udt
updowntrends
How-To Guide

Understand AI stock forecasts, AI target & Trend Radar — from model quality to alerts

Learn step by step how to evaluate AI stock forecasts, use Trend Radar, set up watchlists and alerts, and validate individual assets with Trend Chart, AI target, model band, signal quality and forecast history.

All analyses and forecasts are generated automatically using AI and multiple technical indicators. They are for informational purposes only and do not constitute investment advice.

Model Quality

Metrics Overview – evaluate forecast quality objectively

The Metrics Overview is the starting point for objectively evaluating AI stock forecast quality. Before using an UP or DOWN signal, you should understand how well the model has historically worked for the market, asset and forecast horizon.

The key metrics are accuracy, balanced accuracy, AUC, Brier score and MCC. They show not only hit rates, but also separation, calibration and robustness of the technical AI signals.

Interpretation: Use metrics as a quality filter. Higher values indicate more robust historical forecast signals. Weaker values do not automatically mean a signal is wrong, but they suggest interpreting it more cautiously and checking other horizons.

Definitions are explained in the Technology FAQ.

Metrics Overview for AI stock forecast quality
Quality Over Time

Metrics Development – check forecast stability

Metrics Development shows how model quality evolves over time. This matters because markets change structure: trends, volatility, liquidity and market breadth can all affect forecast quality.

A stable or improving metric indicates consistent model performance. A sharp decline may point to a more difficult market regime where signals should be used more cautiously.

Interpretation: Always combine current signals with their quality trend. A strong signal is more valuable when the underlying model quality remains stable as well.

Metrics Development for AI forecast stability
Trend Radar

Trend Radar – identify strong setups faster

Trend Radar is the central overview for daily AI stock forecasts. It combines technical trend patterns, historical performance, UP probabilities, signal quality, AI target and model band into one compact table.

The left side of the mini chart shows the previous 60 trading days. The right side shows a conservatively derived AI target projection for the possible next 60 trading days. All forecast horizons are included: 1D, 5D, 20D and 60D.

The categories help you find setups more precisely: strong UP and DOWN signals, trend continuations, reversals, breakouts and notable model-band setups. This lets you screen markets faster and then validate individual assets on the detail page.

Trend Radar with AI target and model band
Personal Selection

Watchlist – track relevant setups

The watchlist helps you keep track of relevant stocks, indices, commodities and cryptocurrencies. Instead of scanning the full market every day, you can monitor your personal selection using AI forecasts, trend patterns, AI targets, model bands and classic signals.

You can switch between Trend Radar and detail table. Trend Radar shows the previous 60 trading days, the conservative AI target projection for the next 60 trading days and the model band. The detail table adds compact horizon signals and historical accuracy.

Interpretation: Use the watchlist for candidates you already consider relevant. This helps you spot trend changes, breakouts or changing model-band setups faster without searching for each asset again.

Watchlist with AI target and model band
Alerts

Alerts – monitor trend changes automatically

Alerts let you monitor important changes automatically. Instead of manually checking whether a signal has changed, you receive a notification when defined conditions are met.

Typical alert conditions include trend changes in historical performance or forecast reversals in AI signals. The alert email then shows the trend chart, AI target, model band and the most important quality metrics.

Interpretation: Alerts are especially useful for timing and risk management. They do not replace your own validation, but help you avoid missing relevant market moves and validate setups on the asset page in time.

Alerts with trend chart, AI target and model band
Asset Page

Trend Chart – combine history and AI target

The Trend Chart combines historical price development with the AI target. The left side shows the cumulative performance of the last 60 trading days, while the right side shows a possible path for the next 60 trading days based on forecast horizons, UP probabilities, signal quality and model-band width.

The dashed line is not a fixed price target. It shows a conservatively derived AI target based on model band, UP probability, signal quality and band width. The model band is a quantile-based range of possible returns, not a price target.

Interpretation: Setups are especially interesting when historical trend, signal quality and forecast direction align. Counter-trend moves can indicate reversals, but should always be checked against AI target, model band and quality metrics.

Asset Trend Chart with AI target and model band
Asset Page

Asset Predictions – signals, probabilities, AI target & model band

The enhanced Predictions table shows the most important forecast information per asset across 1D, 5D, 20D and 60D plus the weighted average. You see not only direction, but also UP probability, AI target, model band and signal quality.

The AI target is a conservatively derived model value based on the model band, UP probability, signal quality and band width. The model band additionally shows the quantile-based return range. Together, they help you evaluate direction, quality and range instead of reading signals in isolation.

Signal Quality combines several quality inputs into one compact score. A stronger signal is more useful when historical model quality, UP probability, AI target and model band are aligned.

The lower metrics such as ACC, BALACC, AUC, Brier and MCC additionally show how well the model has historically performed for this asset and horizon. This helps you interpret forecasts based on data instead of only looking at direction.

