The ISR referred to in the more recent assessments of NCAA regional matchups is the Iterative Strength Rating developed by Boyd Nation, who has also kindly provided the pseudo-RPI numbers also used in those summaries.
You can find the full collection of ISRs and pseudo-RPIs linked from Boyd's College Baseball Ratings Page. For a direct link to the ISRs, click here.
For your convenience, I'll summarize the key features of the ISR here. If you want more detail, you should read the FAQ Boyd wrote that answers many common questions about these different ratings. Note that my summary tables quote the ISR rounded to an integer, both for my convenience and reflecting my opinion that the fractional part of the ISR is not significant. This view is backed up by the projected results table shown at the bottom of this page, where you can see that a difference of several points is needed before a team has a clear advantage over another team. If you want more accurate values for the ISRs, you should consult the official tables for a given season.
The ISR takes its name from the iterative process used to generate a self-consistent set of rankings. Boyd's algorithm ensures that the results of all games played to date are interlinked by the fact that your ranking depends on your opponent's ranking as well as whether you won or lost, and your opponent's ranking depends on how well they did against their opponents, and how good those opponents are, etc, etc. Most teams are no more than four or so "jumps" away from any other team in NCAA Division I baseball by the end of the season.
[The algorithm computes a team's ISR by taking the ISR for the opponent in each game, adding 25 for a win and subtracting 25 for a loss, adding these up for every game, and then taking the average. What makes the algorithm iterative is the need to adjust each of the ISRs until the result is self-consistent, that is, until the ISR calculated for each team from all of the other ISRs is the same as what you used to do the calculation.]
The advantage of the ISR over the RPI system is that the ISR uses an interlinked web that connects all of the teams. This additional connectivity compared to the RPI helps make up for the lack of inter-region play in baseball compared to, say, basketball. The ISR rewards wins over quality opponents, and you can also see that a loss to a highly ranked opponent can raise your ISR, and a win over a weak opponent will lower your ISR. Thus the ISR automatically builds in strength of schedule information.
Note well that the ISR cannot be reliable when only a few games have been played. No system can, given the vagaries of a game like baseball. By the end of the season the ISR does give a good estimate of what should be expected if both teams play as they have over the season.
The correlation between higher ISR and higher probability of winning can be quantified. Boyd produced the following prediction table from an analysis of the games played in the four seasons (1998 - 2001) he has been using it. The way you use the table is to calculate the difference in the ISRs of the two teams playing and round it up to an integer. This gives you the "gap" between the teams, and the table then tells you the percentage of the time that the higher ranked team will win.
ISR gap win % ------- ----- 0 51.6 1 53.4 2 55.6 3 59.9 4 61.8 5 65.4 6 65.8 7 67.8 8 70.1 9 71.8 10 73.0 11 76.5 12 77.4 13 79.9 14 82.1 15 86.6 16 83.7 17 87.4 18 88.7 19 87.9 20 90.1 21 92.4 22 94.5 23 94.3 24 92.3 25 96.0 26 96.1 27 95.1 28 96.1 29 95.6 30 97.9 31 98.1 32 100.0 33 98.2 34 100.0 35-44 100.0 46 100.0 47 100.0 ================
Thanks to Boyd for permission to include these numbers from his FAQ (which still has the 1998-1999 ones as I am writing this update).