Market, fundamental & reference data
Stored prices, fundamentals, detailed valuation, technical metrics, news, catalysts and market context.
Data collectionSolid
Tutorial
What AlphaForgeReco is, and how to use every part of it
AlphaForgeReco is a recommendation-first asset engine and research workspace. It reads market data, fundamentals, technicals and structured news; produces ranked recommendations, factor-scored picks, watchlists, per-stock drill-ins and a daily morning report; and later measures whether those recommendations actually made money. The golden rule: every number is deterministic — computed by fixed rules from data. AI is only used to explain results, never to calculate returns or rank stocks. And when a credible number isn’t available (e.g. a proprietary fair-value model), it is omitted rather than faked. Below, every menu entry is explained in plain language.
Knowledge Center
Glossary · indicators · methodology — editable in AdminHow It Works
The deterministic recommendation engine, end to end.
Glossary
Every term used across the app, defined.
Technical Indicators
How each indicator is calculated, read and used.
AI Methodology
Exactly where AI is — and is not — used.
Morning Report
How the Morning Intelligence pipeline builds the brief.
Portfolio Tracking
The transaction ledger, deposit-neutral returns, and Decision Replay “what if” scenarios.
AI Workspace & Research Tools
Portfolio Watch/Review, Thesis Monitor, the Research Agent and Ask AlphaForgeReco.
Recommendation Tracking
Would you have been richer following AlphaForgeReco?
Administrator Guide
Running the jobs, data pipeline and content.
How the system thinks
A deterministic pipeline — each stage feeds the next.
1
Market Data
Prices, volume, market cap, ETF and crypto data, dividends and splits are pulled (keyless via Yahoo) and normalized into one price history. In-use assets refresh hourly; the whole universe daily; anything you open self-heals on the spot.
2
Fundamentals & Technicals
Deterministic metrics are computed per asset: growth, margins, ROE, valuation, plus RSI, moving averages, momentum, volatility, drawdown and trend — and the momentum/quality/growth/value/risk factor scores everything else ranks by.
3
News & Event Detection
Headlines are ingested and the News Intelligence layer classifies, clusters and scores them. Structured events (earnings, upgrades, filings) become Signals when they actually matter — breakouts, oversold RSI, volume spikes, rating changes.
4
Recommendations
Candidates are pre-scored deterministically, then a shortlist is ranked and explained. Each asset gets a rating, score, confidence, thesis and risks. The AI only orders and explains — it never invents a number.
5
Morning Report
Market state, sector/theme rotation, opportunities, risks, news intelligence and accuracy are assembled into one daily, immutable snapshot and a readable report. A model-agnostic AI writes the narrative only after the numbers are final.
6
Outcomes
Later, the system measures what actually happened to every recommendation across eight horizons versus a benchmark — the honest answer to “richer or poorer?”.
How AlphaForgeReco Reaches a Decision
One official path creates a recommendation. Other systems add context, attention, research or later validation.
Authority legend
Canonical decisionContext / explanationAttention / lifecycleExperimental signalProspective validationData collectionData-limitedAdmin / research
Solid, young-data, experimental and data-limited labels describe evidence maturity, not whether an investment is good or bad.
The official recommendation path
Only this blue, top-to-bottom path sets an Official Recommendation.
Eligibility & safety
Rejects stale, missing, future-dated or incomplete inputs before an asset can be scored.
Canonical decisionSolid
Factor evaluation
Calculates stored momentum, quality, growth, valuation and risk factors for eligible assets.
Canonical decisionSolid
Deterministic scoring
Applies the canonical factor model; no contextual or experimental signal changes this score.
Canonical decisionSolid
Shortlist, ranking & selection
Ranks deterministic candidates, applies diversification/selection, and preserves the distinction between a relative pick and an official rating.
Canonical decisionSolid
Daily recommendation publication
Persists the scheduled canonical run, recommendations and snapshots. This is the authority for an official current recommendation.
