THE BOOK--Playing The Percentages In Baseball

Nate tells us who the over and under achieving teams are:
http://baseballprospectus.com/unfiltered/?p=792

I’d like to see the whole list.  After all, it could simply be noise.  But, we can take a guess based on the numbers that have been reported.  Over 5 seasons (810 games), 1 SD = 14.2 wins, if all was random.  What is the SD of Nate’s list?  I don’t know, but I’ll guess it is around 1 SD = 20 wins.  So, the spread is about 1.4 times wider than random, meaning that the difference in forecast was not just by luck.

Among the reasons that Nate listed, I think fielding may probably be one, but it depends how he handles ERA for pitchers.  After all, if a pitcher primarily pitchers with the same fielders behind him, year-to-year, then his forecast will include the fielders themselves.

Teams with alot of young players may be another source of uncertainty, since you have less playing time to forecast with them.

Righting-the-ship, meaning throwing money at the problem, or bringing up players faster than planned, might be another source, as he noted.

I don’t like his chart at the end, as it could very well be that *all* forecasting systems followed that “trend”.  It’s not a trend so much as luck.  That chart only makes sense in conjunction with other similar charts.

I have a hard time buying CHONE’s 89 wins for the Rays. But hey, anything’s possible - but I think that’s a touch optimistic.

I’ve re-run CAIRO since the Santana trade:
http://www.replacementlevel.com/index.php/RLYW/direct/diamond_mind_and_cairo_post_santana_trade
My bigger projections are coming in a couple of weeks, I’ll put up a link when they’re ready.

It probably is, but the Rays projection is the only one that is the least bit interesting to me.  Check all the other projections for CHONE - there’s nothing that I couldn’t have come up with while sitting on the toilet, knowing nothing more than about what the team did last year and who they grabbed in the offseason, and not running any projections or using a computer for anything at all.

Other projections I find interesting- for Cairo it’s the Giants not being horrible and KC in third.  For PECOTA it’s Mariners finishing dead last.

I realize interesting doesn’t necessarily mean correct.  I’m pretty much alone in projecting the Rays as a serious contender, though if they post a slight winning record like PECOTA predicts it will be a huge step forward for them. 

I don’t think I’m better or worse than the other prognosticators, so if I’m alone on the 89 wins, and everyone else has them between 77-84, odds are I’ll be wrong.  But it will mean I’ll follow them a little more closely this year, to see if what my system likes about the Rays has any basis in reality.  This reminds me that I planned to detail exactly how I see the Rays improving so much.

Sean, the latest PECOTA standings have the Rays at 89 wins too.

FWIW, when I run your projections through Diamond Mind 1000 times I get the Rays at 87 wins, so there may be a strength of schedule penalty.  On average the systems/simulations I’ve run to this point have them at 81-81 with 801 RF, 794 RA.

The PECOTA projections have been updated to reflect changes in team defense.  Most notably, Tampa has gone from 82 wins to 89, and the Padres took a hit down to 77.

You can access my projections by clicking my name.

James, can you give me a link to how you create your player forecasts, however nondescript it may be.

The CAIRO ones, as well as CHONE and PECOTA, all have Marcel as some sort of basis.

***

To the DMBers: what is the correlation between runs differential and win differential?  Can I guess it’s r=.99?  Can I further guess that the biggest gap might be 0.50 wins between PythagenPat and simulated wins?

That is, regardless of bullpen strength, type of team offense, etc, etc, simply knowing the team totals of runs scored and runs allowed (i.e., knowing the playing time, and a player’s rate stats) is enough to give us a forecast?

That really, the *only* benefit of running a sim is to appease those people who think that the particular team makeup can be more beneficial than Pythag?

I could be wrong.  I’d like confirmation either way.

Tangom, here’s a sneak peek at my 2008 simulations so far.

