Offseason Moves and New Member
I've hated using ZiPS. There are just too many problems, the lack of splits being the biggest of them. A first glance at the 2010 set looks even worse. Apparently no one is capable of improving in the MLB. We're better off using the season disk than using ZiPS again, where EVERYONE gets worse, no matter the circumstance. Or we start busting our asses coming up with a formula to generate our own projections.
- Cardinals
- Posts: 7787
- Joined: Sat May 18, 2002 1:00 am
- Location: Manch Vegas, CT
- Name: John Paul Starkey
They had a brief partnership with Diamond Mind where they were (likely) going to do this. That partnership ended relatively quickly. Sadly.Brewers wrote:They will never release PECOTA as something that can so easily be transmitted without payment. I don't blame them either.
And I'm sure there would be plenty of people who would buy the PECOTA projection disc for extra money from BP. I'm one of them. Hopefully others would too. It could easily do as well if not better than the Diamond Mind ones I'd have to think.
12, 14, 15, 17, 22
- Cardinals
- Posts: 7787
- Joined: Sat May 18, 2002 1:00 am
- Location: Manch Vegas, CT
- Name: John Paul Starkey
Yeah, everybody seems to regress and not only that, but significantly.Cubs wrote:I've hated using ZiPS. There are just too many problems, the lack of splits being the biggest of them. A first glance at the 2010 set looks even worse. Apparently no one is capable of improving in the MLB. We're better off using the season disk than using ZiPS again, where EVERYONE gets worse, no matter the circumstance. Or we start busting our asses coming up with a formula to generate our own projections.
12, 14, 15, 17, 22
- Rangers
- Site Admin
- Posts: 3939
- Joined: Wed Feb 23, 2005 1:00 am
- Location: Prosper, TX
- Name: Brett Perryman
We asked last year whether there was any way to enter our own projections without manually entering every player individually, and my recollection was that the answer was no. If that's the case, and if we're going to try to create projections ourselves with splits and all, it's going to be a serious group project with a good chunk of the league needing to contribute on the entry. I'm all for it, and one thing that could make it possible is the proliferation of minor league defensive stats. But it's going to take some commitment from quite a few guys, not just 2-3 of us working our asses off for hundreds of hours.
Other than that, it's very doable.
Other than that, it's very doable.
- Brewers
- Posts: 1640
- Joined: Mon Apr 01, 2002 1:00 am
- Location: St. Johnsbury, VT
- Name: Jared Cloutier
Doing our own splits sounds like a way to split the league...it is impossible to make everyone happy no matter what you use. I would favor a third party system due to the arguments that are bound to come up...Outside of PECOTA, I don't know of a great way to handle anything outside of just raw numbers. It sounds like a mess that will have at least one good owner quit.
- Cardinals
- Posts: 7787
- Joined: Sat May 18, 2002 1:00 am
- Location: Manch Vegas, CT
- Name: John Paul Starkey
I tend to agree with this. Combining on our own system is great in theory but there will just be too many opinions conflicting with each other inevitably. I could see it being a major disaster spawning out of a good idea. I'd love to be proven wrong, but I'd just prefer that Pecota or DMB step to the plate this year so we don't have to cross this bridge.Brewers wrote:Doing our own splits sounds like a way to split the league...it is impossible to make everyone happy no matter what you use. I would favor a third party system due to the arguments that are bound to come up...Outside of PECOTA, I don't know of a great way to handle anything outside of just raw numbers. It sounds like a mess that will have at least one good owner quit.
12, 14, 15, 17, 22
- Orioles
- Posts: 3150
- Joined: Fri Jan 02, 2004 1:00 am
- Location: Glastonbury, CT
- Name: Dan Vacek
- Contact:
- Curses. I still hate teal a lot and Joe Robbie Stadium even more, but welcome Z to the ever-improving NLE, which could become a real bear of a division w/ two teams in the minidraft and a whole bunch of elite building blocks to choose from. Guess I shouldn't poop on my own unis too much, since I don't want the Fish to stink too bad in our first NLCS appearance.
- Agree w/ what JP + Ben said re: projs. My preference would be a 3rd party set of projections for 2010, with the caveat that if anyone works out their own complete and workable system at some point before the 2011 season we could then re-evaluate whether our other 3rd-party options are truly better.
