Hacker News Reader

Comment by streptomycin | Hacker News Reader

Even just with normal WAR https://www.baseball-reference.com/leaders/WAR_career.shtml he comes out at #38 all time.

I'm not sure summing up career WAR is the best way to rank players, though. It favors guys who played a long time, like Blyleven who played 22 seasons. Sandy Koufax played half as long but was undoubtedly better at his peak. Who was really "better" overall? You probably want some metric that combines career totals and peak production, but ultimately there is no universal way to come up with one definitive ranking, so we will debate it forever.

View on HN

This edit was added after my initial reply:

>...but ultimately there is no universal way to come up with one definitive ranking, so we will debate it forever.

WAR is a universal way to come up with a definitive ranking. It's a framework for measuring player contributions that is applied equally to all players. That's not to say it is the universal way, but the beauty of WAR is that the framework can be re-created to value or devalue player contributions as one sees fit and then use those criteria to measure all player contributions.

That's why the authors of this paper cite the two most common WAR frameworks (bWAR and fWAR). And it's why they've created their own version which incorporates a measure of the available talent pool at the time of each player's career.

>You probably want some metric that combines career totals and peak production.

How does WAR miss this goal? It credits year-to-year player contributions and gives more credit for better seasonal performances. Longevity alone doesn't ensure a higher WAR total since a negative WAR valuation is possible (and common).

Well there's infinite ways to combine peak and career totals, and "10 years of 1 WAR is equal to 1 year of 10 WAR" is certainly one of them.

So now you're indicating WAR does combine peak and career. That wasn't clear from your earlier comment. Likewise, it's not clear from this comment how you think the framework isn't doing well or what would be better. It's difficult to extrapolate how a measure of performance used to assess about 24,000 players could be improved when just two data points are cited, without detail, as problematic.

FWIW I've implemented a simplistic WAR in a baseball video game I wrote, so I am well aware of what it is :)

> Well there's infinite ways to combine peak and career totals, and "10 years of 1 WAR is equal to 1 year of 10 WAR" is certainly one of them.

That means that you can imagine many different ways of calculating overall career quality, even if you restrict yourself to just using WAR values. Summing up all the seasons is one of them. Others might give some additional value for exceptional seasons. For instance https://a.espncdn.com/magazine/1224NBAHOB.pdf as a simple example. I use a similar formula for Hall of Fame eligibility in my video game.

>Others might give some additional value for exceptional seasons.

This sounds like it would be a form of double counting. I'd be more inclined toward something like WAR/162 games (or whatever season length is in play).

Reply on news.ycombinator.com