Team Wpct Pyth
ARI08 .5270 .5257
ATL08 .5448 .5412
BAL08 .4131 .4102
BOS08 .5714 .5685
CHA08 .4529 .4539
CHN08 .5521 .5492
CIN08 .4794 .4757
CLE08 .5501 .5578
COL08 .5009 .5024
DET08 .5552 .5553
FLA08 .4128 .4227
HOU08 .4500 .4536
KC08 .4523 .4499
LAA08 .5415 .5360
LAD08 .5252 .5175
MIL08 .5244 .5254
MIN08 .4652 .4647
NYA08 .5908 .5866
NYN08 .5948 .5882
OAK08 .4968 .4979
PHI08 .5325 .5352
PIT08 .4278 .4261
SD08 .5238 .5269
SEA08 .4701 .4678
SF08 .4496 .4541
STL08 .4756 .4799
TB08 .5011 .5039
TEX08 .4593 .4615
TOR08 .5272 .5294
WAS08 .4329 .4394
Correlation .9971

Team Proj W Pyth W Difference
ARI08 85.4 85.2 0.2
ATL08 88.3 87.7 0.6
BAL08 66.9 66.5 0.5
BOS08 92.6 92.1 0.5
CHA08 73.4 73.5 -0.2
CHN08 89.4 89.0 0.5
CIN08 77.7 77.1 0.6
CLE08 89.1 90.4 -1.2
COL08 81.2 81.4 -0.2
DET08 90.0 90.0 0.0
FLA08 66.9 68.5 -1.6
HOU08 72.9 73.5 -0.6
KC08 73.3 72.9 0.4
LAA08 87.7 86.8 0.9
LAD08 85.1 83.8 1.2
MIL08 85.0 85.1 -0.2
MIN08 75.4 75.3 0.1
NYA08 95.7 95.0 0.7
NYN08 96.4 95.3 1.1
OAK08 80.5 80.7 -0.2
PHI08 86.3 86.7 -0.4
PIT08 69.3 69.0 0.3
SD08 84.9 85.4 -0.5
SEA08 76.2 75.8 0.4
SF08 72.8 73.6 -0.7
STL08 77.1 77.7 -0.7
TB08 81.2 81.6 -0.4
TEX08 74.4 74.8 -0.4
TOR08 85.4 85.8 -0.4
WAS08 70.1 71.2 -1.1

SG, cool, thanks.  That’s a mighty high correlation.

Can you confirm if you used PythagenPat or straight Pythag? The way to do PythagenPat is to have a custom exponent, rather than “1.83”.  To figure that out, you do:

exp = (RS+RA)^.28

So, if you score 5.565 and allow 4.5, the exponent is 1.91.  If you score 4.5 and allow 3.590, then it’s 1.80. 

In both cases, PythagenPat would produce a win% of .600.  If you use a common exponent, say 1.84, you get a 0.9 win difference over 162 games.  Given the kind of differences you are showing, I’d like to remove that as a potential source for bias.

OK, I did straight Pythag using 1.83 as the standard exponent.  Re-doing it with PythagenPat:

Team Wpct Pyth EXP
ARI08 .5270 .5261 1.859
ATL08 .5448 .5427 1.896
BAL08 .4131 .4059 1.919
BOS08 .5714 .5709 1.894
CHA08 .4529 .4518 1.915
CHN08 .5521 .5509 1.892
CIN08 .4794 .4747 1.899
CLE08 .5501 .5596 1.889
COL08 .5009 .5025 1.913
DET08 .5552 .5575 1.906
FLA08 .4128 .4196 1.903
HOU08 .4500 .4519 1.896
KC08 .4523 .4479 1.902
LAA08 .5415 .5368 1.871
LAD08 .5252 .5179 1.867
MIL08 .5244 .5263 1.894
MIN08 .4652 .4640 1.869
NYA08 .5908 .5911 1.927
NYN08 .5948 .5902 1.872
OAK08 .4968 .4979 1.877
PHI08 .5325 .5370 1.923
PIT08 .4278 .4239 1.884
SD08 .5238 .5271 1.843
SEA08 .4701 .4673 1.861
SF08 .4496 .4538 1.843
STL08 .4756 .4793 1.881
TB08 .5011 .5041 1.897
TEX08 .4593 .4594 1.931
TOR08 .5272 .5299 1.862
WAS08 .4329 .4368 1.908
Correlation .9970

