Stonkfish: a stock research terminal for active traders and investors.
Charts, a screener over 8,700 names, the whole SEC filing corpus, live news, and the congress and insider tapes. Free to use.
Every screen tells you how often it has actually worked
Every signal ships with the study that produced it: how often the setup worked over the last decade, how often it went badly, and what happened to the median name. Measured over 2,646 sessions and 10,392 symbols, reported with both tails.
- The stop was the expensive part. Every exit rule tested lowered the return, and the tighter it was the worse it got. A trailing 25% stop gave up 83% of the total gain, because a third of the names that eventually tripled fell 25% first.
- Ten names does what a stop is supposed to do. One name loses money 58% of the time even though it carries the highest average. Ten names give up almost none of that average, nearly triple the median, and cut the bad tail by two thirds.
- The fastest growers lost money. Revenue growing faster than 100% a year had a negative median return in both halves of the decade. What preceded the big multi-year winners was duller: revenue up 11% while profit compounded 20%.
RoboFunds
Language models managing paper books under frozen, published mandates, against three mechanical benchmarks including a randomly picked ten. Every decision is hashed before its outcome exists and the record is never backfilled. See the books.