Canonical decisionSolid
Context around a decision
Useful evidence and risk context. These systems do not alter the canonical factor score.
Investment thesis lifecycle
Separately evaluates fundamental, price and time invalidation. Time expiry requests reassessment; it never automatically changes a recommendation to Sell.
Attention / lifecycleYoung data
UDI / Attention & decision delta
Qualifies, deduplicates and ranks material changes for review. An attention card is not automatically a recommendation.
Attention / lifecycleSolid
News, catalysts & market context
Provides dated evidence and upcoming events for explanation, monitoring and attention.
Context / explanationSolid
Risk overlay & regime context
Adds risk-aware interpretation and attention context; it is not a second scoring engine.
Context / explanationSolid
Research Memory
Durable, relevance-ranked historical context. It can qualify an explanation but cannot create or alter a recommendation.
Data-limitedData-limited
Investor Intelligence & reliability
Tracks sourced calls and their record over time. It is contextual evidence, not canonical scoring authority.
Context / explanationYoung data
Institutional / 13F evidence
Makes dated institutional filing evidence available as contextual research. Filing disclosure timing is preserved and it is not a recommendation input.
Context / explanationYoung data
Experimental signals
Can flag research or attention; they are not scoring inputs or standalone Buy calls.
Breakout / pre-breakout
Produces a separately labelled technical signal and never contributes to the canonical factor score.
Experimental signalExperimental
Structural Dislocation
Flags a thesis-price divergence with peer and benchmark context as an experimental research prompt, not a Buy.
Experimental signalExperimental
Smart Money & copyability
Collects verified on-chain activity and forward observations. Consensus, wallet skill and copyability remain explicitly immature.
Experimental signalData-limited
Evidence Fusion
Uses agreement or contradiction across independent signal families to create an attention event, never a factor-score input.
Experimental signalData-limited
After publication
Today, asset detail and Copilot explain or monitor the stored decision. Later outcomes measure it prospectively.
Today & Morning Briefing
Presents the canonical recommendation separately from prioritized changes, risks, catalysts and experimental evidence.
Context / explanationSolid
Stock detail
Shows one asset’s canonical rating, factors, thesis/risk state, evidence and outcome context without recomputing authority.
Context / explanationSolid
Copilot
Uses verified deterministic tools and bounded context to explain available evidence. The LLM cannot manufacture canonical state.
Context / explanationSolid
Recommendation outcomes
Prospectively measures published recommendations from their stored entry information; it does not feed back into a historical decision.
Prospective validationYoung data
What each rating means
Every recommendation gets exactly one rating, with a clear definition of success.
STRONG BUYHighest conviction long idea.
Counts as success when: Positive excess return vs the benchmark (or positive absolute return when there is no benchmark).
BUYA positive setup with lower conviction than strong buy.
Counts as success when: Same test as strong buy — beat the benchmark, or rise if no benchmark exists.
HOLDNo edge either way — neither add nor exit.
Counts as success when: A stable return within an acceptable band and no major drawdown over the horizon.
SELLThe setup has deteriorated — reduce or exit.
Counts as success when: The asset underperformed the benchmark or fell; the avoided loss is recorded.
STRONG SELLHighest conviction exit — avoid or close the position.
Counts as success when: The avoided loss is recorded when the asset underperformed or fell after the signal.
One rating per name, everywhere. A symbol's rating comes from a single canonical source — its AI Daily Pick rating, or a band from the stored composite score when it isn't a pick. The dashboard, morning brief, watchlists, screener and stock page all read that same value, so the same stock can never show STRONG BUY on one screen and AVOID on another.
Decisions refresh once per trading day. Ratings and the “Decision of the Day” are computed from the most recent market close. A given day's closing bar only exists after that day's market closes, so re-running the pipeline earlier — or on a weekend or holiday — reuses the last close and produces the same result. Expect the numbers to move once per trading day, after the close.