Team Proj W Pyth W Difference
ARI08 85.4 85.2 0.1
ATL08 88.3 87.9 0.3
BAL08 66.9 65.8 1.2
BOS08 92.6 92.5 0.1
CHA08 73.4 73.2 0.2
CHN08 89.4 89.2 0.2
CIN08 77.7 76.9 0.8
CLE08 89.1 90.7 -1.5
COL08 81.2 81.4 -0.3
DET08 90.0 90.3 -0.4
FLA08 66.9 68.0 -1.1
HOU08 72.9 73.2 -0.3
KC08 73.3 72.6 0.7
LAA08 87.7 87.0 0.8
LAD08 85.1 83.9 1.2
MIL08 85.0 85.3 -0.3
MIN08 75.4 75.2 0.2
NYA08 95.7 95.8 0.0
NYN08 96.4 95.6 0.7
OAK08 80.5 80.7 -0.2
PHI08 86.3 87.0 -0.7
PIT08 69.3 68.7 0.6
SD08 84.9 85.4 -0.5
SEA08 76.2 75.7 0.4
SF08 72.8 73.5 -0.7
STL08 77.1 77.6 -0.6
TB08 81.2 81.7 -0.5
TEX08 74.4 74.4 0.0
TOR08 85.4 85.8 -0.4
WAS08 70.1 70.8 -0.6

Great thanks. 

You have 1 SD = 0.65 wins for the league. 

What would random chance tell us?  Presuming you ran 100 simulations, meaning 16200 games, then 1 SD = .0039 per game, or 0.64 wins per season.

In essence, it seems to me that the gap in sim wins and pythag wins is entirely a function of the number of sims you ran.  Therefore, I would stick with runs only.

The Rivera/Joba combo had zero effect after 100 seasons.

Tango: how about having a fan prediction to go against projection systems and see if we’re THAT much worse? Rally did mention the fact that most of the projections are quite uninteresting really, so I’d like to see if the fans can get any close, and how much, to those systems.
Hey, and fans can take intangibles into consideration

Baseball Prospectus does that every year, and the fans do just as well as the forecasting systems.

Check out my blog here for a thread on the matter.

I might be wrong, but don’t they just do “straight” projections?
I was thinking of something like the fielding fan scouting reports, where fans assess individual areas of the team (rotation, bullpen, lineup, bench, fielding, baserunning) and then add everything up to project standings. That way we could see what fans judge better than projection systems and what projection systems see better than fans (if there are any such patterns). Couldn’t this further improve projection systems by “teaching” them stuff they don’t really do well?
Basically, what types of teams are better judged by fans and which by teams? And why?
If something like that has already been done, could you please link it to me?

Per piaceri signore (hope I got the spelling right), It would help me out that if I ask you to do a search on my site, that you take a few minutes for that. 

This is what I was talking about:
http://www.insidethebook.com/ee/index.php/site/comments/baseball_prospectus_readers_know_what_they_are_doing/

Almost everyone seems to overlook a very important point about forecasting team wins, and I’ve been meaning to write about it.

There are really two kinds of forecasting (actually 3).

One is projecting player value, playing time (and injuries), team intangibles (whatever goes beyond he player projections) and the somewhat nebulous concept of which teams might get help during the season, dump players, bring up young ones, not bring up young ones, etc. (personnel moves and changes).  When the smoke clears after the season is over, you simply compare your projected wins with the actual wins (or withe pythag record perhaps), regardless of what a team did or what happened during the season.

The other is projecting team wins AFTER we know everyone’s playing time, based on pre-season projections - i.e. after the season is over.

Both have their merits and whatever the opposite of merits is (demerits?).

The first one is obviously the whole picture and incorporates many skills.  However, it also includes many things that have nothing to do with the forecaster.  For example, let’s say that everyone projects the Mets to win 95 games, but I project them to win only 90, just because I have a hair up my nose or perhaps even because I have bad player projections for some of the Met players (or other ways in which my projection is genuinely bad).  Now let’s say that Jose Reyes breaks a leg on April 10th and Santana blows out his arm on April 4th and the Mets win only 87 games.

Should I be commended for a great job with my rogue projection?  I probably would be by some people, but that would be nonsense (as far as my skill in projecting the Mets of course).  Same idea for blockbuster in-season trades, etc. Anyway, you get my point.

So the question is, what is the best way to REALLY evaluate projections.  One way is to use my second method - simply using actual playing time.  Of course that would mean that everyone is simply being evaluated on their skill at player projections and their methodology for putting those players together to get a team w/l forecast.  It eliminates any skill there is at projecting paying time, and intangibles (including general organizational and managerial skill and philosophy).

If it were me, I’d go with the second one, if there were only two choices.  I think it is a MUCH better way at truly evaluating forecasts.  Heck, when I do my forecasts, I use Pecota’s playing time estimates anyway.  I don’t do playing time estimates.

