Projection Variations

[viper]

First off, I hope a new Forum area dealing with projections will be started. Otherwise this General Baseball Topics area could get a bit cluttered. If it is made, please move this there.

I'm a SGP person but applying it the both site's projections, what I am looking at is just differences. I find my general ranking of players using Mastersball projections is essentially identical to the dollar amounts Mastersball came up with. I also do the same with BBHQ. Each season there are differences, some small and a few larger. My SGP is based off the 15 team NFBC concept. Here are the variations for players that are clearly draft-worthy in large league formats. I only did hitters.


My guess is that my expected results from the mock draft will be much better using the BBHQ numbers as compared to the MB numbers. Hell, given I drafted both Jacobs and McLouth, they will definitely be better with BBHQ. To be noted, I have Jacobs is a single biggest variation between the two sites.

[Todd Zola]

EVERYONE --

I deleted the reference to Baseball HQ's projections. Even though they were only referred to in comparison to ours, it is not fair for BBHQ subscribers to have 'free' access to their material, even in the innocent form it was presented.

FYI viper, you "won" the mock using Mastersball's projections

Gary, JP and I realize we likely need a dedicated forum to discuss some of our own pay material, to protect your interest in the same manner described above for BBHQ subscribers. If, and it is still in the if stage, it will be solely dedicated to discussing the subscriber content, and will not be a place for subscribers to get answers to questions, like we had before. We may also use it to discuss new 'proprietary' strategies and theory before we unveil them publicly, we'll see.

[viper]

got the OK to post this general stuff. What I was saying was there are several players whose projections vary by more than a tier.

I'm a SGP person but applying it the both site's projections, what I am looking at is just differences. I find my general ranking of players using Mastersball projections is essentially identical to the dollar amounts Mastersball came up with. I also do the same with BBHQ. Each season there are differences, some small and a few larger. My SGP is based off the 15 team NFBC concept. Here are the variations for players that are clearly draft-worthy in large league formats. I only did hitters.

Hanley Rameriz is +4.8 SGP in Mastersball
Delmon Young is +4.2
Ryan Zimmerman is +3.1

On the other side of the world:
Jason Bartlett is -4.1 in Mastersball
David DeJesus is -4.3
Travis Hafner is -4.4
Nate McLouth is -5.4
Rafael Furcal is -6.0
and the biggest loser is
Mike Jacobs at -7.8 [surely make my mock draft pick look real bad]

This shows just how two sites, which I feel use somewhat similar methodologies, can differ by quite a bit. Maybe these sites do things totally different as I really don't know. After a mock draft, if you don't like your results you can just apply different site's projections until you do well. Of course, the only numbers that really count haven't happened yet and those final results won't be known until October.

[Guest]

So I'll be glad to address some of these players more specifically, however, how about we start by pointing out some of the similarites in Viper's two lists? What might the players within those groups generally have in common.

This might be a helpful way to get some conversation going about the pros and cons (yes, we're fine with that here) of how we're projecting players.

[viper]

the "similars" is a very lengthy list. Again I just looked at the first 210 position players that bubbled up on my draft list, excluding one single player not rated by BBHQ as of 2/1/09.

0-1 differences: 72 players
1-2 differences: 70 players
2-3 differences: 40 players
3-4 differences: 18 players
4-8 differences: 9 players
not used: 1 player

Note that the combination of 5 HR, 5 SB, 7 Runs and 7 RBIs will make about a 3 SGP difference. That is not all that much. I actually consider players within 2 SGPs as pretty much the same player and I think it probably could be extended to 3 SGPs

In general, MB is less hitter friendly - not San Diego but think Oakland. If history plays out like last year, the outliers will decrease to a precious few in four weeks.

[Todd Zola]

Of course I have the advantage of knowing where Gary is going with this

By similarities, focus on the specific players viper used as examples -- is there some commonality with the members of the plus group? The minus group?

[viper]

AB is always a driving factor but several of the outliers are within 20 or so ABs. BBHQ is lower on all the MB plus players and MB is lower on all but two of the MB minus players.

[viper]

If I have a little free time, I might normalize the numbers by recalculating all based on the average ABs for the two sites. I could then see where the bigger differences were. That task is not real high on my list and this occurs naturally in most cases over the next four weeks.

[msugray]

Hanley Ramirez is +4.8
Delmon Young is +4.2
Ryan Zimmerman is +3.1

Jason Bartlett is -4.1
David DeJesus is -4.3
Travis Hafner is -4.4
Nate McLouth is -5.4
Rafael Furcal is -6.0
Mike Jacobs at -7.8

The players in the "plus" group are younger in age.
Hanley 26, Delmon 23, Zimmerman 24.
Bartlett 29, DeJesus 29, Hafner 31, McLouth 27, Furcal 31, Jacobs 28.

