Forecast Methodology

How our AI crypto forecasts work

We combine live market data, token fundamentals, historical behaviour, and multi-model AI reasoning to produce transparent forecasts, consensus signals, and confidence scores for supported crypto assets.

Live market dataMulti-model analysisConsensus scoringRisk-aware forecasts

The process

From raw market data to a transparent forecast

Every forecast follows the same disciplined pipeline. Nothing is hand-picked — the same steps run for each supported asset.

  1. 01

    Collect live market data

    We pull live data for each supported token — current price, market cap, trading volume, 24h and 7-day movement, recent historical behaviour, and available token metadata — so every forecast starts from an accurate market picture.

  2. 02

    Build token context

    Each asset is framed with relevant context: its category and market structure, recent price action, volatility, trend direction, relative strength, and any available fundamentals. This context is what the models reason against.

  3. 03

    Query multiple AI models

    We ask several leading AI models for an independent view on the token. Each model assesses it across six horizons (End of Month, 3 Months, 6 Months, 1 Year, 3 Years and 10 Years) and returns a base case (its single most-likely price) with a min/max range per horizon, percentage upside/downside, sentiment, a bull case, a bear case, key assumptions, a confidence estimate, and a short rationale.

  4. 04

    Standardise the outputs

    Model responses are normalised into one consistent structure. That makes forecasts directly comparable across models and tokens, instead of a pile of free-form text in different formats.

  5. 05

    Generate AI consensus

    We take the consensus base case as the MEDIAN of the models' base cases, and an agreement band from the trimmed spread of those base cases (dropping the single highest and lowest) — so the range reflects where the models actually agree, not the sum of their hedges. We then blend sentiment, confidence and agreement into a clear directional view: bearish, neutral, bullish, or very bullish.

  6. 06

    Display forecasts transparently

    You see both the individual model forecasts and the aggregate consensus, side by side — so it's clear where models agree, where they diverge, and where uncertainty is high.

  7. 07

    Keep predictions refreshed

    Forecasts are cached and refreshed on a defined schedule and as new market data becomes available. Data-freshness indicators across the app show how recent each figure is.

The panel

What each AI model analyses

We query a panel of leading models and treat each as one independent analyst. No single model is treated as authoritative — the value is in comparing their views.

ChatGPT

OpenAI · GPT-5.5

Broad market reasoning, structured scenario analysis, and balanced bull/bear framing.

Typical strengths
General reasoning, scenario coverage
Forecast style
Structured and balanced

Claude

Anthropic · Claude Sonnet 4.6

Long-form reasoning, careful risk analysis, and explicit assumption tracking.

Typical strengths
Risk framing, assumption clarity
Forecast style
Measured and detailed

Gemini

Google DeepMind · Gemini 3.5 Flash

Market context synthesis and data-aware interpretation of recent trends.

Typical strengths
Context synthesis, trend reading
Forecast style
Context-led

Perplexity

Perplexity AI · Sonar Pro

Research-oriented reasoning and source-aware market context where available.

Typical strengths
Research framing, recency awareness
Forecast style
Evidence-oriented

Grok

xAI · Grok 4.3

Fast-moving market sentiment and awareness of prevailing narratives.

Typical strengths
Sentiment, narrative awareness
Forecast style
Narrative-aware

Mistral

Mistral AI · Mistral Large

Concise analytical forecasting and clear, structured directional views.

Typical strengths
Concision, directional clarity
Forecast style
Lean and analytical

Every model returns the same structured output

Regardless of provider, each response is normalised into the same fields so forecasts stay directly comparable:

Base case (most-likely price)Min/max price rangesUpside / downside rangeSentimentConfidenceBull / bear caseRationale

Forecast horizons

Six horizons, base case + range

Every supported asset is forecast across six horizons — End of Month, 3 Months, 6 Months, 1 Year, 3 Years and 10 Years. Each leads with a base case (the most-likely price) and a min/max range around it: nearer horizons lean on current conditions, while longer horizons widen into broader scenarios. We show that uncertainty rather than hiding it.

EOM

End of Month

A near-term range built on current price structure, momentum and sentiment — reading whether the prevailing trend is likely to continue or reverse.

Lower uncertainty
3M

3 Months

A short-term range that blends momentum with the early stages of the prevailing market cycle and liquidity conditions.

Lower uncertainty
6M

6 Months

A medium-term range weighing the broader cycle, adoption narratives and the prevailing regulatory and macro regime.

Moderate uncertainty
1Y

1 Year

A full-cycle range that leans on market structure, liquidity and adoption rather than short-term price action.

Moderate uncertainty
3Y

3 Years

A multi-year scenario range shaped by adoption, real utility, network effects, token economics and macro trends. Read as a scenario, not a precise target.

High uncertainty
10Y

10 Years

A long-horizon scenario range. The widest band by design — a directional thought experiment about where the asset could sit across a full decade.

Very high uncertainty

Consensus & confidence

Two signals, clearly separated

Consensus tells you which direction the panel leans. Confidence tells you how much agreement sits behind that lean. We keep them distinct so a strong direction with weak agreement never looks like a sure thing.

AI Consensus

A directional signal derived from the full panel of model outputs. It resolves to one of four ratings and is built from:

  • Directional agreement between models
  • Median base-case upside / downside
  • Sentiment alignment across the panel
  • Spread between individual forecasts
  • The forecast horizon in question
  • Each model's self-reported confidence

Confidence signal

A measure of how aligned and stable the model outputs are. Higher confidence means the panel is telling a consistent story — not that the forecast is guaranteed.

High confidence

Models are directionally aligned and forecast dispersion is relatively low.

Medium confidence

Models broadly agree, but with some variation in price targets or assumptions.

Low confidence

Models disagree meaningfully, or market conditions are too uncertain to call.

Example flow

How a single forecast comes together

A simplified walk-through using ETH. This is a generic illustration of the pipeline — not a live forecast.

01

ETH selected

Asset chosen

02

Live market data

Price, volume, history

03

Models queried

Independent views

04

Forecasts returned

Base case + range

05

Consensus calculated

Median base case + agreement

06

Outlook shown

Base case + agreement band

Why it's useful

An intelligence layer, not just a tracker

Most tools show you what a market did. We help you understand what a panel of AI models thinks could come next — and how much conviction sits behind it.

Many models, one view

See what multiple leading AI models think without manually prompting each one yourself.

Short and long term together

Compare near-term and long-range outlooks for an asset in a single, consistent view.

Agreement vs disagreement

Instantly see where models align and where they diverge, instead of reading them in isolation.

Consensus over noise

Separate a broad consensus from a single outlier prediction that may not be representative.

Spot signal and uncertainty

Quickly surface high-upside assets — and just as importantly, high-uncertainty ones.

A research input

Use forecasts as one input into your own research process, not as a trading signal.

What this is not

We built this to be useful and honest. Forecasts are a lens on the future, not a guarantee of it — so it's worth being clear about the limits.

  • This is not financial, investment, or trading advice.
  • AI forecasts are estimates and scenarios — they are not guaranteed.
  • Models can be wrong, and they can be confidently wrong.
  • Crypto markets are volatile and can move sharply against any forecast.
  • Predictions should be treated as research inputs, not decisions.
  • Always do your own research and consider your own circumstances.

CoinAugur is operated by COINAUGUR LTD. For how to treat these forecasts, see our disclaimer.