-
tangotiger David, can you post a file with all the data used in the article?-
David Gassko Hey Tom, I'll glaldy send it to you.
-
-
pizzacutter David, it's not clear whether you're using absolute error or RMSE or something else. For example, you project two hitters to have a .800 OPS. One actually has a .775, the other an .825. You missed, on average, by 25 points. However, if you take the one's deviation (+25) and the other's (-25), and average them, you get zero! Your projections were perfect!-
David Gassko Sorry if that wasn't clear. I am using average absolute error.
-
-
ger8ry If I understand the bit about "adjusting the numbers so that averages were equal," then I'm not sure I like it. If there are only 2 ballplayers, and you project OPS of .900 and 1.000, and I project .725 and .775, and they actually come in at .700 and .800, then it sounds like you subtract .200 from each of your projections and report an average error of 0 for yours, .025 for mine. Is that really a good way to compare rating systems?-
David Gassko I think so. After all, all we care about is the relative ranking of players. What's the difference if someone gets closer to predicting the league average than someone else?-
ger8ry I see your point. But I can't help thinking that if I predict league average OPS to be .700, and you predict it to be .900, and it actually comes in at .725, then I have a firmer grip on reality than you do.-
tangotiger The reality is that all forecasting systems will presume some average around .750 to .770. It will be fairly tight. Does it really matter if one forecaster guesses 4.7 runs per 27 outs in the league while another forecasts 4.8, and the actual is 4.72? What counts is the relative ranking. Now, if you want to argue that someone forecasts 3.6 runs per game, and it comes it at 3.7 while everyone else forecasts 4.6 to 4.8, then fine. When that happens, then we'll have something to talk about. Otherwise, when we are talking about such minute differences in player-to-player forecasts, you can't give one forecaster an edge because he happened to guess the overall league average better than someone else. After all, this is all used for Fantasy purposes, and if you asked all the forecasters to use dollars instead, with a cap of 260$, then we wouldn't have this issue.
-
-
-
-
philosofool Nice work guys.
Now, if you'll just include a spreadsheet with a purchase of the 2009 THT season preview, you will get several of my dollars! (well, you'll probably get them anyway, but you'll be sure to get them, and a lot more, if you give me a spreadsheet.)
Any chance you could include accuracy of Marcels in a future study?
-
studes All THT's books come with spreadsheets. Folks who buy the 2009 Preview will get one, too.
-
Evaluating the 2008 Forecasting Systems
THE BOOK--Playing The Percentages In Baseball —
David compares THT, Chone, PECOTA, and ZiPS.
Heroes
Viva El Birdos —
... (subscription only) Unsurprisingly, they appear to favor their own PECOTA projections so take that for what you will. Here are the ones flimtotheflam found from The Hardball Times for this year. In any case, neither evaluated the James projections as part of their comparison so it doesn’t help us any even looking back a year. ...
Blue Jay Projections - Wave I - Hitters
Batter's Box Interactive Magazine —
... Disputes are (50+ point OPS spread) with Lind, Snider, (both over 70 points) and Rolen. Lets hope ZIPS is wrong for all 3. PECOTA doesn't appear to have come out yet, and isn't free anyways. CHONE, MINER, and THT aren't out yet also. The good news is ZIPS came in last for projecting hitters when compared to THT/CHONE/PECOTA but sadly we don't have how Bill James and Marcel did (source: http://www.hardballtimes.com/main/article/so-how-did-tht-projections-do/). ...
Bill James Likes Jason Kubel, But Not Denard Span
Twinkie Town —
... Usually, most of the major projection specialists (Bill James, Zips, Chone, Pecota, Marcels, THT) don't vary too much in terms of how they expect individual players to perform. This makes sense, since they all have their formulas and calculations for players, positions, player types, that help sculpt career arcs and have an effect on what numbers they expect. And in the end, they're all pretty accurate and ...
Pondering Nelson Cruz & The Texas Rangers' 2009 Lineup
Baseball Time in Arlington —
... Conversely, the CHONE projections measure up quite well from a statistical standpoint with the industry leader in player forecasting, Baseball Prospectus's PECOTA, and seem quite well-rooted in practicability and logic, and that's what I wanted to strive for here. ...
The Greatest Book Ever Written
Lookout Landing —
... Like the previous editions, this year's book features chapters for every team, with essays and player comments written by people who know more about their organizations than anybody who doesn't already work for said organizations. In addition, the book includes solid projections and a trio of essays at the end, all in an effort to make you better prepared for the season ahead. Seriously, it's an excellent book put together by a lot of excellent people, like David Gassko and, more sexily, me, and if you've been thinking about ways to trade money for smarts, this is a pretty good ...
Chone Smith can see the future
Angels blog —
... , by some measures the most accurate statistical forecaster for hitters. We talked about this season’s surprises (i.e. stuff I got wrong) and about the nature of statistical projections. chone-figgins1 me: I want to take about some of the mistakes I made in anticipating this season. Since I lean heavily on your projections, I figure I can just blame you. Sean Smith: I didn’t anticipate the injuries - I’m pretty sure I had Ervin set for a big year. me: Before I go to some players outperforming or underperforming projections, though, this is quoting you… ...
