A seismic-based reservoir properties estimation is implemented and tested in this work. The main goal in this work is to map oil saturated sands based on a sandshale oil field system. We consider petrophysical measurements as source of information to construct a conditional probability density function (PDF) for water saturated sand and a conditional PDF for oil saturated sand. Using these PDFs’ and seismic attributes from a reservoir cell, we compute the probability for water saturation given the attributes and the probability for oil saturation given the attributes. From these probabilities and following a Bayesian criterion we create an indicator for saturating fluid to this cell and the associated Bayes error. We analyze each reservoir cell to create a map for oil saturated indicator, water saturated indicator and the associated uncertainty. Several seismic attributes are analyzed in this work and using the maximum entropy measured from these PDFs’ we decide the most informative attributes and attribute pair to reduce uncertainty. In the current time, the methodology was successful tested in well log data. Our next step is to test the methodology with seismic attributes and apply the methodology in a real situation. Introduction The 4D seismic have a wide potential in the monitorating and administration of reservoirs, allowing to identify fluids contacts, no drained oil portions, injection fronts and permeability barriers. The appearance of preferential water paths creates no drained oil areas. The 4D seismic becomes an important tool to identify the reservoir areas that still contains oil and, consequently, to elaborate recovery strategies. Studies for fluid-saturation have been written frequently in the literature. Based on the Biot-Gassman theory, we study the effect of fluid saturation on seismic properties (Han and Batzle, 2004). Relationships between saturation and uncertainty, as predicted from Sengupta and Mavko (1998) and Mukerji et al. (2001), and principally methods based on Bayes criterion to reduce uncertainty (Gonzalez et al., 2002) were also of great interest for our work. A similar work, but more developed, can be founded in Mukerji, Mavko and Takahashi (2002). Only for visualization we study an application using Bayesian estimation theory through a workflow for analysis and prediction (Bachrach et al., 2004). Our goal is to map sand reservoirs filled out by oil, through the seismic attribute answers in different saturation conditions. We use well log data as source of information to build a conditional probability density function (PDF) for an oil filed out sand/shale reservoir situation and to build a PDF for a water filed out sand/shale reservoir situation. Starting from these PDFs’ and the seismic attributes estimated from seismic data associated to a reservoir cell we determine the most probable situation for the associated cell: ifiled oil reservoir cell or iifiled water reservoir cell. A Bayes criterion is applied to this decision. We follow the mathematical theory developed by Takahashi (2000) based in previous works, that involves methods to quantify the information through the probability theory and methods to estimate based in the information, were used thoroughly. Theoretical background To express quantitatively the “state of knowledge” of rocks properties; probability density functions (PDFs’) about these parameters, given a set of well log data representing in situ petrophysical measurement are computed. The “state of knowledge” (Takahashi, 2000), expressed by PDFs’, can describe how well we know the targets and how uncertain our targets are. In estimation problems, the PDFs’, supply a complete and quantitative description of the "state of knowledge " of each observed parameter, becoming a valuable information source. However, as it is waited of the own statistics, the measures don't have a perfect state of knowledge in other words, uncertainty zero. And it doesn't change in geophysical measures. This limitation can have cause in data acquisition, in the present noises, in the complexity of the nature, and in many other difficulties. Therefore, all and any form of minimizing the values uncertainty of properties in subsurface are been worth. In this point of view, it is noticed easily that a unique value they are not sufficient for estimates and that’s the reason because the use of PDFs became viable in this work. Initially, we worked with one-dimension PDFs’, computing PDFs’ for water saturation and oil saturation situations for each one of a series of petrophysical properties. We consider seismic velocities, Impedances, density, and others. To increase the amount of information we starts to work with pair of petrophysical properties, building twodimensional PDFs’ for oil saturation and water saturation situations. This work with a pair of properties was accomplished by the greatest reliability offered when we working with a pair.