Alchemy Desk Weekly Outlook - Week of 2026-05-25
Monday weekly mover watch
This report looks for setups that may need several sessions to resolve. It is not a promise that Monday's move carries through Friday.
Source Integrity
As of 2026-05-25T00:00:53+00:00, the full board shows 156 live instruments and 612 fallback/simulated instruments. 10 live candidates in this article; fallback/simulated rows were excluded from the main board. The numbers below come from the current local snapshot and local research memory, with no external language API involved.
Candidate Board
- ABR.PRE: predicted 55.3% signal, 74.0% confidence, setup open-close long, source live
- AGRZ: predicted 52.1% signal, 44.6% confidence, setup reject, source live
- AXIL: predicted 50.3% signal, 49.0% confidence, setup reject, source live
- BIYA: predicted 49.3% signal, 39.4% confidence, setup reject, source live
- DCMT: predicted 43.6% signal, 71.7% confidence, setup open-close long, source live
- GMED: predicted 41.7% signal, 47.7% confidence, setup reject, source live
- QNC: predicted 39.4% signal, 47.5% confidence, setup reject, source live
- ABX: predicted 38.4% signal, 44.6% confidence, setup reject, source live
- BNTC: predicted 33.2% signal, 47.8% confidence, setup reject, source live
- AXPG: predicted 30.7% signal, 52.6% confidence, setup watch only, source live
Signal Table
| Symbol | Source | Predicted | Latest actual | Confidence | Relative movement | Risk band | Setup |
|---|---|---|---|---|---|---|---|
| ABR.PRE | live | +3.29% | pending | 74% | normal | low | open-close long |
| AGRZ | live | +1.93% | pending | 45% | normal | high | reject |
| AXIL | live | +1.96% | pending | 49% | normal | high | reject |
| BIYA | live | +1.23% | pending | 39% | normal | high | reject |
| DCMT | live | +2.09% | pending | 72% | normal | low | open-close long |
| GMED | live | +1.03% | pending | 48% | normal | high | reject |
| QNC | live | +0.94% | pending | 48% | normal | high | reject |
| ABX | live | +0.96% | pending | 45% | normal | high | reject |
Numerical Provenance
| Number family | Source | Meaning | Caution |
|---|---|---|---|
| Live instruments / fallback count | current snapshot feed integrity | How much of the board is live versus fallback or simulated | Live-ish Yahoo/public feed is still prototype-grade, not broker-grade |
| Expected move / confidence | proprietary local scoring stack | Open-to-close research estimate and confidence label | Prediction candidate, not trade instruction |
| Relative movement score | market-context adjustment layer | Whether a name is behaving unusually versus its backdrop | Can be distorted by gaps, stale quotes, or news |
| Risk score | proprietary fragility screen | Penalizes candidates with unstable context | Not a complete risk model |
| Research fragments | SQLite research spine | Prior lab lessons used to shape article language | Lab fragments do not promote a signal by themselves |
How To Read The Engine
The article is generated from structured records rather than free-form prediction copy: source flags, market regime, past-week outcomes, and research fragments all have to agree before the language gets stronger.
Current regime context is mixed, with breadth at 47.9%. That is the market weather around the candidate board, not a guarantee that any one name will move.
The research spine currently contributes these reusable lessons:
- failure: Corrected alpha returned -6.52% in alpaca_tradable versus best dumb baseline contrarian at 6.75%; verdict keep_in_lab. (lab)
- caution: Weakest adversarial control for alpaca_tradable was slippage_100bps at -51.58%, showing where the current signal is most fragile. (lab)
- caution: Feedback-look sharpening was hurt in alpaca_tradable: corrected alpha with feedback returned -6.52% versus -2.77% without feedback, a -3.75 point lift. (lab)
For today's board, the language focuses on ABR.PRE, AGRZ, AXIL, BIYA: these are candidates because the feature stack ranked them, not because the article is trying to manufacture conviction.
In plain English: the system is looking for evidence that a stock is moving differently from the market for a reason. The public report shows the conclusion layer; the weighting recipe stays inside the local research engine.
This Day In Finance
This section is a small market-memory intake. It gives the system a historical analog to chew on before it over-trusts today's tape.
- market history: No local finance-history item is pinned to this date yet. Lesson: The desk still records the empty slot: over time, this becomes a local calendar of market memory.
What The Desk Will Watch
The system will compare these candidates against actual movement, market breadth, source integrity, divergence, and whether a simple baseline would have done better.
The important question is not whether the list looks clever at the open. The important question is what the close teaches.