Posted: Thu Jan 11, 2007 1:56 pm Post subject: 75.7 wins on average
Not a pretty sight there. I ran 250 simulated diamondback seasons using a couple assumptions.
Thats 75.7 wins on average. 92 wins max, 59 wins min. 57 of 250 were at or above 81 wins. Most common win total was 77 wins. More details and accuracy assumed later
Some of the assumptions going in:
1) I had to assume a starting lineup and rotation for every team in baseball including the dbacks. for the diamondbacks the lineup and rotation would be
Chris B. Young
Orlando Hudson
Eric Byrnes
Chad Tracy
Conor Jackson
Stephen Drew
Carlos Quentin
Chris Snyder
Pitcher
Tony Clark (PH1)
Alberto Callaspo (PH2)
Brandon Webb
Livan Hernandez
Doug Davis
Enrique Gonzalez
Randy Johnson
with RJ assumed the #5 pitcher only because I assumed he would be missing some games this year.
2) I assumed that aside from the closer, the teams all had the same bullpen and bullpen roles as last year. This will eventually be updated
3) I used last years data where needed but mostly used ZiPs projections where available. I augmented the Zips projections to have home/road/left/right splits as needed.
I used a modeling program that I wrote last season that does a decent job at predicting individual games. From there, I ran each game the dbacks are set to play 250 times. I then line up the first simulation for all 162 games and call that a season. I do that 250 times and I end up with 250 supposed simulated seasons.
The distribution is bell shaped. The odds of winning only 59 games is 1 in 250 just like the odds of winning 92 games. I think this distribution is pretty defendable. If I told you the min was 25 games and the max was 145, then I could see some speculation.
Ive projected the pitching matchups for all 162 games. i used the currently available Zips projections for those pitchers and hitters on each team. When Adam Eaton faces Randy Johnson in Arizona on May 08, I skew the phillies stats to reflect the fact they are facing a lefty at a ballpark with the various park factors chase has.
The simulation then tries to simulate every event in the game from first pitch to last out. Ground balls, fly balls, home runs, strike outs, errors, HBP, triples, doubles, singles as well as the baserunning choices that occur when say a guy gets a single with a man on first (does the runner advance to second or third). It will make pitching changes if the pitcher passes some criterion on IP and ER and will pinch hit for the pitcher if hes a reliever or a starter close to that criterion. It will choose different relievers (setup, mr, lr, or closer) depending on the game situation. It even has an option of generating a box score for each simulated game.
Oh, and this is all done in excel. Takes about 3 minutes to run 250 games.
Joined: 10 Aug 2006
Posts: 3242
Location: In front of my computer
Posted: Thu Jan 11, 2007 3:11 pm Post subject:
What is your performance baseline though.....IOW, what projections did you use as the basis for the simulation? Your own, ZIPS, some other projection service?
The problem that I have with any set of simulations is that there are too many significant variables to nail down a true 100 year probability.
We went over this after the 2004 season where the prediction was that we had 0.0% chance of winning our division. I couldn't buy into there being no chance that we had enough good luck and/or the others having enough bad luck, injuries or both to give us a chance.
I agree that with our team 92 wins seems high, but I would bet in 250 different seasons we'd beat that at least once.
i agree that a 59 win season is possible if you run enough seasons; that's your "worst case scenario" when shoewiz becomes my daddy.
fwiw, there's also this...
http://yankeefan.blogspot.com/
Scroll to SG's simulations run on Dec. 29th... Mind you, in those simulations, he did not have RJ in the rotation. Not sure if that was a plus or a minus
Anyhow, there's also a pretty interesting piece on the RJ trade
Joined: 10 Aug 2006
Posts: 3242
Location: In front of my computer
Posted: Thu Jan 11, 2007 3:38 pm Post subject:
Yuck....how the heck did the gnats come out that well inhis sim?
I think I am going to be ill.
As for D backs, I think they are dead on. I had 790 runs scored, they have 782. And if they stuck RJ in there, the runs allowed would probably come closer to my 768
What is your performance baseline though.....IOW, what projections did you use as the basis for the simulation? Your own, ZIPS, some other projection service?