- I think we should set a hierarchy for 2010 projs now (either the ExCo, or through a vote). My personal order of pref would be: 1. DMB 2. PECOTA 3. ZiPs. Unfortunately we don't know if we'll have 1 or 2. If we are left w/ ZiPs, do we want to apply some standard formula across the board to sort out splits, or would that just create the same issues we're trying to avoid by not using our own system of projs?
- Agree w/ what JP + Ben said re: projs. My preference would be a 3rd party set of projections for 2010, with the caveat that if anyone works out their own complete and workable system at some point before the 2011 season we could then re-evaluate whether our other 3rd-party options are truly better.
- I think we should set a hierarchy for 2010 projs now (either the ExCo, or through a vote). My personal order of pref would be: 1. DMB 2. PECOTA 3. ZiPs. Unfortunately we don't know if we'll have 1 or 2. If we are left w/ ZiPs, do we want to apply some standard formula across the board to sort out splits, or would that just create the same issues we're trying to avoid by not using our own system of projs?
2023 GM Totals: 1780 W - 1460 L | 0.549 wpct | 89-73 (avg 162 G record)
Once we agree on how the projections are formed, whatever formula we come up with, then the arguments over. Its just plugging in numbers. I agree, it will be a bit of a hassle and there will be some disagreements, but it is literally impossible for our system to be worse than ZiPS. Do we really want to put the league through another year of this crap? It's a huge step backwards. Those of us who play H2H a great deal can tell you that the lack of splits has dumbed down this season significantly.
It's not something that can wait. If we sit here and do nothing and hope that DMB or PECOTA will come to our rescue, then the odds are we use ZiPS again in 2010.
I know it would be a significant undertaking, but this is a significant change. A huge part of what our league is based on, the projection disk, is gone and its not likely to come back. We need to start thinking of long-term solutions. It could also be a way to improve the league overall, and some of the gripes we have with some of the unexplainable projections that pop up in the DMBs of old and in ZiPS now. I appreciate the work Dan does over at BBTF, but in the end, I can't stand ZiPS or the thought of leaving another IBC season at their mercy.
It's worth the debate and the hassle IMO, but if the league isn't behind then there's no point in trying. We should at least take another look at CHONE or CAIRO. Anything but those god awful ZiPS projections.
It's not something that can wait. If we sit here and do nothing and hope that DMB or PECOTA will come to our rescue, then the odds are we use ZiPS again in 2010.
I know it would be a significant undertaking, but this is a significant change. A huge part of what our league is based on, the projection disk, is gone and its not likely to come back. We need to start thinking of long-term solutions. It could also be a way to improve the league overall, and some of the gripes we have with some of the unexplainable projections that pop up in the DMBs of old and in ZiPS now. I appreciate the work Dan does over at BBTF, but in the end, I can't stand ZiPS or the thought of leaving another IBC season at their mercy.
It's worth the debate and the hassle IMO, but if the league isn't behind then there's no point in trying. We should at least take another look at CHONE or CAIRO. Anything but those god awful ZiPS projections.
- Brewers
- Posts: 1640
- Joined: Mon Apr 01, 2002 1:00 am
- Location: St. Johnsbury, VT
- Name: Jared Cloutier
Any situation you come up with spit out unexplained projections. It happened in DMB, it happens in Zips, and it happens in PECOTA. Let's not pretend we can rid the system of that. Numbers for guys with three years of major league data are fairly easy, but we would have all kinds of arguments on minor league equivalencies which is the main cause of whacko projections. Fielding would also be impossible to settle on, it is too much of a system. How do you decide what stats to use and what positions to qualify a guy for. Finally, the best part of DMB was that it wasn't just a numbers system. It was a guy who used numbers, but then honestly looked at second half improvements, young players getting better, and guys having hit the wall and made bold predictions. Any system without a human element will battle regression models that plague Zips.
- Cardinals
- Posts: 7787
- Joined: Sat May 18, 2002 1:00 am
- Location: Manch Vegas, CT
- Name: John Paul Starkey
Were the Zips projections really that bad? Aside from the inordinate amount of passed balls this year, there weren't too many ridiculous players such as Hiram Bocachica or Cy Mark DiFelice of years past. That said, I don't really like them too much. From what I've seen so far, he refuses to allow growth to players and instead seems to weigh them all down significantly. It's not perfect nor will it ever be.