Official Recommendation vs Relative Pick vs Screen result vs Playbook Fit
Four different kinds of statement AlphaForgeReco can make about a security. Mixing them up is the single most common source of confusion in the app.
Official Recommendation
The engine's standing rating for one security — STRONG BUY / BUY / HOLD / SELL / STRONG SELL — produced by the canonical scoring pipeline. It exists on its own; it does not depend on what else is available today. This is what Recommendations, AI Daily/Monthly Picks, the stock page and Compare all show as "the rating", and it is the only one of these four that Playbooks can never move.
Relative Pick / ranking
Where a security sits against the rest of the universe right now (Daily Picks, Monthly Picks, Live Signals). A name can rank #1 today and still carry no BUY rating — a ranking answers "best of what's here", not "worth owning".
Screen result
A security that matched filters you chose on the Screener. It carries no opinion at all — appearing in a screen implies nothing about whether the engine likes it.
Playbook Fit
How well a security matches your chosen strategy's stated criteria (sector, factor weights, risk limits) — a personalization score, computed live from real factor data but purely additive. It appears alongside a rating; it never replaces or overrides one, and multiple active playbooks can each show a different Fit for the same stock.
The Discover page labels every tab with exactly one of these four kinds so you always know which question you're looking at the answer to.
What shows up on Today, and why
Today shows changes, not just ratings — a card can appear even when the Official Recommendation itself hasn't moved.
Recommendation
The Official Recommendation itself — STRONG BUY through STRONG SELL, unchanged unless the underlying score actually moved. See the vocabulary above.
Decision event
What actually changed — a new opportunity opening up, conviction strengthening or weakening, a breakout confirming, or a caution to reassess. This is why a stock can show up on Today with an unchanged HOLD rating: something material happened even though the rating itself hasn't moved (yet).
Warning
A caution, not an opportunity — e.g. rising risk on a holding, or a setup that looks tempting but has already moved too far to chase safely. Warnings are never buried under routine confirmations; they're treated as at least as important as a new opportunity.
Thesis change
If you've written a thesis in Thesis Monitor, a pillar strengthening, weakening or breaking shows up here too — the same evidence, read against the specific case you made for owning the position.
Confirmation (corroboration)
When two or more independent evidence sources — say, a technical breakout and a tracked investor's call — agree on the same stock, that agreement counts for more than either alone. AlphaForgeReco looks for genuinely independent agreement, not the same underlying fact counted twice.
New evidence
A single fresh observation — a technical setup, an on-chain wallet move, a tracked investor's post — that hasn't yet been confirmed by anything else. It can still show up, but it's weighted lighter than corroborated evidence until (if) something independent agrees with it.
Experimental signal
A signal family AlphaForgeReco tracks and measures but does not fold into the Official Recommendation while it's still proving itself. Breakouts and Potential Dislocation (below) are the current examples — labeled experimental both here and on their own pages, never presented as a recommendation. A high Breakout readiness score does not by itself mean BUY.
Potential Dislocation
AlphaForgeReco may flag cases where a price has fallen materially while the current structural thesis remains intact and current evidence does not show proportional fundamental deterioration — a divergence between how far the market has moved and how much has actually changed. This is not “buy the dip”: it never creates a BUY on its own, and a weakening/broken thesis or material negative news blocks it from qualifying at all. Where reliable, it also notes whether the decline looks specific to the company or shared with its sector/the broader market. Experimental and still prospectively collecting evidence — treat it as a research prompt to look closer, not a signal to act on.
How the ranking works
Today doesn't show everything, and it doesn't show things in the order they happened. Each item is ranked by how much it actually deserves your attention: how proven the underlying signal family has been historically, how many independent sources corroborate it, whether it's a warning (which are never suppressed), whether it touches something you hold or watch, and how fresh it is. A category that keeps repeating the same kind of story becomes progressively less prominent against everything else waiting for a slot — so one noisy signal family can't crowd out the rest of Today, though it can still win a slot when nothing else is genuinely material.
