Skip to content

Timing signals

Pick a market and a trigger. Every row is one buy rule with the record of how often it has actually worked there.

Current market: Normal Alliance CN. This browser opens every tool here until you pick another. Address for this market: ahledger.com/signals?market=forever/normal/alliance/cn

▸How this is worked out

One rule, tested honestly: buy when an item's price for the day falls at least 20% under its own average since prices began, which becomes a trailing 30 day average once the market has 30 days of prices (it has 9), while the price is otherwise steady. A day's price is the average of that day's scans, so a two hour crash that recovers by evening will not fire it. An item with fewer than 10 days of prices is not judged yet.

Steady means the seven days before today average less than 10% under the seven days before those, with at least three days seen in each. An item that far under its average on a falling trend is listed as still falling, not as a buy: after a launch most prices settle for weeks, and a dip bought on the way down waits for a recovery that is not coming. An item without enough days to read its trend is neither.

A firing counts as a win when the price comes back within five percent of the average inside 7 days. Each day under the trigger is a firing of its own, so a dip that lasts five days counts five times. A rule needs 5 scored firings before it shows a hit rate, and "if it recovers" is what one unit makes sold back at 95% of the average, the least a win needs, less the house cut (the deposit comes back when it sells).

Not enough Normal Alliance CN history yet

An item needs 10 days of prices before the rule can judge it, and no item on Normal Alliance CN has them yet: its prices begin November 8, so the first firings can come around November 18. The rule is running; it reports the moment there is enough history to judge. Set up scanning to build that history faster.

Common questions

What is a timing signal?
A buy rule tested against the market's own history: buy when an item's price for the day falls at least 20% under its average since prices began (9 days) while its price is otherwise holding steady. Until a market has 30 days of prices, the average covers every day it has, and an item with fewer than 10 days of prices is not judged yet. A day's price is the average of that day's scans, so a short crash that recovers by evening does not fire it. A firing counts as a win when the price comes back within five percent of the average inside 7 days.
Why is an item under its average not always a buy signal?
Because it may still be falling. When the seven days before today average at least 10% under the seven days before those, the item is listed as still falling instead of firing. After a launch most prices settle for weeks, and an item bought on the way down waits for a recovery that is not coming. The rule fires once the price holds steady.
Why do some signals show no hit rate?
A rate from two or three firings says nothing, so a rule needs 5 scoreable firings before the page shows one. A firing is scored only once its 7 day recovery window has passed, and each day under the trigger counts as one, so a dip that lasts five days is five firings. On this market the first rates can appear around November 29. Until then the page shows what is firing today, the backtest table lists each rule's scored firings from the first one on, and every scan adds to that history.