When the engine publishes a score of 75 next to a stock, it's fair to ask: 75 out of what, and how? A number is only as useful as the reasoning behind it, so here's exactly how the AlphaScore is built — no black box. And because every read the engine publishes is sealed before the market opens and stored in a public database, we can go one step further than showing the recipe: we can show you every score it has ever produced, and let the distribution speak for itself.
The recipe: four reads, equal weight
Every day, for each stock in the universe, the engine produces four separate reads, each on a 1-to-10 scale:
- Technical — the trend and price setup. Is the stock riding above its longer-term average, or falling through support? Is momentum with it or against it?
- Sentiment — the mood around the name. What are the headlines and the market's tone saying, filtered from noise into a single read.
- Macro — how the stock fits the broader regime. A name can be strong on its own but swimming against a risk-off tide; this score captures that fit (or friction) with the wider market.
- Catalyst — the near-term trigger. Earnings, a product event, a sector rotation — whatever might actually move the stock in the sessions just ahead.
The AlphaScore is simply the average of those four, scaled to 100. Deliberately, no single read dominates — a great chart with no catalyst and poor macro fit won't run away with the number, and neither will a hot headline on a technically broken stock. The four have to broadly agree for a score to be high. Here's the arithmetic on the highest-scoring read in our sealed record — not a made-up illustration, the actual row from the database:
| Read | What it asks | AMZN, Jul 31 |
|---|---|---|
| Technical | Is the trend and price setup with the stock? | 7 / 10 |
| Sentiment | What are headlines and market tone saying? | 9 / 10 |
| Macro | Does it fit the wider regime, or fight it? | 8 / 10 |
| Catalyst | Is there a near-term trigger that could move it? | 9 / 10 |
| AlphaScore | (7 + 9 + 8 + 9) ÷ 4 × 10 | 82.5 |
What 1,300+ sealed reads actually look like
A scoring system's honesty shows up in its distribution. If every stock the engine liked scored 85, the number would be marketing, not measurement. So here is the full histogram of every AlphaScore published since we began storing the four sub-reads — each one sealed before the open, none edited after:
The shape is the point. The median read sits in the low 50s — meaning the typical stock on a typical day looks, honestly, average. The bulk of scores live between 45 and 65, and the tails are earned: a single read above 80 in the entire record. That's what "the four reads have to agree" looks like in practice. When you see a 75+ on one of our morning reports, it's rare by construction, not by marketing.
Where each read lives on its scale
Zooming from the composite into its parts: each of the four reads is anchored near the middle of its 1–10 scale, with the 10th–90th percentile band spanning just a few points. Catalyst runs slightly hottest — news flow usually gives the engine something to point to — while macro runs coolest, because most stocks on most days neither fit nor fight the regime strongly:
Why equal weights, and not something cleverer? We could tune the weights — say, 40% technical, 30% catalyst — until some backtest looked great. We deliberately don't. Every tuned weight is a degree of freedom, and degrees of freedom are how scoring systems quietly overfit the past. Equal weights are a guardrail: fewer knobs, fewer ways to fool ourselves. The same philosophy runs through the whole engine — simple, inspectable rules over optimized ones.
Two things the AlphaScore is not
First, it isn't a "buy signal" — it's a read on the setup, one input among several, and it says nothing about how much (if any) of the stock the engine actually holds. Position size is a separate calculation driven by conviction and volatility, which is why a name can carry a big weight with a middling score, or a small weight with a high one. Second, it isn't a prediction of the price. It's a structured snapshot of how four dimensions line up today — and as the distribution above shows, most days most stocks simply don't line up.
Why we show the math at all. Most scoring models ask you to trust a number you can't inspect. We'd rather you see the recipe and the output: four honest reads, equally weighted, averaged, published, sealed. You can watch each name's score against what actually happens over time and judge for yourself whether the reasoning holds. That's the whole point of scoring in public.
1. The AlphaScore is four equally-weighted 1–10 reads — technical, sentiment, macro, catalyst — averaged and scaled to 100. The record high needed all four to fire at once.
2. The sealed distribution is the honesty check: median in the low 50s, one read above 80 in 1,300+. No grade inflation — high scores are rare by construction.
3. A score is a setup read, not a buy signal and not a price prediction — position size is a separate, volatility-aware calculation.
Informational and entertainment content about how the Alphixir engine works — not investment advice.