Evidence & signal quality
AlphaForgeReco does not treat every signal equally. Raw information (a news article, a wallet transaction, an investor's post) becomes evidence once it's extracted into something specific and checkable — never a vague guess. Evidence becomes a signal once it clears a minimum quality bar. A signal is then validated over time by what actually happened afterward — proven, promising, still experimental, or not holding up — before it's allowed much weight. Independent, corroborating signals reinforce a read; contradicting ones are flagged as a genuine disagreement rather than quietly averaged away. None of this touches the Official Recommendation directly — it's supporting context that sits alongside it, on Today and on the stock page.
Smart Money / Crypto Intelligence & Investor Intelligence
Two evidence sources run in the background today. Smart Money & Crypto Intelligence watches on-chain wallet activity for Bitcoin, Ethereum, Solana and XRP — exchange inflows/outflows, dormant-wallet reactivations, and staking/delegation moves — each chain kept separate, since evidence on one chain (say, ETH wallets accumulating) does not automatically say anything about another (say, SOL). It is an internal research signal family without its own dedicated page yet. A wallet moving to or from an exchange is not by itself proof of a sale or purchase — it is watched, weighted by how independently corroborated it is across separate wallets and how verified the activity is, and a single wallet movement is never treated as proof a trade should be copied (execution delay, liquidity and fees all matter, and are explicitly not assumed away — “copy this wallet” style guidance is not currently offered, since there isn't yet reliable intraday execution-price data to support it honestly). X Investor Intelligence tracks public calls and commentary from a small, curated roster of investors and traders (added and reviewed by AlphaForgeReco, not a live automated feed), weighted by each source's own historical accuracy — including how much history exists behind that grade (a source with a long track record is weighted differently than one with only a handful of recent calls) and whether its accuracy has been trending better or worse lately, not just its lifetime average. In both cases: this is evidence AlphaForgeReco considers, never AlphaForgeReco's own conclusion — an external investor's call is not the same thing as an AlphaForgeReco recommendation, and it is always shown as one input among several, never repackaged as if the engine itself reached that view.
Research Memory
AlphaForgeReco keeps a running record of durable research findings per security — what past evidence, theses and thesis changes said, and when. Ask AlphaForgeReco can recall that history when it is relevant to your question (“what did we think about this before?”), and will say plainly when current evidence has since moved past or contradicts an older finding. Older research is context for how the picture has evolved, not a standing view — current evidence and the Official Recommendation always take priority over anything in Research Memory, which is never treated as recommendation authority on its own.
Score vs confidence
Two different numbers you will see on every recommendation.
Score
A deterministic quality measure blending valuation, growth, quality, momentum and risk. Higher means a stronger overall setup. It drives the rating.
Confidence
How sure the engine is about that score, given data quality and signal agreement. Use it as a filter — e.g. only follow recommendations above a minimum confidence.
The Outcomes screen tells you whether score and confidence were actually predictive — i.e. whether higher values really led to better returns.
Outcome metrics & horizons
For every recommendation, measured at eight fixed horizons.
1d7d14d30d60d90d180d365d
| Entry / Exit price | The price on the recommendation date and at the end of each horizon. |
| Price return % | How much the asset moved from entry to the horizon exit. |
| Benchmark return % | How the chosen benchmark (e.g. QQQ) moved over the same window. |
| Excess return % | Price return minus benchmark return — the value the recommendation added. |
| Max gain % | The best point reached during the horizon, relative to entry. |
| Max drawdown % | The worst point reached during the horizon, relative to entry. |
| Time to peak | How many days it took to reach the best point. |
| Avoided loss % | For SELL / STRONG SELL: the loss you would have taken by holding. |
| Hit or miss | Whether the recommendation succeeded by its rating’s own definition. |
| Status | PENDING (horizon not reached), CALCULATED (priced), or MISSING_DATA (no prices). |
Simulation settings
Every rule you can set when replaying “what if I had followed the recommendations”.
