The 1976 determinate sentencing law did not specify how sentences should actually be reviewed, but it became apparent that some type of computerized statistical approach was appropriate. The board initially hired a small in-house staff to design the system and collect and code data, using a IV-Phase minicomputer connected to a State data center. The board collects at least 172 variables on every case and edits data extensively when entering them into the system. Overall, 1,400 to 2,000 cases are entered each month. The ASR calculates a sentence for each prisoner using Monte Carlo techniques. Sentences for each offender are calculated 10,000 times. If the simulation indicates that fewer than 10.5 percent of these 10,000 theoretical sentencings would have resulted in a sentence as high or higher than the actual sentence imposed, the case is identified as potentially disparate. A similar process identifies cases that may be disparately low. Such cases are then submitted to board staff for a secondary screening. The ASR reads both inmate-specific data and odds or probabilities distributions relating to the several sentencing decisions. Its output file contains a distribution of 10,000 sentences for each offender, statistics describing each distribution, and a comparison of the actual sentence imposed with the mean simulated sentence. The review process has reduced several prisoners' sentences, saving the State at least $312,000. When the system is perfected, it is expected to uncover at least 20 persons per month with disparate sentences. Charts, sample outputs, and two references are included. See NCJ-87442. (Author summary modified)