Well, the way they calculate it is this (I warn, it's kinda lengthy and can get confusing):
First, they take the major league average in each of the following categories:
Advancing a base on a groundout, advancing a base on a flyball, stolen bases/pickoffs, and advancing extra bases on a hit (1st to 3rd on a single, score from second on a single, score from first on a double).
So, for instance, let's say there were 10,000 instances in all of MLB of a runner on first base with no out when a hitter gets a single. If runners advances to 3rd base 2,500 times, then you can say that the average baserunner will go first-to-third on a single with no outs 25% of the time.
Then, they look at each player, and look at how many time they went first to third with nobody out. Let's say Ichiro was on first with no outs 100 times when there was a single, and advanced to 3rd base 30 times. That means he advanced 30% of the time, compared to the MLB average of 25%. Thus, Ichiro is above average in going first-to-third on a single with nobody out.
Comparitively, let's say Jason Giambi also has 100 first-to-third opportunities with no out, but he only advances 20 times. Thus, his percentage of 20% is well below the MLB average of 25%. So, Giambi would be a below-average first-to-third baserunner.
Then, the job is to quantify just how valuable Ichiro's additional first-to-third ability is to the team. For that, you use a run expectancy matrix, such as this one:
http://www.tangotiger.net/RE9902.html
What this tells you is the average number of runs that score in an inning, given a certain out/baserunners situation. (Sometimes no runs will score, sometimes exactly one run will score, sometimes exactly two runs will score, etc. The matrix gives you the average) So, looking at the table, you see that with nobody out and a runner on first base, a team is expected to score .953 runs. On a single, if the runner advances to second, the run expectancy increases to 1.573 (no outs, runners on first and second).
However, if the runner advances to third, the run expectancy increases to 1.904. The difference between the expected runs with 1st & 3rd no out and 1st & 2nd no out is .331 runs. Thus, every time a runner goes first to third, he is adding .331 runs to the average number of runs the team will score that inning.
So, in the example I gave before, the MLB average is 25 times and Ichiro advances 30 times. Thus, there were five additional times that Ichiro advances first-to-third. Each of those adds .331 runs to the team's expected run total, so the total contribution is 1.655 runs. Thus, a team with 24 exactly-average players plus Ichiro on it will score an average of 1.655 more runs than a team of 25 exactly-average players, just because of Ichiro's ability to go first-to-third.
Plus, you can factor in getting thrown out. Let's say in those 10,000 MLB first-to-third opportunities, a runner gets thrown out 100 times. Thus, an average runner will be thrown out 1% of the time. Now, let's say in those 100 opportunities for Ichiro, he gets thrown out twice. Thus, he gets thrown out one extra time than he should, according to the MLB average. If he had stayed on second that one extra time that he got thrown out, his team's average runs scored is 1.573. But, since he got thrown, the situation is now runner on first (the hitter that got the single) and one out. The expected runs in this situation is .573. Ichiro's ultra-aggressive baserunning cost the team exactly one run. (.573 minus 1.573). Add that to the 1.655 we got before, and Ichiro's total first-to-third, nobody out contribution is .655 additional runs
That's how you do it for one specific situation. Then, you do it for all the possible baserunning situations (for instance: score from first on a single, or advance to third on a flyball out, or score from third on a groundball out, etc.). And when you add up all the "additional" runs, you get the total number of additional runs that a player contributes via his baserunning over an exactly-average player.
There are additional things that they consider, such as where the ball is hit, the ball park (a ball that goes to the wall in spacious Petco Park is more likely to score a runner from first than a ball that goes to the wall in tiny Citizens Bank Park, for example), but that's beyond our realm of calculation.
So, adding all those "additional runs" up for the entire season, they saw that Chone Figgins added about 9.19 runs for the Angels last year. Thus, if the Angels replaced Figgins with a player of equal offensive and defensive prowess, but was only an average baserunner, it would have cost them 9.19 runs, on average, over the course of the season. Figgins was the MLB leader.
Victor Martinez was the worst. His baserunning cost the Indians 5.73 runs in 2006.
I also found the best and worst for the 2000-2005 period. The best: Carlos Beltran (25.44 total runs added...Jeter was second at 25.09). The worst: Jorge Posada (-20.39...Jim Thome was second at -19.16)