| Buy ratings | Which ratings trigger a purchase — e.g. STRONG BUY and BUY. |
| Sell ratings | Which ratings force an exit — e.g. STRONG SELL and SELL. |
| Max positions | The most holdings the simulated portfolio can hold at once. |
| Position sizing | Equal weight (split evenly) or a fixed percentage per position. |
| Rebalance frequency | How often the strategy re-evaluates: daily, weekly, or monthly. |
| Min confidence / score | Skip recommendations below a confidence or score threshold. |
| Stop loss / take profit | Optional automatic exits when a position falls or rises by a set %. |
| Benchmark | The symbol the strategy is measured against to compute alpha. |
The result shows final value, total return, benchmark return, alpha (return minus benchmark), win rate, max drawdown, every trade, current holdings, and the best and worst recommendation.
Glossary
The terms used throughout the app, in one place.
| RSI | Relative Strength Index — a 0–100 momentum gauge. Below ~30 is “oversold” (possibly cheap), above ~70 is “overbought” (possibly extended). |
| Momentum | How strongly a price has been trending recently, measured over 1, 3 and 6 months. |
| Volatility | How much a price swings around. Higher volatility means higher risk. |
| Drawdown | The drop from a recent peak. Max drawdown is the worst peak-to-trough fall. |
| Benchmark | A reference index (e.g. QQQ) you compare performance against. |
| Alpha / Excess return | Return above the benchmark. Positive alpha means the pick or strategy added value. |
| Win rate | The share of recommendations (or trades) that succeeded. |
| Equity curve | A line of portfolio value over time. |
| Deterministic | Computed from data by fixed rules — same inputs always give the same result. Auditable, not guesswork. |
| Factor scores | The 0–100 scores AlphaForgeReco computes per asset — momentum, quality, growth, value and risk — combined into a composite. They drive the Screener, AI Daily/Monthly Picks and the drill-in. |
| Market regime | Whether the market is broadly risk-on (bull) or risk-off (bear), scored from breadth, momentum, volatility and risk appetite across the whole universe. |
| Sector / Theme rotation | Which sectors (e.g. Energy) or investment themes (e.g. AI, Semiconductors) are gaining or losing momentum strength — how money is rotating. |
| Market cap | A company’s size (price × shares). Tiers used here: Large-Cap ≥ $10B, Mid-Cap $2–10B, Small-Cap $0.3–2B. |
| Analyst target (fair value) | The consensus price target from Wall Street analysts (mean/high/low). Shown as a fair-value guide — it is their estimate, not our model. |
| Free cash flow (FCF) | Cash a company generates after capital spending — a core measure of financial strength. |
| P/E, P/B, P/S | Price relative to earnings, book value and sales — common valuation ratios. Lower can mean cheaper, but context matters. |
| 13F | A quarterly filing large investment managers must submit to the SEC disclosing their US equity holdings. Powers Guru Portfolios. |
| News Intelligence | The layer that classifies, clusters and scores news 0–100 for importance, tags sentiment, and flags which stories affect your holdings — all from stored data. |
| Position Analysis | A deterministic per-holding suggestion (Add / Hold / Trim / Review / Reduce-Exit) from the latest rating, momentum, P/L and your profile’s risk limits. |
| Shadow alpha | How a decision performed versus the alternative you passed on — did switching actually help? |
| Decision Alpha | One relative-performance number for a portfolio: your actual return minus the return of a shadow portfolio that never made your tracked decisions. Blank (not 0%) until at least one decision with a valid before/after basis has been recorded. |
| Decision Analytics | The absolute-dollar counterpart to Decision Alpha: which individual tracked buys and sells created or destroyed value, with a best and worst decision — the outcome, kept separate from whether the process behind it was sound. |
| Risk Budget | A portfolio’s concentration, diversification and volatility read at a glance, with a trend sparkline for each — plus the full detail (drawdown, correlation, sector/asset-class/currency exposure, risk limits) one click down. |
| Lazy fetch / cache | Data (a stale price, a company’s fundamentals) is fetched only when you open something that needs it, then cached — fast pages without hammering data sources. |
