 
Baseball Primer Newsblog — The Best News Links from the Baseball Newsstand Sunday, January 26, 2014Clay Davenport: First Projections for 2014
Thanks to Carlos. 
BookmarksYou must be logged in to view your Bookmarks. Hot TopicsNewsblog: Josh Lueke Is A Rapist, You Say? Keep Saying It.
(57  11:13am, Apr 23) Last: The Id of SugarBear Blanks Newsblog: The rise and fall of Ike Davis' New York Mets  Capital New York (12  11:12am, Apr 23) Last: Captain Supporter Newsblog: OTP April 2014: BurstNET Sued for Not Making Equipment Lease Payments (2107  11:12am, Apr 23) Last: Slivers of Maranville descends into chaos (SdeB) Newsblog: OT: The NHL is finally back thread, part 2 (193  11:08am, Apr 23) Last: Pooty Lederhosen Newsblog: OT: NBA Monthly Thread  April 2014 (473  11:07am, Apr 23) Last: jmurph Newsblog: ESPN: W. P. Kinsella: Where It Began: “Shoeless Joe” (95  11:07am, Apr 23) Last: BDC Newsblog: J.R. Gamble: Albert Pujols' 500Homer Chase Is A Bore, But That's Baseball's Fault (41  11:06am, Apr 23) Last: Ron J2 Newsblog: Primer Dugout (and link of the day) 4232014 (3  11:06am, Apr 23) Last: Harveys Wallbangers Newsblog: OMNICHATTER for 4/23/2014 (7  11:05am, Apr 23) Last: Rickey! In a van on 95 south... Newsblog: Cameron: Numbers don't lie: The decline of Pujols is stunning (219  11:03am, Apr 23) Last: Ray (RDP) Newsblog: OT: The Soccer Thread March, 2014 (1057  10:31am, Apr 23) Last: Shooty Survived the Shutdown of '14! Newsblog: Royals G.M. Dayton Moore believes hitting will come around (14  10:17am, Apr 23) Last: JE (Jason Epstein) Newsblog: Daniel Bryan's 'YES!' chant has spread to the Pirates' dugout (178  9:11am, Apr 23) Last: You Know Nothing JT Snow (YR) Newsblog: High and outside (8  8:55am, Apr 23) Last: RMc's desperate, often sordid world Newsblog: OMNICHATTER for APRIL 22, 2014 (92  8:23am, Apr 23) Last: Sunday silence 



Page rendered in 0.5321 seconds 
Reader Comments and Retorts
Go to end of page
Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.
1. Steve Balboni's Personal Trainer Posted: January 26, 2014 at 08:54 PM (#4646565)442/636
393/606
448/586
Some other projections:
Steamer: 418/594
Oliver: 413/592
ZiPS: 404/581
Maybe all of those are wrong and this is right, but that seems a significant outlier forecast and I'd be interested in hearing why. Rinse and repeat for the Vottos, Tulos, etc of the world.
These projections are regressed to the mean. The average disparity between the best team and the worst team for each simulated season is going to be greater than the disparity between the best average projection and the worst average projection.
I hope I explained that well.
I have to imagine that Chris Davis' projection is closer to 1.5 WAR than 6.
In fairness, Keith Law says they were never good to begin with.
Projections will always have a tighter range because they are mean projections. We expect 3 teams, on average, to perform to their 90th percentile, 6 to 80th or better, etc, but we don't know which ones they are yet.
loathe as I am to agree with Law, I kind agree with your simplified take on law's take. They were never that good, maybe like 85 wins good, you can't win a 80% of your one run games very often, ask the 2005 White Sox. Doesn't mean we shouldn't appreciate it when it happens, but it's not repeatable.
By a full three games.
As big a mess as the Cubs were when Thed took over, and as much as 90 losses seems near certain... It's awfully hard for me to see how this regime gets anything more than one more 90 loss season. I think I'm being more patient than most  I like the farm system, I think we're starting to see some real depth, and if Rizzo/Castro can hopefully rebound, the MLB cupboard isn't wholly bare  but really, you can't have more than 4 years of complete futility.
The Marlins had 4 starters last year who made 17 or more starts with ERA+ of better than 100 who are younger than 25 years old. They also have Giancarlo Stanton. You can only project so badly when you should have a solid rotation and a young superstar hitter.
That's not how projections (or basic probability) work. They're supposed to have a tighter spread because they represent the mean projection for each team. Those projections aren't saying that 91 wins will lead MLB, only that there's no team that has an *average expectation* of more than 91 wins. Obviously, some teams will perform to levels they only have a 10% or 20% chance of reaching (or falling to).
If teams were coin flips, the mean projection for every team would be 81 wins. But that's not the same as saying that 81 wins will lead the league because on average, you'd expect around 92 wins to be the average leaguebest in a league of 162 coin flips and 30 teams.
People are statistically illiterate #######.