Asset Predictions with AI target, model band and Signal Quality
Asset Page

Prediction History – check signal stability

Prediction History shows how UP probabilities and signals have changed over time. It helps you see whether the model has provided a stable view or whether signals changed frequently.

Interpretation: Stable signal clusters suggest more robust forecasts. Frequent changes may point to sideways phases, higher volatility or a difficult market regime.

Prediction History
Asset Page

Prediction Quality: Signals & Hits – hits per horizon

This chart shows how many signals were generated per horizon and how many were confirmed in hindsight. It helps you quickly identify whether a specific horizon has historically been especially reliable for this asset.

Interpretation: A horizon with many signals and a high hit count is often more meaningful than a single strong signal without historical confirmation.

Signals & Hits
Asset Page

Prediction Quality: Correlation – forecast vs later performance

The correlation chart compares normalized forecast values with later price moves. It shows whether higher UP probabilities historically aligned with better subsequent returns.

Interpretation: A visible positive structure indicates genuine forecast quality. A diffuse cloud suggests weaker or regime-dependent signals.

Correlation
Articles

Understand AI signals, AI target and model band & apply them systematically

These guides explain the most important questions around AI stock forecasts, technical chart analysis, UP probabilities, AI target, model band, signal quality, Trend Radar, watchlists and alerts.

Workflow

1) The optimal workflow: from model quality to actionable setups

A good workflow does not start with a single UP or DOWN arrow. It starts with model quality: which assets and horizons have historically produced robust signals? Only then should you use Trend Radar, watchlists, alerts and the asset page to prioritize concrete setups.

In practice: check metrics first, then screen Trend Radar for strong signals, trend continuations, reversals, breakouts or model-band setups. Interesting candidates go on the watchlist. Critical conditions are monitored with alerts. Final validation happens on the asset page using Trend Chart, AI target, signal quality, model band and forecast history.

The advantage of this workflow: you reduce gut feeling and avoid jumping blindly into single signals. Instead, you combine historical model quality, technical trend patterns and current AI forecasts into a structured decision.

Signals

2) How to interpret UP probability, DOWN signals and UNCLEAR

The UP probability is the core value of the platform. It describes how likely a positive price move is over a defined forecast horizon. From this, three practical states are derived: UP, DOWN and UNCLEAR.

An UP signal means historically similar situations more often led to rising prices. A DOWN signal indicates elevated downside risk. UNCLEAR is intentionally not a weak signal but a quality feature: the model does not detect a sufficiently clear statistical edge.

The combination with signal quality, AI target and model band is especially important. A signal with high UP probability is more useful when historical model quality is solid and the model band is plausible as well.

Model Band

3) Understanding model band and AI target: why a range is more useful than a price target

The model band is not a fixed price target. It describes a quantile-based range of possible returns. The AI target is conservatively derived from the model band, UP probability, signal quality and band width.

A positive signal with a narrow model band and high signal quality should be interpreted differently from a very wide model band with high uncertainty. The AI target helps soften extreme upper quantiles and makes the forecast interpretation more conservative.

In the Trend Chart, the model band is visualized as a range around the AI target projection. This helps you see faster whether a forecast is narrow, wide, aggressive or defensive.

Watchlist & Alerts

4) Watchlists and alerts: monitor setups without screening everything daily

Many investors miss opportunities because they track too many markets at once. The watchlist reduces this complexity: you collect relevant candidates and monitor them daily with Trend Radar, detail table, AI target and model band.

Alerts go one step further. They monitor defined conditions such as trend changes, forecast reversals or new signal directions. This draws your attention to relevant changes without manually checking every chart.

Best practice: use the watchlist for strategically relevant assets and alerts for tactical triggers. This separates long-term monitoring from concrete timing.

Horizons

5) Using multiple forecast horizons: focus on structure, not noise

1D, 5D, 20D and 60D answer different questions. Short-term horizons react faster to momentum and sentiment. 20D and 60D are often more stable for medium-term trend decisions, position building and risk management.

A typical setup: the short-term horizon is volatile, but 20D and 60D remain positive. Then a pullback may be more of a timing issue than a true trend break. If longer horizons turn while short-term strength remains visible, risk increases.

Use multiple horizons not as contradictions but as time layers: short-term for timing, medium-term for structure, longer-term for risk and exposure.

Risk Management

6) Risk management with AI signals: dose positions more effectively

AI signals are strongest when they improve discipline. They do not replace a strategy, but they help dose risk more deliberately: larger exposure when quality and setup align, smaller exposure when signals are mixed or model quality is weak.

AI target, model band, signal quality and Trend Chart provide four perspectives: how large is the conservative model projection? How wide is the quantile-based range? How reliable was the signal historically? And does the forecast fit the current trend regime?

Especially around earnings, macro events or high volatility, signals should not be read as buy or sell recommendations but as data-driven support for position sizing, timing and hedging.

Definitions for probabilities, metrics, hit rate, walk-forward, out-of-sample, signal quality, AI target and model band are explained in the Technology FAQ.