There probably is a much better happy medium though.  Something like taking each team’s final results (actual w/l) and factoring out relatively unpredictable things (like a player being out for the season, or some major in-season trade), and THEN comparing a team’s “adjusted” w/l to the forecasters.  Or something like that.

Actually, what you are talking about is part of your article in the THT book, right?  I have no complaints with it.

But, when it comes to making a final evaluation, one that you are betting on, you really have no choice but to evaluate it based on the totality of it all, and comparing the wins you forecasted as of March 31 to the wins they had on Oct 1.

I agree with you in principle, but people will only buy this method.

On BP Nate shows the average error in the team record forecasts, his error was about 4.3 last year, the best in the last 5.  That easily beats mine, which came out at exactly 6.0.  Assuming I’m doing it right, I just took the absolute value of wins - projected wins and averaged, I didn’t square anything like you’d do in a StDev.

This is despite the CHONE projections being as good or possibly better than the PECOTA projections, at least for 2007.  So I guess I didn’t do a very good job projecting playing time and manager preferences.  I did much better in the American League, nailing all 4 playoff teams, but only picked the Cubs in the NL.

It could very well be me being more familiar with the AL, being an Angel fan living in Baltimore.  Or just luck.

Tango, I agree totally with #18.  While I understand MGL’s point, that won’t go over in Vegas.  The people trying to collect on their Cardinal bets who say “Nobody could have known Carpenter would get injured” are SOL.

Its too arbitrary.  For an intellectually honest analyst like in MGL’s THT article, that approach makes sense to a sober reader. But when money and/or bragging rights are being exchanged, we all need something concrete and beyond dispute.  The Cards only won 78 games last year.  Doesn’t matter how.  No ifs, ands or buts, if you put money on them to repeat, time to pay up.

Of course, “in practice” you can’t do that, but if someone nailed the Cardinals last year, there would be no doubt in my mind that they had no idea what they were doing.  Simple as that.  It makes absolutely no difference to me (in terms of evaluating the quality of the forecast) whether someone nails a team or the whole league or not, unless I look at all the underlying factors.  Just like you can easily do all the right (optimal) things in a game (or season) and either win or lose, so to with team projections.

And of course if you use playing time after the fact, all you are doing is testing your projections - nothing more.  After all, all anyone (like Chone, me, BP, etc.) does is to take their projections, pro-rate them for playing time (and leverage I guess for relievers - at least the closer and maybe the set-up guy) and put them in a sim or just add up all the RAA and then use a pythag.  Even strength of schedule is not that big a deal.

The other thing that is “silly”, IMO, is to give some credit for nailing a team’s w/l record when their w/l record is far from their pythag record - like the D-backs last year. If someone truly has the “skill” to predict/forecast a team outperforming or under-performing their pythag by more than a few wins, I’d like to get hold of that person.  We can make a lot of money.

IOW, the whole idea of forecasting w/l records, at least in so far as comparing them to actual w/l records at the end of the year, is a joke - to me at least.

And yes, as I discuss a little in my THT article, the only thing that really helps to analyze a team both retrospectively and prospectively is comparing how they did to how they “should have done,” after the fact.

If I am in charge of a team and my team is projected to win 85 games before the season starts, but they win 78 games, I want to know, what I was supposed to win AFTER THE FACT.  If that number was 85, then I have to wonder why my team performed so badly, at least in terms of w/l (and of course I would want to look at my pythag record and my underlying WAA).  But if it was 85 before the season started and it was 77 after the fact because of injury or because I traded my best player, that is a completely different story.  I could actually say that my team overperformed.  Of course most GM’s have NO IDEA what their team is supposed to do before the season starts.

In fact, I would guess that if you polled all the GM’s (and managers and players) and it was a secret, anonymous ballot, and you implored them to give a fair assessment, the average number of wins would be around 83, maybe more.

Tango, I use other people’s projections (mainly PECOTA/ZiPS) but I calculate the team records differently.  Instead of running Diamond Mind sims or anything like that, I use a spreadsheet I designed that models run scoring distributions.  Its primary purpose is to estimate the odds of Team A beating Team B in one game, but it’s not hard to extrapolate to 162 games.

Speaking of ZiPS, post 55 here has projected standings:

http://www.baseballthinkfactory.org/files/oracle/discussion/2008_zips_projections_for_diamond_mind_9_and_microsoft_excel_build_1/

However these are a month old so they might not reflect proper playing time/injuries/roster moves etc.