I guess Im not understanding the question. I use the Zips stats that say in a given plate attempt, Chris young has a 11% chance at taking a walk, 8% chance at getting a single, 6% chance at a double, 1% chance at a triple, 5% chance at a homerun, 19% chance at a strikeout, some minor shot at getting HBP or reaching on an error depending on the team hes facing and then whatever left % chance at getting out by a gb or a fb. If he is facing say, Adam Eaton, I would say that any non-K out would be a fly out about 65% of the time and a ground out 35%.
I augment those stats though depending on ballpark, l/r splits and pitcher. In any plate appearance i assume Eaton has a 8% probability of issuing a walk, a 23% probability of a hit of any non-HR variety (I assume whether the hit is a single/double/triple is not pitcher dependant), a 18% chance at a K, a 3% chance at a HR, and the rest are outs.
I am going to try to attach a sample picture to this.... er.. pic is a little wide
http://img237.imageshack.us/img237/8877/bbpspicfl0.jpg
Last edited by rgndvo on Thu Jan 11, 2007 4:16 pm; edited 2 times in total
The problem that I have with any set of simulations is that there are too many significant variables to nail down a true 100 year probability.
We went over this after the 2004 season where the prediction was that we had 0.0% chance of winning our division. I couldn't buy into there being no chance that we had enough good luck and/or the others having enough bad luck, injuries or both to give us a chance.
I agree that with our team 92 wins seems high, but I would bet in 250 different seasons we'd beat that at least once.
Yes, this is just for conversations sake. It doesnt know if hudson is going to miss any games. It doesnt take into account that drew might hit 20% better than his projection but snyder might hit 20% worse.
I will say though that I went through this same process last year at this time on much more of a macro level (I basically took runs per game by team A, ERA of starter for team B and randomly generated a run total for each matchup) and I had pretty good results overall. these are listed below, with the preseason vegas line on o/u for each team.
AL East
NY Yankees - 93 predicted, 97 actual, 97.5, W
Boston - 88 predicted, 86 actual, 91.5, W
Baltimore - 77 predicted, 70 actual, 74.5, L
Toronto - 75 predicted, 87 actual, 87, P
Tampa Bay - 74 predicted, 61 actual, 67.5, L
AL Central
Cleveland - 86 predicted, 78 actual, 89.5, W
Chicago Sox - 81 predicted, 90 actual, 91.5, W
Detroit - 81 predicted, 95 actual, 78.5, W
Minnesota - 83 predicted, 96 actual, 82.5, W
Kansas City - 61 predicted, 62 actual, 63.5, W
AL West
Oakland - 91 predicted, 93 actual, 89.5, W
Texas - 78 predicted, 80 actual, 80.5, W
LA Angels - 79 predicted, 89 actual, 87.5, L
Seattle - 85 predicted, 78 actual, 75, W
NL East
NY Mets - 92 predicted, 97 actual, 90.5, W
Philadelphia - 84 predicted, 85 actual, 82.5, W
Atlanta - 83 predicted, 79 actual, 87.5, W
Washington - 72 predicted, 71 actual, 74.5, W
Florida - 69 predicted, 78 actual, 64.5, W
NL Central
St. Louis - 86 predicted, 83 actual, 92.5, W
Milwaukee - 80 predicted, 75 actual, 81.5, W
Chicago Cubs - 84 predicted, 66 actual, 85.5, W
Pittsburgh - 79 predicted, 67 actual, 75.5, L
Houston - 80 predicted, 82 actual, 82.5, W
Cincinnati - 80 predicted, 80 actual, 73.5, W
NL West
LA Dodgers - 83 predicted, 88 actual, 85, L
Arizona - 75 predicted, 76 actual, 73.5, W
San Francisco - 82 predicted, 76 actual, 84, W
Colorado - 77 predicted, 76 actual, 69.5, W
San Diego - 78 predicted, 88 actual, 78, P
No love from the vegas insider sports odds site. Below is their chart <pre RJ trade> of odds by team of winning the NL.