It's worth debate, sure, we can sit here all winter and debate on a formula. 40/30/30 for the past few years, then Minor league equivs need to factor in somehow too, as well as weighing the players with minor league and major leaguer numbers, coming up with a regression and progression formula which would have to be age dependent and then the ever mysterious injury mystique, as well as calculating playing times, speed ratings, range ratings, error ratings, PB tendencies, WP tendencies, balk tendencies, catchers arms, OF's arms- which for the MLB may seem simple but the data isn't as easy to interpret at minor league levels for things like that. We would also need our system to calculate platoon splits since that's something that seems to be pretty important to our league.
While it's nice to think we could agree on something, I just don't see it as being feasible. A more likely scenario that we could do if we were to go this route as BP suggested to me would somehow combine projection systems and weigh them ourselves and add in past MLB years into it as well. Something like 40% PECOTA, 30% CHONE/ZIPS combo, 15% 2009, 10% 2008, 5% 2007 and then the ML Equivs. And that would require us to wait until February when PECOTA comes out. I also think that we are tremendously underestimating the amount of work and effort that would go into something like this too.
Again, I don't want to sound like I am against it, I think at the very least it's interesting discussion. What if we do all this work over the winter and then DMB or PECOTA does announce they'll release a disc? A lot of hard work for nothing in that case. I'll bet you at least $20 that DMB won't announce if they will make a projection disc until after they see how V10 sells - and that won't go on sale till December. It's interesting discussion and it would be intriguing to see if we could even agree on a formula to project players. We still get shit about how we handle injuries...would love to see us somehow agree on a projection system.
It's worth debate, sure, we can sit here all winter and debate on a formula. 40/30/30 for the past few years, then Minor league equivs need to factor in somehow too, as well as weighing the players with minor league and major leaguer numbers, coming up with a regression and progression formula which would have to be age dependent and then the ever mysterious injury mystique, as well as calculating playing times, speed ratings, range ratings, error ratings, PB tendencies, WP tendencies, balk tendencies, catchers arms, OF's arms- which for the MLB may seem simple but the data isn't as easy to interpret at minor league levels for things like that. We would also need our system to calculate platoon splits since that's something that seems to be pretty important to our league.
While it's nice to think we could agree on something, I just don't see it as being feasible. A more likely scenario that we could do if we were to go this route as BP suggested to me would somehow combine projection systems and weigh them ourselves and add in past MLB years into it as well. Something like 40% PECOTA, 30% CHONE/ZIPS combo, 15% 2009, 10% 2008, 5% 2007 and then the ML Equivs. And that would require us to wait until February when PECOTA comes out. I also think that we are tremendously underestimating the amount of work and effort that would go into something like this too.
Again, I don't want to sound like I am against it, I think at the very least it's interesting discussion. What if we do all this work over the winter and then DMB or PECOTA does announce they'll release a disc? A lot of hard work for nothing in that case. I'll bet you at least $20 that DMB won't announce if they will make a projection disc until after they see how V10 sells - and that won't go on sale till December. It's interesting discussion and it would be intriguing to see if we could even agree on a formula to project players. We still get shit about how we handle injuries...would love to see us somehow agree on a projection system.
12, 14, 15, 17, 22
My bad. I wasn't clear. This is what I was talking about, too. I remember the idea came up last season and that's what has been in my head ever since. I don't think we should start from scratch creating projections. I thought the debate would simply be how to properly combine all the projections and find an average. If I remember right, it was suggested that we could keep ZiPS defensive ratings, combine offensive projections, and try to come up with a way to factor splits.Pirates wrote:A more likely scenario that we could do if we were to go this route as BP suggested to me would somehow combine projection systems and weigh them ourselves and add in past MLB years into it as well. Something like 40% PECOTA, 30% CHONE/ZIPS combo, 15% 2009, 10% 2008, 5% 2007 and then the ML Equivs. And that would require us to wait until February when PECOTA comes out. I also think that we are tremendously underestimating the amount of work and effort that would go into something like this too.
But, yes, it will be a shitload of work. Worth it though, in my opinion. And yeah, I personally do think ZiPS is that bad. It's horrendous towards younger players. Those of us who really try to keep our rosters full of players under 30 really get screwed on those guys who have only been in the league only a year or two. They seem to always regress, and almost never improve, which is far from true to life.
Yes, DMB has some crazy Bocachica like projections sometimes, but that was part of the fun, and those crazy SIM monsters were a small price to pay for a system that worked better and was a better product in every measurable way.
