Diamond Mind Projections & Predictions

[Crash Course]

http://sports.espn.go.com/mlb/spring2004/news/story?id=1774245

Whoa! The Cards to beat the Cubs and Astros?
White Sox to miss 1st by just one game?
How about the AL East race? Imagine if it just came down to one game?
World Champ Marlins to top at at 78 wins?
Texas won the West in one sim?

Any thoughts on these?

[Crash Course]

Just did a sort on their data:


#DIV #WC
Philadelphia 86.5 3.2
Oakland 56.3 1.5
Boston 54.5 42
San Francisco 50.5 3.5
St. Louis 45.5 19.3
New York AL 45.5 46.5
Minnesota 42 0
Chicago AL 34 0
San Diego 29.5 4.5
Houston 29.5 17.2
Seattle 25.8 2.5
Chicago NL 25 22.8
Anaheim 16.8 6.5
Cleveland 15.5 0
Atlanta 12 26
Los Angeles 10 0
Kansas City 8.5 0
Arizona 8 1
Colorado 2 0
Texas 1 0
Montreal 1 1
New York NL 0.5 0
Detroit 0 0
Tampa Bay 0 0
Baltimore 0 0
Toronto 0 1
Milwaukee 0 0
Pittsburgh 0 0
Cincinnati 0 0
Florida 0 1.5

22 teams out of 30 have a chance to finish 1st, according to the DM study. Does Bud have his hope and faith in place now?

[Crash Course]

What I reall like is when you add the DIV and WC columns. DM says that there's a 96.5% probability of Boston making the post-season - and that the Cubs are 47.8%. That's going to hurt all those predicting a Cubs-Sox WS.

[PianoMonkey]

Interesting, I guess, but just as worthless as all preseason predictions. 100 sims isn't nearly enough, and even if it was, there is no way to predict injuries and their effect on a team. You can simulate injuries, as they claim to have done, but they are random guesses.

Chaos theory states: no one can predict any reasonably complicated event with any accuracy whatsoever.

In the divisions with only one or two real contenders (like both easts) the projections of postseason appearaces may have more validity, though no one can be sure. But when it's a three- or four-team race, 100 sims is pathetically small to predict with any kind of precision (let alone accuracy). Exhibit A-the Royals averaged 79.4 wins to the Indians 79.3, but the Indians took the title twice as often. They say they don't know what that means. IT MEANS YOUR SAMPLE SIZE IS INSIGNIFICANT, MORONS!!

Aargh. I thought these people were GOOD at statistics. In any case, the records they predict the teams are much more trustworthy than drawing some inference from their "projected" postseason appearances, but that isn't saying much.

[Craig S.]

Interesting, I guess, but just as worthless as all preseason predictions.

That sums up my feelings pretty well. DM is a fun game, but I've never been a fan of their projections.

[sweaver]

Chaos theory states: no one can predict any reasonably complicated event with any accuracy whatsoever.

Well, that's not exactly what it states. It says that the more complicated the system, the more difficult it is to model and therefore to make a specific prediction. Butterfly in Kansas causing a Florida thunderstorm and all that.

Exhibit A-the Royals averaged 79.4 wins to the Indians 79.3, but the Indians took the title twice as often. They say they don't know what that means.
Largely, it means that the Indians players have more variability than the KC players, because they are younger and less "set in their ways."

I like the DM predictions, because they are usually based on more than a shot in the dark. Someone should take all the different predictions year-to-year and see who does the best.

[nyy26wc]

Largely, it means that the Indians players have more variability than the KC players, because they are younger and less "set in their ways."


I disagree with the more variability answer. I don't disagree the merits of the Indians having more variability than the Royals. But, DM's projection system doesn't take that into affect.

Let's take C.C. Sabathia as a random example. Over the course of his career, he's had 3, 4, 13 RSAA seasons.

He might revert back to the just a little bit average pitcher that he was during his 1st couple of years. He might continue to be the 13 RSAA pitcher he was in 2003. Or, his increase in 2003 might be an indication of things to come and he's even better in 2004.

But, DM's system doesn't take that variability into consideration. What they do is come up with a "projections disk" in which they project what each player is going to do and then, in each one of their 100 seasons, that same set of projections is used.

That's very different from sometimes running the simulation with the 2001-02 Sabathia level, sometimes with 2003 and sometimes with an even better version.

[ronh]

Someone should take all the different predictions year-to-year and see who does the best.

Prediction comparisons
http://www.diamond-mind.com/articles/tmpred03.htm