| Decision Score | A transparent 0–100 rating of one portfolio decision across seven weighted components (outcome, sizing, risk-adjusted, timing, AI alignment, process, calibration). Components without enough history are left blank, never invented. |
| Analysis Trust Score | A 0–100 grade of how much data-quality confidence backs a rating (High/Medium/Low) — informational only, it never influences the rating itself. |
| Investment thesis (pillar) | The written reason you hold a position, broken into pillars — specific, checkable conditions the system tracks evidence for over time, in Thesis Monitor. |
| Smart Alert | A system-detected change (rating, technical trigger, catalyst or portfolio-risk breach) surfaced automatically in Notifications — distinct from a user-defined price/RSI Alert. |
| As-of / end-of-day | The date a price or level was actually last recorded. Market Pulse, the stock chart and every price-derived figure show end-of-day closes, never an intraday quote — a number without its as-of date is treated as untrustworthy, not shown. |
| Playbook Fit | How well a security matches a chosen strategy’s stated criteria — see “Official Recommendation vs Relative Pick vs Screen result vs Playbook Fit” above. Additive only; never changes the Official Recommendation. |
| Decision event | What changed for a stock today — a new opportunity, strengthened or weakened conviction, a confirmed breakout, or a caution to reassess or avoid chasing. See “What shows up on Today, and why”. |
| Evidence Fusion | When two or more independent signal families (e.g. a breakout and a tracked investor’s call) agree or disagree on the same stock, and how much that agreement counts for. |
| Validation state | How proven a signal family has been historically, graded from what actually happened after past signals fired — proven, promising, still experimental, or not holding up. |
| Breakout | An experimental momentum signal — Pre-Breakout Watch, Breakout Triggered, Momentum Expansion, Extended (Do Not Chase), or Failed Breakout. Tracked and measured, never itself a recommendation; a high readiness score does not by itself mean BUY. |
| Potential Dislocation | An experimental signal for a security whose price has fallen materially while its structural thesis and fundamentals still look intact. Not "buy the dip" — it never creates a BUY on its own, and a weakening/broken thesis or material bad news blocks it outright. Still prospectively collecting evidence. |
| Research Memory | AlphaForgeReco’s running record of past research and thesis history per security, so Ask AlphaForgeReco can recall "what we thought before" versus now. Context for how the picture evolved — current evidence and the Official Recommendation always take priority over it. |
| Smart Money / Crypto Intelligence | On-chain wallet, exchange-flow, dormant-wallet and staking evidence for BTC, ETH, SOL and XRP, kept separate per chain. An internal research signal family — a wallet movement alone (including a move to/from an exchange) is never treated as proof a trade should be copied. |
| X Investor Intelligence | Public calls and commentary from a small, curated roster of investors (added and reviewed by AlphaForgeReco, not a live feed), weighted by each source’s own historical accuracy and how much track record backs it. Evidence AlphaForgeReco considers — never AlphaForgeReco’s own recommendation. |
Getting started
Six steps from sign-in to your first measured outcome.
- 1Sign in — local-first by default (no password needed unless an admin turns authentication on) — you land on Today, the daily home.
- 2Open Data Pipeline (or Assets) to seed the universe and refresh prices and metrics — or just let the scheduled jobs do it.
- 3Explore: skim Market Overview for the mood, Market Movers and AI Daily Picks for ideas, and click any ticker to open its full drill-in.
- 4Build a Watchlist of names you care about, and run Recommendations for ranked, rated, explained ideas.
- 5Generate a Morning Report for the daily snapshot, news intelligence and action center.
- 6Add a Portfolio to get Position Analysis and holdings news; set risk limits in your Investment Profile.
- 7Open Outcomes and click Calculate to score past recommendations — the honest “richer or poorer?” answer.