Anyway, just my two cents.
- Brewers
- Posts: 1640
- Joined: Mon Apr 01, 2002 1:00 am
- Location: St. Johnsbury, VT
- Name: Jared Cloutier
One of my biggest complaints with Zips is that the fielding system is just lazy. I don't see any rational behind his numbers. Derek Jeter has an AV range rating...blech.
I also think the amount of work is underestimated. It was horrible how many little mistakes were made like which way guys hit and numbers being added wrong and Dan has a history of knocking these out. These mistakes were caught because tons of league were using the numbers. It would be tough to check through a home made database and could cause weird things to happen that no one would know till it was too late.
I also think the amount of work is underestimated. It was horrible how many little mistakes were made like which way guys hit and numbers being added wrong and Dan has a history of knocking these out. These mistakes were caught because tons of league were using the numbers. It would be tough to check through a home made database and could cause weird things to happen that no one would know till it was too late.
The sim monsters are a huge part of the fun, IMO, because they create extra assets. Obviously not having sim monsters had a bigger impact on my team than most because of the way I play (100 wins down to 88, yikes), so I'm biased, but they were I thought a crucial part of the projections. Zips are ultimately very conservative projections, he's basically betting that there will be no outliers or breakouts and that makes him right more often than not, which is good for padding your record but not so great for a fun sim league. I always thought that the sim monsters from DMB added those wild card breakout players that make baseball seasons so exciting, and that they were a positive rather than a negative. Zips is fine, it's just not that fun.
Is there a point where we consider an OOTP discussion?
Is there a point where we consider an OOTP discussion?
Your REIGNING AND DEFENDING #evenyear IBC CHAMPION
2015- #torture #evenyears 179-145
2006-2014 Gritty Gutty A's 828-631
2005 Texas Rangers 65-97
Total: 1072-873 .551
2015- #torture #evenyears 179-145
2006-2014 Gritty Gutty A's 828-631
2005 Texas Rangers 65-97
Total: 1072-873 .551
I think for the most part I agree with JP on this issue. While creating our own system sounds great now, I think it will take way more work than we expect and will create more controvery and disagreements than it might end up being worth.
Another alternative that we might consider if a decent projection disk isn't developed by one of the DMB/BP/ etc... would be to do a hybrid type system using the DMB Season Disk (so it includes splits/defensive ratings/ etc) and then create MLE projections for those players that don't appear in the season disk. There are definately pros and cons to this type of system but would be another compromise to consider.
Another alternative that we might consider if a decent projection disk isn't developed by one of the DMB/BP/ etc... would be to do a hybrid type system using the DMB Season Disk (so it includes splits/defensive ratings/ etc) and then create MLE projections for those players that don't appear in the season disk. There are definately pros and cons to this type of system but would be another compromise to consider.
- Orioles
- Posts: 3150
- Joined: Fri Jan 02, 2004 1:00 am
- Location: Glastonbury, CT
- Name: Dan Vacek
- Contact:
I've always been against using the season disk, b/c I think a big part of what makes this fun is having to predict future performance and balance the risk of a below-expectation projection against your long-term expectations for a player, but given that recent performance is probably the most significant factor in valuing players anyway, maybe there's something to this if we make some adjustments like those mentioned above factoring in other projection systems/recent seasons.Mariners wrote:Another alternative that we might consider if a decent projection disk isn't developed by one of the DMB/BP/ etc... would be to do a hybrid type system using the DMB Season Disk (so it includes splits/defensive ratings/ etc) and then create MLE projections for those players that don't appear in the season disk. There are definately pros and cons to this type of system but would be another compromise to consider.
If the season disk were our baseline for player defensive/ speed/ durability ratings, we would avoid delving into that mess, and could be sure that there would be no errors in terms of L/R-handedness. If we found a good way to factor in one or more of ZiPs/CHONE/PECOTA/2008+2007 we'd have workable numbers for everyone who played last year. Our problem then would be dealing w/ MLEs and any other player w/ a reasonable expectation of a big jump in MLB ABs. I don't know how much this reduces the overall complexity of developing a proj system, but at the very least, it would eliminate a lot of the more contentious and labor-intensive elements, like settling on defensive ratings (for most players).