#17 is an extremely important point, although you’d have to talk everyone into re-running their projections at the end of the season.  MGL is the only person I’ve seen do this.

#19, this could be a function of Diamond Mind quirks.  I find that no matter how much I try to restrict the playing time of certain chronically injured players, Rich Harden still pitches 150 innings most seasons.

Finally, the five-year standard deviation for the PECOTA spreadsheet Nate put out is 8.49 wins per season/team, versus an average error of 5.6 wins.

MGL/21 - At that point, aren’t you just doing the same thing that people do when evaluating individual player forecasts, just using a different presentation of the information? We already know the accuracy of forecasting systems on the player level.

Maybe there’s some value in knowing that, say, PECOTA does better with the player forecasts of the Blue Jays than the Devil Rays, but I don’t know what that would be.

There are really two elements in play here: how good Forecasting System X is at projecting player performance, and how good Forecaster X is at forecasting playing time. I think the latter is a much more interesting question than the former; any of the Marcel-like player forecasting systems (there really needs to be a term to differentiate real forecasts from fantasy forecasts) are close enough to each other for these purposes. Being able to accurately predict playing time is probably the bigger indicator of forecasting success.

MGL/21 - At that point, aren’t you just doing the same thing that people do when evaluating individual player forecasts, just using a different presentation of the information? We already know the accuracy of forecasting systems on the player level.

Absolutely.

And you are right about forecasting playing time.  But as with the statement above, you don’t need or want to use team w/l forecasts to evaluate playing time forecasts!

So what we are left with is evaluating forecasters’ player rate projections and their playing time projections.  Doing the team w/l projections is nothing more than a fun exercise which means very little.

Re #7 and other comments on PECOTA projections.  The full projections by team posted above were the preliminary ones before defense was considered AND before depth charts were set.  There were many changes in the predicted wins in the revised projections based on establishing depth charts.

If last year is a precedent, those team projections will be revised a couple more times before the season starts but further revisions will be quite minor. If anybody wants to see the “latest” (and they have asubscription) then they should go to the “Depth Charts” here: http://baseballprospectus.com/fantasy/dc/

Doing the team w/l projections is nothing more than a fun exercise which means very little.

That’s the way I always looked at it.  I do them because they’re fun to look at, not because I expect them to be especially predictive.

The Vegas over/unders:

http://www.battersbox.ca/article.php?story=20080310150539866

The average is 81.03 wins.  Better than in some past years, where the average was 81.5.

Question: what happens if the season is cut short to say 154 games for some reason (say a walkout by the players)?  And what if a team plays 160 games, while another plays 163?

It would seem that a W/L differential would be more appropriate.  So, an over/under of 93.5 is really an over/under of 25 win differential (games over “.500").

If last year is a precedent, those team projections will be revised a couple more times before the season starts but further revisions will be quite minor. If anybody wants to see the “latest” (and they have asubscription) then they should go to the “Depth Charts” here:

Yup, if you didn’t know that they were in the “depth charts” section, you would have to spend about half an hour trying to find them because it is apparently too much trouble for them to put a link somewhere easy to find that says, “2008 Team w/l projections” or something like that.

I have been railing (not for a while though) about their BAD web site (as far as navigation and ease of use) for years!  Plus it is WAY too busy!

Tango, in order to bet on the Vegas over/unders they have a minimum number of games (it might be 162 actually) for the bet to stand, otherwise it is a push.

Now hold on about the 162.  If let’s say you bet for the Yanks on the over 93.5, and they have already won 94 games, then regardless of how many games played, the bet stands, right?

Similarly, if they won 92 games after 161 games, and if they don’t play game 162, then the bet also stands, right?

In effect, you are better on being over the 93.5 wins AND being under the 68.5 losses.  If you play enough games that this is true, then you win.

Similarly, on a “False” and “False” (Yanks have fewer than 93.5 wins and more than 68.5 losses), then you lose.

It’s only on the “True/False” or “False/True” in terms of the “over WinValue / underLossValue” that it should be a push.

Right?

I just checked the odds on the BetCRIS sportsbook, and the minimum games for “action” is 160. Otherwise, it’s as if you never made the bet at all.

Now hold on about the 162.  If let’s say you bet for the Yanks on the over 93.5, and they have already won 94 games, then regardless of how many games played, the bet stands, right?

