.0. That is 25 assets out of 50 clearing a Bitcoin move of +4.5%, which puts the reading in the band called neither season.

Everything above is arithmetic anyone can repeat from the same two endpoints. What follows is the part that needs a second data set.

A season index is a statement about one lookback period, and the period is a choice rather than a fact. The original blockchaincenter version uses 90 days. Change the lookback and the same market can read as breadth or as concentration, so publishing one window hides the most important assumption in the metric.

The CoinGecko markets endpoint serves five change windows on a single call, 24 hours, 7 days, 30 days, 200 days and 1 year. It does not serve 90 days. This page therefore does not publish a 90 day figure, and the 30 day default here is not the same measurement as the 90 day figure quoted elsewhere. Comparing the two numbers directly is a category error, and the difference between them is not a signal.

One number at one depth cannot tell you whether outperformance is broad. A reading of 60 across the top 10 and a reading of 60 across the top 200 describe very different markets, and only the second one means the move has reached past the largest names. The ranked results for this query all publish a single depth, so the shape is missing from every one of them.

Read the ladder from the top down. A curve that falls as depth grows means the leaders are carrying the reading and the long tail is lagging. A curve that rises with depth means smaller assets are doing the work. A flat curve is a market moving together.

The standard index treats every asset as one vote. A protocol with real fee revenue and a finished vesting schedule counts exactly as much as a token whose only distinguishing feature is a listing. That is fine as a definition and useless as a description, because the two cases mean opposite things about what is driving the move.

Early Thunder scores 251 tokens on 25 named variables, including protocol revenue, supply inflation, token release schedule, circulating to fully diluted ratio, holder concentration and competitive moat. That research file is dated 2026-09-18. Joining it to today prices on ticker symbol splits the index into two halves at the median score, and the two halves can then be compared against the same Bitcoin benchmark over the same window.

At the top 200, 104 assets matched a scored token. The higher scoring half reads 63.5 and the lower scoring half reads 73.1, both against the same Bitcoin move of +4.5%. Whichever half is higher, that is the half doing the work behind the headline number.

Two limits on this split, stated plainly. The scores are a dated snapshot and the prices are live, so this is research from 2026-09-18 tested against a later market, never a live score. And the join runs on ticker symbol, which is the weakest identifier in this asset class, so a renamed or reused ticker will fail to match and simply drop out of both halves rather than land in the wrong one. The matched count in the table is the honest measure of how much of the slice the split actually covers. The full scoring method sits on the methodology page, and the per token detail sits on the scorecard.

It measures breadth against one benchmark over one lookback, and that is the whole of it. It is not a forecast. A high reading says a lot of assets have already outperformed, which is a description of the recent past and carries no claim about the next month.

Survivorship is baked in. The top 50 is recomputed from today ranks, so an asset that collapsed out of the top 50 during the lookback is simply absent from the count, while an asset that climbed into the top 50 on the strength of that same move is present and counted as a win. The measured population is selected partly on the outcome being measured, and no version of this index avoids that, including the original.

Equal weighting is a second choice with consequences. A hundred billion dollar asset and a two billion dollar asset each contribute one vote, so a reading can be high while the market capitalisation weighted move is flat. The depth ladder is the partial answer here, since comparing the top 10 with the top 200 exposes where the votes are coming from.

Missing data is excluded rather than assumed. An asset with no percentage change for the selected window is dropped from the numerator and from the denominator, and the count of those drops is printed next to every figure. Below 10 comparable assets this page prints no index at all, because a share of eight moves twelve points on a single row.

The reading also changes with the fetch. This page holds one dated snapshot taken at build time, so the number here is already older than the market by the time it is read. For trade level math the crypto profit calculator is the tool, and the current research shortlist sits on the opportunities page.