That seems crazy to me. I'll be stunned if they fall below 70 and really think 85 is as likely as 75.
Consistent performance? Their pitching was a train wreck last year!
Hmmm, that seems to hurt the credibility of these projections quite a bit.
Didn't they significantly underperform last year?
I didn't do it for the National league teams but if you total up the runs scored by the American League teams it comes out to exactly the same 10,525 runs that were scored by AL teams in 2013. It doesn't look like it's a reduced environment, just a more even environment. The same reasons for the more smoothed out win/loss projections that Dan lays out in #26 probably explain that.
Someone from that top 34 teams is going to score over 800 runs and someone from the group of Minnesota, Houston and Kansas City is probably going to score around 625.
Someone from that top 34 teams is going to score over 800 runs and someone from the group of Minnesota, Houston and Kansas City is probably going to score around 625.
Right, you have to remember that every year ~33% of teams will exceed or fall short of expectations by 1 SD.
The only stats class I took was Intro to Stats, so bear with me, but if you ran enough simulations, would the results eventually be that every team finished 8181? Or do they not regress like that?
If you start with all teams being equal 8181 teams, then just by chance some of them will win 90 games, others will win 70 games.
While acknowledging this is right, why do ZiPS and the other projection systems in #2 spit out consistently higher results? Obviously Dan is well respected around here, and justifiably so, yet it seems there is a materially different overriding approach in Clay's* projections.
*The systems agree on some players and has some normal variation. That said, the number of players that are projected lower in Clay's far outweighs the number projected higher in Clay's, particularly for established players. This isn't 'Clay doesn't like Miguel Cabrera/Tulo/Votto/etc', it's 'Clay is regressing Miguel Cabrera/Tulo/Votto/etc. more than anyone else'. I find the outlier approach interesting and worth exploring, especially for someone with Clay's track record, but barring a better explanation I'd rather just use other sources. Of course, a Dan/Clay debate would be the best outcome. Get on it, boys.
As you run more simulation, the average of each team will get closer to their true mean (in the case of a coin that would be 81).
That's basically the progress by which these projections are arrived at. They take the average of thousands of simulations, to get the most likely outcome. But you have to remember that the most likely outcome, isn't very likely at all. If you looked at each individual simulation however, you would probably find at least one 95+ team in most. It's just evened out by the time that team finished with 85 in another sim.
Edited for crappy grammar.
I think we both read the question a bit differently by the way.
Looking at it Sean's way, there is about a 6.26% chance of any individual "team" finishing at exactly 81. Not accounting for interdependency of results, the odds of that happening 30 times in a row would be 0.000079%, or a bit less than one in a million.
Yeah, I think so. Looking at in another way, the more games you play in your sim, the less spread in the results you'll have. Was that what was meant in the original question?
With 162 games, an average team will have a +1 SD result of 87 wins, or .540 winning percentage.
Play 1 million games, then +1 SD will be a percentage of .5005. With 1 million trials, a team playing .506 ball (the equivalent of 82 wins in 162) will be 12 SD from the mean, which means it pretty much doesn't happen. (an average team playing 12 SD above the mean in 162 games would be 1575)
That was probably really unclear, so here's a thought experiment (and again, those who know this better can correct me). If you played the season a thousand times (actually playing the games, not calculations) without changing the teams at all (impossible, sure, but bear with me) then you wouldn't have to regress much at all because the observed value would be quite accurate. At that point the best team would have a lower average win total than the best team has in a given season, but it would have a greater average win total than even a very good projection.
I assumed it meant repeatedly simulate the season, and average the results, but I am not certain.
Looking at it Sean's way, there is about a 6.26% chance of any individual "team" finishing at exactly 81. Not accounting for interdependency of results, the odds of that happening 30 times in a row would be 0.000079%, or a bit less than one in a million.
Just to be obnoxious, 6.26% is binomial. There are a set number of wins in a 2430game season, so if you're not flipping the coin 2430 times, you want hypergeometric. So 6.37%.
I think SG does that in the RLYW blowouts.
Edit: Yes, he does. Looking back at last year's blowout is fun. Blue Jays at 29% for the division (Red Sox at 15), Angels at 40%, Nationals at 45%, Giants at 28%. Whoops!
And their 2nd order win percentage was .503, their 3rd order was .513. It's a mediocre team that got extremely lucky to win 91 games in '12. I think it is what it is.
I see the error, you got .00079% by taking .626 and raising to the 30th power. That's the equivalent of taking a likely event (.626 winning percentage is about 100 wins) such as the best team in baseball winning a single game. The best team in baseball winning 30 in a row? Very unlikely  less than one in a million. Take an unlikely event and make it happen 30 times in a row and the numbers get silly.
Sorry to be snarky, but it's a pet peeve of mine.
C'mon, man, undecillion!
Hmm, clearly I needed more coffee. It did seem to big at the time.
You must be Registered and Logged In to post comments.
<< Back to main