The crypto calendar, every month since 2011

Monthly return. For a month m, return is (close(m) - close(m-1)) / close(m-1), each close being the last traded price of that calendar month in UTC. Nothing adjusted, nothing annualised.
Average against median. Average is the arithmetic mean of every completed instance of that month, median is the middle value. They separate badly when one year is extreme, which is exactly why both are offered. Most sites show one and let you assume.
Win rate. Completed instances of that month closing above the month before, divided by completed instances. Always shown with its observation count, because eight observations and eighteen are not the same evidence.
Range. (high - low) / low inside the month, a plain measure of the ground it covered.
Quarter and year. Compounded from the monthly returns inside the period, prod(1 + r) - 1, never a mean of months.
Day of week. From daily closes, (close(d) - close(d-1)) / close(d-1), grouped by the UTC weekday of the closing day.
Year against year. Cumulative product of monthly returns from January, so every line starts at zero and the shape is comparable across years of wildly different price.
Projection. Each remaining month takes its own historical median return, compounded forward from the last close. The band is the 25th to 75th percentile of that month's historical returns, compounded the same way. It is a restatement of the past, not a forecast, and it carries no information about what is coming.
The running month. Excluded from every average, median, win rate and streak. It appears once, dashed, as month to date, so an unfinished period can never flatter the record.
Halving countdown. (1050000 - height) x 600 seconds from the live block height, so it tracks the network's real pace instead of assuming a date. Halving months are ringed gold in the grid.
The market-wide indices. TOTAL, TOTAL2, TOTAL3, OTHERS, TOTAL-ES and TOTAL-E50 are built here, not bought in. Basket is the top 100 by market cap with wrapped and staked duplicates removed, so nothing is counted twice. For each month, every member that traded in that month and the previous one contributes its return weighted by its previous market value, weight = supply x close(m-1), giving R(m) = sum(w x r) / sum(w), chained into a level. Members joining or leaving change the next return, never the level, so there are no phantom jumps. Stablecoins are held at one dollar rather than fetched. This is an honest approximation of a market-cap index, it is not any vendor's TOTAL2, and it will not match one to the decimal.
Altcoin season. For each month, the share of the largest alts whose return beat Bitcoin's return that same month. The pool is the top 50 by market cap with stablecoins and wrapped or staked duplicates removed, and only those with a full monthly history on a major venue, which in practice is around two dozen names rather than fifty. Every cell prints its own count, and months with fewer than eight are dropped entirely. Above 75% is the usual definition of alt season, below 25% is Bitcoin season. It measures BREADTH, how many beat Bitcoin, and says nothing about size: a month can be broadly green for alts while everything still falls.
History depth. Bitcoin reaches 2011 by stitching an older daily record onto the modern one. Other coins begin at their own first traded month, printed beside the coin name so a short history is never mistaken for a long one.