The answer is that no, the bet does not stand even if they already won the over.  The same is true on totals bets. The game has to go at least 8.5 innings even if the game already went over the total.  Otherwise everyone would bet the over on the game when it was raining, since they would either win or get their money back if it got rained out before 8.5 or 9 innings.  The books either need to make all bets count for any number of games or no bets count for a min number of games.  And of course if they made all bets count for totals, then if it was raining when the game started or there was a threat of rain, they would have to make the total really low and even then, people with good knowledge about the weather might be able to get an edge.  In any case, the best thing is to do what they do, which is to have a min number of games or innings.  For the season over/unders it doesn’t make much of a difference of course, since the chances of a team playing less than 160 games is infinitesimal.  But they just want to make sure that if the season only goes 100 games (like in a strike shortened season), that the over betters are not going to scream bloody murder (which they would of course).  Anyway, that’s the way they do it with all of these kinds of bets including totals on individual games - and it makes sense.

#28: You’re looking at some outdated numbers.  Right now, the major online sportsbooks average about 81.25 wins per team, so there’s a clear bias towards the over.

MGL is dead on in #32.  For season win over/unders, I wish they would make a rule where you can still cash a ticket if it’s a lock, but it’s unlikely to ever matter.

THT has their forecasts here:
http://www.hardballtimes.com/main/blog_article/tht-projections-update/

My projections with Diamond Mind are up now. I ran the following forecasting systems 1000 times each:
CHONE
Diamond Mind
THT
PECOTA
ZiPS
CAIRO

Then I have the aggregate rolled up into one final standings.

I had to split it into two parts.

http://www.replacementlevel.com/index.php/RLYW/direct/the_2008_diamond_mind_projection_blowout_pt_1

http://www.replacementlevel.com/index.php/RLYW/direct/the_2008_diamond_mind_projection_blowout_pt_2

tango, the THT forecasts you linked above this post are 100 that I ran for David last week, the 1000 I ran are the same thing but in a larger sample.

SG, great work.  Since you seem to have all the players mapped, could you produce a file with all the forecasts?

***

And: any reason you didn’t do Marcel as well?  They are just as good as all the other forecasts, and in some cases, better.  Some people fall into the trap that just because they are “simple” and “open source” that it must be inferior.  At the very least, they provide a benchmark against which all other systems are evaluated, considering it’s the only system that does not get altered ever.

Sometimes Marcel gets treated like Barry Bonds!

Since you seem to have all the players mapped, could you produce a file with all the forecasts?

Yeah, I can see if I can put that together.  The ZiPS and Diamond Mind disks won’t map to the other ones so it’ll take a little time but it shouldn’t be too bad.

And: any reason you didn’t do Marcel as well?

Mainly the minor league issue.  Since Marcel doesn’t use any minor league data I was concerned about having minor leaguers projecting as league average.  I can add it in for next year.

SG, got it.  Note though that you need have no concern, since Marcel’s official forecast is league average for all players with 0 prior MLB PA.

Note though that you need have no concern, since Marcel’s official forecast is league average for all players with 0 prior MLB PA.

Actually, that is my concern.  I think it may overstate the talent level of a team like Florida if we assume that all their minor leaguers are good enough to be league average.  But I may be overthinking it, I’ll probably include Marcel going forward.

If you knew that a team was going with a bare-minimum payroll, and decided to field a team exclusively of guys who had been in the minors, would you expect that team to finish 81-81?

Yes, but that is Marcel’s concern, not yours!  Marcel is saying: every single player that is not forecast is being forecasted as league average.

Consider for example these guys who have an explicit forecast in Marcel:
Cruz Enrique (1 PA since 2005)
Greenberg Adam (1 PA since 2005)
Melillo Kevin (1 PA since 2005)
Morton Colt (1 PA since 2005)
Clark Jermaine (1 PA since 2005)

They all have the exact same forecast.  Their forecast was based on 99.5% league average and 0.5% their performance in that 1 PA.  Needless to say, whether they went 0-1 or 1-1, they are getting virtually the same forecast.

Marcel is simply extending that to its logical conclusion: everyone who is breathing, but has 0 MLB PA since 2005, is forecast for league average.

Since I have no desire to put out a file with thousands of records of exactly the same hitting line (with just the name differing), I just put out a blanket statement that everyone not in that file is league average, whether he is Fukodome, or some schlub in A ball.

Rally: yes, that is Marcel’s position.  Like I said, this is Marcel’s concern.