ComboCurve: Probit Plot Fit Statistics

ComboCurve: Probit Plot Fit Statistics

Probit Plot Chart 

When modelling a type curve, Probit plots can provide a deeper understanding of your rep (analog) wells. Once you creating a type curve you can visualize the Probit plot chart from the following dropdown menu. 


What is a Probit Plot 

Statistically, the outcome of multiplying 2 or more independent variables (such as b factor, Di, qi) results in a lognormal distribution. When calculating the EURs for a group of wells from DCA, the resulting EUR distribution becomes lognormal. Probit plots compare the EUR distribution of the type curve rep wells to a perfectly lognormal EUR distribution. In the plot, the x-axis represents the EUR volumes and the y-axis represents the Percentile (probability of occurring). The "Probit Model Fit" represents a perfectly lognormally distributed EUR whereas the green dots are the actual EURs of the rep wells. The closer the dots are to the black line, the more lognormal the rep wells EUR distribution is. The 3 horizontal grey lines, represent the P10/P50/P90 of a perfectly lognormal distribution. 

Probit Fit Statistics

ComboCurve provides additional statics in the Probit plot that can be accessed by toggling on the "Enable Probit Fit Statistics Legend" 


  1. X ^2 (pronounced Chi Square) represents the goodness of fit by comparing predicted vs. actuals. In our case predicted is a perfect lognormal distribution of EURs vs our actuals distribution of EURs
    1.  Lower the X^2 the better correlation between predicted and actuals, i.e. a X^2 of zero would be a perfect fit, meaning a perfect lognormal distribution of EURs
  2. The “model statistics” displayed in the legend are that of the perfect distribution of the data set using the mean and standard deviation of the raw data model distribution rather than the actual p-series values of the raw data
  3. The degrees of freedom (df) value is achieved by binning wells in groups of 4-8 to get a total number of categories then subtracting 1 for degrees of freedom
    1. Maximum degrees of freedom being 100, and min of 3
  4. Utilizing the X^2 value and df in a statistics X^2 table you can get a better idea of how well the raw data is fit to the perfect lognormal distribution in this case the higher the p-value, the better the fit where a p-value of 1 would be a perfect fit to the predicted data