Last time, I showed the "efficient frontier" for roster total SWP/G as a function of increasing franchise values. That article spurred a number of email inquiries, some of which I'll try to respond to here.

Q. Where is there a good source of data for Smallworld points per game?

A. I post tables which show SWP/G for all players on my personal website at http://rotoguru.com. In fact, if you use a browser that is JavaScript enabled, you can use my "Assimilator" program to add up your roster's SWP/G averages for you.

Q. What is the breakdown of expenditures on pitching versus hitting?

A. For the lower franchise values (below $70 million), the cost of pitching is around 50% of the total roster value. Above that level, the percentage allocated to pitchers grades down, comprising only about one-third of the total for rosters exceeding $115 million. For the majority of franchise values, the efficient pitching staff averages between $6-$7 million per pitcher.

Q. How many different players were represented on the efficient rosters?

A. There were a total of 43 different players who appeared on at least one roster. Of these, there were 14 different pitchers, and 29 different hitters. Two players appeared on the rosters in every single value range: Orlando Hernandez, and Robb Nen. Greg Maddux appeared on 17 out of 18 rosters. The most frequently appearing hitter was Ivan Rodriguez, who showed up on 15 out of 18 rosters.

Q. How do these rosters perform in price appreciation?

A. Only one repricing has occurred since I assembled those rosters. Of the 18 rosters, only 6 showed a net value gain in this latest repricing. Albert Belle, the biggest price gainer last week, was not on any of the rosters. The best of the roster gains was $480,000 and the worst was -$1,170,000. Looking at the data this way, it appears that value generation was poor.

However, when I look at the 43 individual players, 19 showed a price gain, 3 were unchanged, and 21 showed a decline. The largest price drops belonged to Sammy Sosa and Juan Gonzalez. The largest gains were for Kenny Rogers and Andy Ashby.

All in all, I'd say the data is inconclusive on this score.

Q. How stable are the SWP/G averages?

A. That's hard to answer in general. However, as I mentioned last time, it is likely that a roster chosen this way will tend to select players who have just been on a hot streak, so it seems likely that the SWP/G averages might be slightly high vs. the longer term likelihood. In fact, I checked the totals for these rosters ten days after the All Star break, and every single roster showed a decline in its total SWP/G, with the point declines ranging from 7-11 SWP/G. That seems pretty significant, averaging more than 0.5 SWP/G per player.

Drilling down a little deeper, of the 43 players, 11 showed a subsequent increase in SWP/G, 3 were virtually unchanged, and 29 showed a decline. The largest increase was for Kenny Rogers (who increased from 13.2 SWP/G to 14.2 SWP/G - and his recent price increase shows the reward), while the largest decrease was for El Duque, who dropped from 20.7 SWP/G to 18.4 SWP/G.

I also noted last week that the top worldwide rosters generally had averages that were around 15 SWP/G below the efficient frontier. It appears that maybe even the efficient rosters can't keep up to their lofty standards.

I haven't had a chance to rerun new efficient rosters to see how similar they would look using updated SWP/G averages and new prices. I'll save that analysis for a future column. I'd also like to take a backward looking view at roster efficiency - that is, what would have been the best producing rosters, now that we know what the latest production has been. And how closely do these "20/20 hindsight" rosters resemble the ones assembled without the benefit of a crystal ball. So stay tuned.