Code:
New York Mets 14/5
Los Angeles Dodgers 4/1
St. Louis Cardinals 9/2
Philadelphia Phillies 8/1
Atlanta Braves 14/1
Florida Marlins 14/1
San Diego Padres 14/1
Chicago Cubs 15/1
Houston Astros 15/1
San Francisco Giants 16/1
Cincinnati Reds 25/1
Milwaukee Brewers 25/1
Arizona Diamondbacks 35/1
Colorado Rockies 50/1
Pittsburgh Pirates 75/1
Washington Nationals 150/1
Joined: 10 Aug 2006
Posts: 161
Location: Tucson, Arizona
Posted: Thu Jan 11, 2007 6:53 pm Post subject:
I think what throws the projections out the window is that our team is full of guys with little to no expierence. Jackson, Quentin, Drew, Young, Callaspo, Montero just don't have enough ml expierence. If the offense clicks, this team will win.
_________________
-Zephon
I think what throws the projections out the window is that our team is full of guys with little to no expierence. Jackson, Quentin, Drew, Young, Callaspo, Montero just don't have enough ml expierence. If the offense clicks, this team will win.
I think you could say that with just about any team really. If they are older teams you never know how healthy they will be and there are always key guys changing teams or manager changes or whatnot that just cant be accounted for.
During the season though, you can predict better what is going to happen using say the first 30 games of data. The nice thing here too is that if a player gets injured you can predict that impact because the loss of a single player does not affect the players around them as it would in other sports.
I was playing with some of the numbers a little bit just to compare the NL West position by position. I used the Zips from the BBTF for the players I could find and had to really guess as to who was going to play which slot for some of the teams, but here it is.
Code:
1B OBP SLG OPS
AZ JACKSON .372 .429 .801
LA GARCIAPARRA .354 .469 .823
SD GONZALEZ .350 .475 .825
CO HELTON .432 .517 .949
SF KLESKO .364 .411 .775
2B OBP SLG OPS
AZ HUDSON .348 .440 .788
LA KENT .352 .473 .825
SD WALKER .344 .381 .725
CO MATSUI .330 .385 .715
SF DURHAM .348 .452 .800
SS OBP SLG OPS
AZ DREW .335 .467 .802
LA FURCAL .358 .431 .789
SD GREENE .325 .418 .743
CO BARMES .296 .374 .670
SF VIZQUEL .340 .349 .689
3B OBP SLG OPS
AZ TRACY .349 .465 .814
LA MUELLER .352 .390 .742
SD BRANYAN .340 .454 .794
CO CASTILLA .297 .398 .695
SF FELIZ .292 .410 .702
C OBP SLG OPS
AZ SNYDER .325 .392 .717
LA LIEBERTHAL .316 .469 .785
SD BARD .356 .424 .780
CO TORREALBA .297 .409 .706
SF MOLINA .317 .411 .728
LF OBP SLG OPS
AZ BYRNES .323 .462 .785
LA GONZALEZ .353 .435 .788
SD CRUZ .353 .381 .734
CO HOLLIDAY .383 .570 .953
SF BONDS .464 .539 1.003
CF OBP SLG OPS
AZ YOUNG .336 .525 .861
LA PIERRE .342 .385 .727
SD CAMERON .334 .446 .780
CO FINLEY .320 .394 .714
SF ORTMEIER .292 .369 .661
RF OBP SLG OPS
AZ QUENTIN .361 .472 .833
LA ETHIER .366 .463 .829
SD GILES .386 .423 .809
CO HAWPE .368 .493 .861
SF WINN .337 .413 .750
S1 ERA K9
AZ WEBB 3.85 6.95
LA SCHMIDT 3.59 7.61
SD PEAVY 3.23 8.82
CO FRANCIS 4.48 6.31
SF ZITO 4.01 6.88
S2 ERA K9
AZ RJ 3.71 7.91
LA LOWE 3.83 5.17
SD YOUNG 3.82 7.91
CO COOK 4.37 3.65
SF CAIN 4.01 8.11
S3 ERA K9
AZ DAVIS 4.23 7.76
LA PENNY 3.95 6.45
SD MADDUX 3.90 5.23
CO KIM 4.44 6.87
SF LOWRY 4.15 6.43
S4 ERA K9
AZ HERNANDEZ 5.14 5.53
LA WOLF 5.56 7.07
SD HENSLEY 4.14 6.64
CO BAUTISTA 5.25 6.95
SF MORRIS 4.31 5.34
S5 ERA K9
AZ GONZALEZ 4.96 5.66
LA TOMKO 4.47 5.73
SD WELLS 4.76 4.71
CO JIMINEZ 5.73 6.48
SF SANCHEZ 4.57 7.50
BTW I know ERA isn't a great pitching stat, but I didn't have the time to go too crazy on the list.