2023 GM Totals: 1780 W - 1460 L | 0.549 wpct | 89-73 (avg 162 G record)
- Brewers
- Posts: 1640
- Joined: Mon Apr 01, 2002 1:00 am
- Location: St. Johnsbury, VT
- Name: Jared Cloutier
I would agree with that as a good way to solve defensive issues, but then you still have to go into minor league defensive rating which would be even tougher then major league guys. It just feels like something that burns one or two people completely out and causes one or two to quit because of someone's conspiracy against them. I think have this dialogue now is good though.
- Dodgers
- Site Admin
- Posts: 5771
- Joined: Fri May 30, 2003 1:00 am
- Location: Fort Lauderdale
- Name: Shawn Walsh
I can't be 100% sure on this until they release more info, but I believe with DMB's API announcement, it may be possible to generate a database with players from a data source, which would allow relatively quick creation of a database from say Pecota or something else. Also can't be 100% sure that I'd have time to program such a thing since it'd be a relatively new language for me, but something to think about (I believe when v10 ships they will announce more information).
- Orioles
- Posts: 3150
- Joined: Fri Jan 02, 2004 1:00 am
- Location: Glastonbury, CT
- Name: Dan Vacek
- Contact:
We could also add in ZiPs bat/pch rtgs for any players not covered in the season disk. I guess it depends what peoples' biggest issue with ZiPs is. ZiPs does kind of suck, but we lived with it this year, and I'd settle for doing so one more year if it ends up being our best option (while hoping DMB or PECOTA provide compete projs). Barring another usable set of projs coming our for this year, or Dan S's final version being improved enough, maybe there's some way we can tweak what we've got to improve it temporarily.
Admittedly w/out looking at any numbers or anything it seemed like hitters underperformed and pitchers performed above expectations. The things that jumped out as the biggest problems were definitely the overly pessimistic offensive performances (causing some of the pitching numbers to seem like 1968), the lack of splits, and the defensive/player ratings.
Defensive ratings I think we can solve w/out much of a problem. Splits I don't think we should touch (at least for 2010) if there's not a set of usable projs that provides them. General sense was that in-play avg was way lower in ZiPs for all players than in both MLB and DMB (a stat I think I remember being projected out by DMB for both LH + RH splits), and just from netplaying a few games it seemed like as a result there were way too many GDPs. Lots of runners erased on the basepaths.
Are the 2008 stats out on the site? Tried this link: http://ibc.poweralleycircuit.com/reports/2008/index.php, but doesn't have 08 season stats up. Am I looking in the wrong place?
I have no idea if or how IP avg and GB% are used by the simulation, but maybe it's a matter of giving in-play avg an across the board bump then running a few seasons to see if they produce more accurate looking numbers. Maybe it's just a tweak to the "Era" settings in the organizer, which I'd assume were set by DMB to produce good numbers in conjunction w/ the DMB 2008 season disk, not ZiPs. Haven't tried this stuff, just hoping somebody knows more than I do, and can shoot this down before I waste a bundle of hours fumbling around with it.
Admittedly w/out looking at any numbers or anything it seemed like hitters underperformed and pitchers performed above expectations. The things that jumped out as the biggest problems were definitely the overly pessimistic offensive performances (causing some of the pitching numbers to seem like 1968), the lack of splits, and the defensive/player ratings.
Defensive ratings I think we can solve w/out much of a problem. Splits I don't think we should touch (at least for 2010) if there's not a set of usable projs that provides them. General sense was that in-play avg was way lower in ZiPs for all players than in both MLB and DMB (a stat I think I remember being projected out by DMB for both LH + RH splits), and just from netplaying a few games it seemed like as a result there were way too many GDPs. Lots of runners erased on the basepaths.
Are the 2008 stats out on the site? Tried this link: http://ibc.poweralleycircuit.com/reports/2008/index.php, but doesn't have 08 season stats up. Am I looking in the wrong place?
I have no idea if or how IP avg and GB% are used by the simulation, but maybe it's a matter of giving in-play avg an across the board bump then running a few seasons to see if they produce more accurate looking numbers. Maybe it's just a tweak to the "Era" settings in the organizer, which I'd assume were set by DMB to produce good numbers in conjunction w/ the DMB 2008 season disk, not ZiPs. Haven't tried this stuff, just hoping somebody knows more than I do, and can shoot this down before I waste a bundle of hours fumbling around with it.
2023 GM Totals: 1780 W - 1460 L | 0.549 wpct | 89-73 (avg 162 G record)