- Thread starter
- #1
TDs3nOut
Well-Known Member
I have had a fantasy team in an espn roto league for each of the past four years. After never finishing above fourth place before last year, I finally won my league last year. I am looking forward to this year’s draft and a chance to “defend my title” (LOL). To that end, I am trying to devise a strategy for identifying starting pitchers that are strong in the following four statistical categories: Ws, Ks, ERA, and WHIP.
The other day I mentioned to a poster on another forum that chasing the W stat for pitchers in a roto league has often led to bad decisions that end up hurting both my WHIP and ERA. That poster, who is a very experienced roto fantasy player posted that he tries to get SPs who have low WHIPs and high Ks/9, since he finds that by doing that Ws tend to take care of themselves. He also feels that such pitchers, in addition to getting Ws, tend to have low ERAs.
Accordingly, in order to investigate this proposed strategy, I constructed a data set for all MLB pitchers who last year made ten or more starts. I then used these data to estimate two regression models. In the first I model Ws as a linear function of both WHIP and Ks/9. While WHIP is highly significant in this model, Ks/9 is not. Likewise, in the second model, where ERA is estimated as a linear function of WHIP and Ks/9, I again find that WHIP is highly significant but Ks/9 is not.
Since Ks/9 was not a reliable predator of either Ws or ERA, I began thinking about how it seems to me that guys who get a lot of strikeouts also often throw a lot of pitches and aren’t able to stay in the game long enough to earn a W. Accordingly, I modified each of the two models above by substituting Ks for Ks/9.
In the first of these two new models, where Ws are modeled as a linear function of both WHIP and Ks, both WHIP and Ks are highly significant. In the second, where ERA is modeled as a linear function of these two variables, however, while WHIP is again highly significant, Ks is not.
So, of the four models that I estimated, here is the only one in which both independent variables are statistically significant:
A pitcher’s estimated wins = 6.69-3.56(his WHIP)+.06(his Ks)
A couple of pitchers whose performances fit this model almost perfectly last season are Mike Minor and Hiroki Kuroda. The two pitchers last year who most underperformed (in the sense that they won far fewer games than their WHIP and Ks predicted) are Cole Hammels and Tyson Ross. And the two pitchers who most outperformed the model (in the sense that they won far more games than their WHIP and Ks predicted) are Bartolo Colon and Jorge De La Rosa.
Finally, if anyone has made it this far in this post, feel free to post your thoughts on either this approach or alternative approaches to identifying starting pitchers in a roto league who can help your team in the Ws, Ks, ERA, and WHIP categories, without doing so at the expense of any of the others of these categories.
The other day I mentioned to a poster on another forum that chasing the W stat for pitchers in a roto league has often led to bad decisions that end up hurting both my WHIP and ERA. That poster, who is a very experienced roto fantasy player posted that he tries to get SPs who have low WHIPs and high Ks/9, since he finds that by doing that Ws tend to take care of themselves. He also feels that such pitchers, in addition to getting Ws, tend to have low ERAs.
Accordingly, in order to investigate this proposed strategy, I constructed a data set for all MLB pitchers who last year made ten or more starts. I then used these data to estimate two regression models. In the first I model Ws as a linear function of both WHIP and Ks/9. While WHIP is highly significant in this model, Ks/9 is not. Likewise, in the second model, where ERA is estimated as a linear function of WHIP and Ks/9, I again find that WHIP is highly significant but Ks/9 is not.
Since Ks/9 was not a reliable predator of either Ws or ERA, I began thinking about how it seems to me that guys who get a lot of strikeouts also often throw a lot of pitches and aren’t able to stay in the game long enough to earn a W. Accordingly, I modified each of the two models above by substituting Ks for Ks/9.
In the first of these two new models, where Ws are modeled as a linear function of both WHIP and Ks, both WHIP and Ks are highly significant. In the second, where ERA is modeled as a linear function of these two variables, however, while WHIP is again highly significant, Ks is not.
So, of the four models that I estimated, here is the only one in which both independent variables are statistically significant:
A pitcher’s estimated wins = 6.69-3.56(his WHIP)+.06(his Ks)
A couple of pitchers whose performances fit this model almost perfectly last season are Mike Minor and Hiroki Kuroda. The two pitchers last year who most underperformed (in the sense that they won far fewer games than their WHIP and Ks predicted) are Cole Hammels and Tyson Ross. And the two pitchers who most outperformed the model (in the sense that they won far more games than their WHIP and Ks predicted) are Bartolo Colon and Jorge De La Rosa.
Finally, if anyone has made it this far in this post, feel free to post your thoughts on either this approach or alternative approaches to identifying starting pitchers in a roto league who can help your team in the Ws, Ks, ERA, and WHIP categories, without doing so at the expense of any of the others of these categories.