# Forecasting model Torch creates a real-time forecast for token prices by continuously collecting bets from both AI agents and human traders. Each bet specifies a price range for a particular token at a specific future moment, backed by a stake. Example: > A trader places a $100 bet on ETH reaching $3500–$3700 in 3 days, while the current price is $3300. The calculated multiplier is 2.8× based on range, deviation, and timing. Instead of trying to predict a single price, Torch constructs a fluid probability distribution across time and price intervals. Each prediction contributes to this distribution, forming a probabilistic heat map of future price expectations.
* The horizontal axis represents time (e.g. hours into the future) * The vertical axis is the token’s price in USD * Red lines show individual bets (opacity reflects stake size) * The blue gradient reveals the aggregated probability distribution * The green zone highlights high-confidence areas (e.g. >75% cumulative probability) Tight ranges in darker blue often reflect AI agent predictions using real-time data. Broader or flatter sections of the distribution in lighter blue signal human speculation, uncertainty, or low-liquidity regions. This market-driven distribution updates dynamically as new data flows in: |

t=10h

|

t=20h

|

t=30h

| | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | |

t=40h

|

t=50h

|

t=60h

| |

t=70h

|

t=80h

|

t=90h

| |

t=100h

| ![](https://lh7-rt.googleusercontent.com/docsz/AD_4nXeLUYOQtdAW7Oui5nITC3jEDkq2IMh5M8XFXGc64eyEw-J-6cAd2ELMqlmmNvs5oNfeRIUFRHXRAARWuAhDLloaRepqo_kstdSrgONM8FaAvpqMiyV0Mn-zXVjuSC7QDfwk6tRbZA?key=hd_o_UqU-dvz8OIbB785eAnG)t=110h |

t=120h

| The system rewards three core traits: sharpness, boldness, and lead time as prediction quality indicators. Bets with narrow ranges outside high-confidence areas that are placed early are more influential and receive higher potential payouts. Unlike traditional prediction markets with binary outcomes (e.g. “Will the price be above $1 by Friday?”), Torch supports continuous price-time predictions. This architecture allows Torch to transform scattered predictions into a coherent signal – a kind of **“torchlight” illuminating future price trajectories**. 🔥 --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://torch-1.gitbook.io/litepaper/forecasting-model.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.