This repository contains multiple signal-generation algorithms.
They all serve the same product, but they do not solve the same problem, operate on the same time horizon, or rely on the same assumptions. Treating them as one unified model would be misleading.
This document explains how the system is structured at a high level, and where to find the detailed logic for each signal type.
- [V0] WSB scoring algorithm
- [V1] WSB scoring algorithm
- [V0] House trades scoring algorithm
- [V1] House trades scoring algorithm
Quantarded is not a price prediction engine. It is a signal extraction system.
Each algorithm answers a specific question:
-
WSB signals
What is attracting disproportionate retail attention right now, and is that attention clearly bullish or bearish?
-
House trades signals
Which disclosed congressional trades look non-random, clustered, and meaningful enough to matter over a mid-to-long time horizon?
Both produce a “confidence” number, but those numbers are not comparable across algorithms. They live in different contexts and should be interpreted differently.
This separation is intentional.
| Signal type | Typical horizon | Update cadence |
|---|---|---|
| WSB | days to 1–2 weeks | continuous, intraweek |
| House trades | weeks to months | event-driven, slow |
Trying to force these into a single scoring model would either:
- make WSB too slow, or
- make House trades too noisy
So we do not do that.
All confidence scores exposed by Quantarded are:
- relative
- capped
- intentionally conservative
They express conviction within a signal type, not probability of profit.
A 75 confidence on WSB and a 75 confidence on House trades do not mean the same thing, and they are not meant to. If that ever becomes confusing, the algorithm failed, not the user.
Every algorithm in this system should be able to answer one question clearly:
“What does this signal know that the market might not fully reflect yet?”
If it cannot answer that honestly, it does not belong here.