Players younger in age may be "under-projected" if appropriate growth rates are not taken into account.
Older players are not likely to increase skill sets and will generally remain static in production or regress.

[shif6]

One of the things I appreciate about Viper's post is that it avoids a trap that is easy to fall into. If you just compare HQ and MB's dollar values, some differences will be accentuated that are not real. This is because HQ has historically used a different split between hitting and pitching than MB, the former giving more weight to pitching and less to hitting than does MB. The way I have compared the two sets of projections is to use draft preparation software (rotolab) which is equipped with HQ projections and $ and compare the $ values with those produced by Paul using the same splits in both cases (the ones appropriate for the particular league). Some of the dollar differences are based on different methods of category evaluation, but some are based on different assessments of player performance. In any event, the differences are always interesting. I think Viper's calculations avoid some of the rough edges of the crude methods I have used, but I would be interested to know how time consuming Viper's effort to put this data together was.

[Todd Zola]

Hanley Ramirez is +4.8
Delmon Young is +4.2
Ryan Zimmerman is +3.1

Jason Bartlett is -4.1
David DeJesus is -4.3
Travis Hafner is -4.4
Nate McLouth is -5.4
Rafael Furcal is -6.0
Mike Jacobs at -7.8

The players in the "plus" group are younger in age.
Hanley 26, Delmon 23, Zimmerman 24.
Bartlett 29, DeJesus 29, Hafner 31, McLouth 27, Furcal 31, Jacobs 28.

Players younger in age may be "under-projected" if appropriate growth rates are not taken into account.
Older players are not likely to increase skill sets and will generally remain static in production or regress.

DING!! We have a winner. Gary - I'm tagging you in

[Guest]

As Todd said, there's your link.

Our biggest "overs" are players with a full three year track record of relative success at a young age, and we likely project the possibility of upside to these players while perhaps other systems see their skills as more flat (I cannot say for sure how they approached the projection). Said another way, we cannot be comfortable that at their ages, they're at their peak. It's certainly a possibility but not the only one. Again, its not just that they're young, it's their stable skill set at a young age that is a common bond here.

Our "unders" are:

- Jacobs, a guy with a pretty steady offensive track record with a major power burst which we're regressing a bit based on the risk its an outlier (likely others consider it permanent)
-McLouth, who had a huge increase in his contact rate while other indicators were basically stable, again its likely we regressed a bit on the contact rate given history
- Hafner, who is a flat out guess by anyone doing a projection, regardless of method.
- DeJesus, who had a relatively big leap in BABIP without a ton of other indicator movement, where we're probably hesitant on the BABIP
- Bartlett, whose very slight skills regression was masked by some BABIP movement (but whose entire value is basically SB driven, so not sure there).
- Furcal, probably a playing time thing would be my guess given his skills are pretty stable (though not sure what the other projection did or didnt project). The injury from last year

[Kstan]

...The way I have compared the two sets of projections is to use draft preparation software (rotolab) which is equipped with HQ projections and $ and compare the $ values with those produced by Paul using the same splits in both cases (the ones appropriate for the particular league). ....

True you can import HQ projections into RotoLab. But HQ's CDG does not match RotoLab values exactly, so Im not sure this is a 100% valid comparison.

[viper]

I think Viper's calculations avoid some of the rough edges of the crude methods I have used, but I would be interested to know how time consuming Viper's effort to put this data together was.

Over the past several years I have developed a hitting spreadsheet that goes to column BP and a pitching spreadsheet that goes to BB. It was a labor of love but did take some time. Part of the sheet is columns for AB, H, etc for hitters as projected by the two sites. Same situation for pitchers. My basic scheme is to average the numbers and apply my SGP delta for each category. I do this for both my averaged results and for each sites unique projections. Excel is real easy once things are in place. Each year i add a little and delete a little. There is a column of differences right next to the column which tracks raw AB differences. Right now I am tracking far too many players but I start cutting back as spring training starts and projections get firmed up. I typically limit the players to those with 200 ABs except for steal merchants. My pitching limit is 100 for a starter. For non-starters, they must be closers or setup men who are next in line. I may add a couple more players on a whim. My leagues typically do not draft middle men, even my AL-only and NL-only leagues. I know they are all either $1 or $2 players.

When new projections come out, it takes about an hour to get both pitchers and catchers inserted.