One thing I did find is that if you average the OPS for the teams, AZ comes out as #1
Joined: 12 Aug 2006
Posts: 23
Location: Gilbert AZ
Posted: Thu Jan 11, 2007 9:46 pm Post subject:
I have no problem with a range of 59 to 92 wins when we're talking about 250 simulated seasons.
I have a problem with season simulations when there's no accounting for personnel changes such as major injuries or surprise callups. When you don't account for the variation in personnel changes, you can get results like zero chance in a thousand seasons.
_________________
-Steve
I have no problem with a range of 59 to 92 wins when we're talking about 250 simulated seasons.
I have a problem with season simulations when there's no accounting for personnel changes such as major injuries or surprise callups. When you don't account for the variation in personnel changes, you can get results like zero chance in a thousand seasons.
I dont really agree. I think that the unforseen major events can affect the team in either a good or bad way and including those would only stretch the curve a little. I like to think the idea is more, if the individual components have the year that some of the better stat prognosticators think they will, they should end up with about 76 wins plus or minus say 8-10 wins.
I do think that if I could figure out how to read the expected error range on the predictions. Like when PECOTA projects say 175 innings for RJ, is that +/- 10 innings or +/- 25? If I could work in a random number generator that puts each player at having ups and downs from their projections, this would perhaps do the same widening of the bell youre talking about
As I mentioned, I'm not entirely familiar with the other team's anticipated roster and likely there are several mis-matches for positions. If you bring them to my attention, I'll update the numbers <if anyone cares to see them>.
Joined: 10 Aug 2006
Posts: 279
Location: Phoenix, AZ
Posted: Fri Jan 12, 2007 12:07 pm Post subject:
rgndvo wrote:
I will say though that I went through this same process last year at this time on much more of a macro level (I basically took runs per game by team A, ERA of starter for team B and randomly generated a run total for each matchup) and <b>I had pretty good results overall</b>.
I've never been a fan of season predictions this specific, because there are just too many variables to consider. And I respectfully have to disagree that your predictions for last year were "pretty good." Eleven of the teams were off by more than 8 games (which is, granted, an arbitrary cut off), and eight were off by 10 or more. If we look at an over/under split, 14 were off by more than 5, and only 16 were off by 5 or less. The average error is 6.23 games. I just don't see how this is "pretty good" accuracy.
I'm not trying to denigrate your efforts, rgndvo--I just don't think it's possible to accurately predict a season with reasonable accuracy because there are simply too many variables to consider, and not all of them can be defined statistically. Of course, I seem to recall that some people told Bill James something similar, so if you do develop an accurate system, I'll be the first to use it to place bets in Vegas, and I might even give you a cut!
It just seems that any prognosticator using last year's W/L records and a gut feeling about offseason transactions could be just as accurate.
No love from the vegas insider sports odds site. Below is their chart <pre RJ trade> of odds by team of winning the NL.
Code:
New York Mets 14/5
Los Angeles Dodgers 4/1
St. Louis Cardinals 9/2
Philadelphia Phillies 8/1
Atlanta Braves 14/1
Florida Marlins 14/1
San Diego Padres 14/1
Chicago Cubs 15/1
Houston Astros 15/1
San Francisco Giants 16/1
Cincinnati Reds 25/1
Milwaukee Brewers 25/1
Arizona Diamondbacks 35/1
Colorado Rockies 50/1
Pittsburgh Pirates 75/1
Washington Nationals 150/1
in vegas the week after next with work......might have to throw down on that!!!!!! IM not a big gambler, so confirm for me, if i did bet them and they make it i win 35 bucks for every dollar i put down?? what casino?
_________________
Not many D-Back fans in NJ.....
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
IM not a big gambler, so confirm for me, if i did bet them and they make it i win 35 bucks for every dollar i put down?? what casino?
