Abstract Unconventional gas exploration and development is the new challenge that faces the Colombian energy industry, and understanding of the reservoirs is the first step to achieve the goal of increasing gas reserves. Coalbed Methane (CBM) has become an important resource given that Colombia holds the largest coal reserves in Latin America. Exploration of CBM began in 2008 with characterization of coal reservoirs, gas content, and storage capacity of four coal bearing formations in the main Colombian basins. In addition, the coals were assessed using techniques such as proximal and ultimate analysis, Rock-Eval pyrolysis, organic petrography, gas chromatography, isotopic analysis and desorption tests. From the four basins evaluated, this article will be focused in "Los Cuervos" Formation where four stratigraphic wells were drilled. Based on ultimate analysis and gas desorption measurements of cores, a mathematical model for estimating gas content was developed. The gas content showed excellent correlation and statistical significance with depth, total carbon and moisture percentage. 53 samples, classified as High Volatile Bituminous C using heating value from proximate analysis, were included in this study. Non parametric regression techniques based on statistical and optimization theory were applied to find the best correlation for predicting gas content. The equation set was generated with Grace98® software applying the ACE method. The non parametric model obtained explains the 92.5% variability of gas content and makes predictions with exactitude of 91.8%. The average percent deviation between experimental and calculated gas content was 7%, indicating good fitting between data and the proposed model. Based on the Anderson Darling test and residual plots, the errors generated from prediction were found to be randomly dispersed without any pattern. The obtained results showed that non parametric techniques were more efficient to predict gas content than conventional multiple regression. It was also found that there is not a kind of "universal correlation" but specific basin correlations since the gas content depends on geologic characteristics.