PATTERN+ PREDICTIONS
Updated

Scoreboard · which events move price · 341 past events in 7 kinds

Which events really move price

Every kind of event we count, against the same coins' ordinary days: how often the price was higher after it, how often it is higher after any day, and how big the move was. A kind gets a call only when its whole 90% range clears the ordinary days. Past patterns, not financial advice.

Kinds of event

7

each against its coins' ordinary days

With a clear edge

5

the whole 90% range clear of ordinary days

Too few cases

0

under 20 measured, no call yet

Past events

341

measured, with ordinary days to compare

The boardevery kind of event and its edge, next day, 7 and 30 days

§ 01 · Every kind of event5 with a clear edge

Every kind of event, clearest edge first

as of · against the same coins' ordinary days

Each cell is the edge in points: the share of past cases that were higher that many days later, minus the share of the same coins' ordinary days that were. Green or red only when the whole 90% range is clear of the ordinary days; grey when it is not; faded and in brackets when there are under 20 cases, too few to say. Hover a cell for its numbers.

EventCasesNext day7 days30 daysMove size
DELIST Delistings 77 -39 pts -32 pts (-31) 3.1×
LISTING Listings 30 +12 pts -1 pts -27 pts 0.9×
CPI US CPI releases (Bitcoin's move) 29 +20 pts +26 pts +6 pts 1.1×
HACK Hacks 69 -18 pts -11 pts -5 pts 1.3×
UNLOCK Unlocks 80 -2 pts +1 pts +13 pts 1.1×
PERPS Perp listings 36 +8 pts +6 pts -13 pts 1.5×
FED Fed rate decisions (Bitcoin's move) 20 -2 pts -9 pts (-18) 1.4×
  • After hacks, the price was higher the next day in 28% of 69 cases, against 45% of the same coins' ordinary days: -18 points (90% range -26 to -8).
  • After unlocks, the price was higher 30 days later in 45% of 67 cases, against 32% of the same coins' ordinary days: +13 points (90% range +3 to +23).

Source: every measured event on record against its coin's daily closes · fetched

§ 02 · How it is counted

How it is counted

as of · methodology

Each kind's past cases are counted against the same coins' ordinary days in the year before each case, so a kind is never credited with a market that was simply rising.

A case is a past event with a measured move: from the coin's close the day before the event to its close 1, 7 or 30 days after the event day, the same rule as every pattern page. Events that are only known once their day is over, like a day's funding or a day's stablecoin printing, are dated the day after, so no move is counted before the news was out.

A case's ordinary days are every window of the same length in the year before it, on the same coin, that ends before the event: about 330 to 365 of them. The share of those windows that ended higher is averaged over the cases, so a kind of event that happens mostly in rising markets is compared with rising markets.

The edge is the cases' share higher minus the ordinary share, in points; its 90% range is the Wilson interval of the cases' share. A call needs at least 20 cases and a range clear of zero. With 7 kinds and three horizons, a few ranges will clear zero by chance: the pattern pages check every kind on its newest cases, and an edge that holds there is the one to lean on. The move size is the median size of the move 7 days after the cases over the median size of the ordinary 7-day moves, up or down.