Feb 11, 2011

Introduction to the Baseball Projections



When unlocking the power of these fantasy baseball cheatsheets (or browsing the new printable PDF versions), you're presented with an option to choose which projection system you want to use. This can be very confusing without knowing the difference between each of them. Each system calculates player projections differently and one may suit your tastes a bit better than another. So, let's take you through a quick tour with some brief introductions to each:

Marcels

Marcels is a long-time staple among baseball projections and has been quite successful despite being unabashedly simple. Tom Tango developed it and considers it the most basic forecasting system one can have (so easy a monkey named Marcel could do it, he would say). The system takes the past 3 years of baseball data (weighting most recent data heavier) and regresses the players towards a mean based on age factor. Tango wrote a full introduction to the system years ago that goes into more detail and also wrote an explanation about why the HR totals are lower than some may like.

In one version of a 12-team fantasy league using 2011 projections, here is an idea of what an average fantasy hitter and pitcher would produce with Marcels (which would be used for comparison purposes when calculating a player's WERTH value):

Hitter: .273 avg, 16.6 HR, 68.2 R, 65.7 RBI, 10.7 SB 
Pitcher: 8.75 W, 3.62 ERA, 1.25 WHIP, 7.98 K/9

CAIRO

A somewhat newer projection system, CAIRO, was developed by SG over at Replacement Level Yankees Weblog as a response to some of Marcel's shortcomings from being so basic. For instance, CAIRO factors in minor league performance, park factors and league factors (which Marcels does not). Also, while Marcel ages all stats equally as players get older, CAIRO ages each stat individually as stolen bases might not decline as fast as home runs. A few other differences are that CAIRO uses four years of weighted baseball data and also incorporates a player's position when regressing players towards the mean.  It takes the simplicity of Marcels and adds some more complicated layers to it.

In that same 12-team fantasy league using 2011 projections, here's what an average fantasy hitter and pitcher would produce with CAIRO:

Hitter: .272 avg, 18.3 HR, 75.3 R, 72.7 RBI, 10.8 SB 
Pitcher: 9.07 W, 3.86 ERA, 1.26 WHIP, 7.81 K/9

ZiPS

Similarly to Marcels, ZiPS is a system that has been around for a long time, developed by Dan Szymborski. This system mostly uses 4 years of statistics (a weight of 8, 5, 4 and 3 respectively for each year) but uses 3 years for pitchers and younger/older players. A player's growth or decline is determined based on player type (locating large groups of players with similar statistical characteristics). So, while the other systems are somewhat designed to find averages and regress to the mean, this system finds similar player types and then follows a unique regression or progression based on that player type.

For that 12-team fantasy league with 2011 projections, here's what an average fantasy hitter and pitcher would produce with ZiPS in 2011:

Hitter: .273 avg, 18.4 HR, 74.7 R, 73.0 RBI, 11.7 SB 
Pitcher: 10.06 W, 3.58 ERA, 1.25 WHIP, 8.56 K/9

Fangraphs

Unlike the formulatic projections above, Fangraphs is simply a collection of projections made by fans themselves at their website. Only players that have 15 fan projections or more are used here. The projections are typically a bit more optimistic than a traditional projection system which may speak to the nature of fans themselves. This system relies on the knowledge of the people and may give you a better idea on what the general public is thinking about a player. There were a total of 50,000 ballots last season and the system finished 10 out of 21 in the Forecaster Challenge despite it's possible shortcomings (it should be noted that CAIRO did not fare well in this while Marcels finished tops and ZiPS was not included and it should also be noted that the creator of Marcels ran the challenge based on the metrics of his choosing).

Using Fangraphs for our 12-team league with 2011 projections, we get a more optimistic view of both fantasy hitters and pitchers:

Hitter: .281 avg, 20.0 HR, 82.3 R, 85.2 RBI, 12.3 SB 
Pitcher: 11.45 W, 3.48 ERA, 1.23 WHIP, 8.34 K/9

Combined

If you aren't comfortable with any of the four systems here, the Combined projection option offers a safer view of projections by averaging out these four to come to some sort of consensus. The advantage is that it compensates for the shortcomings of any system while the disadvantage would be that it doesn't fully reward the potential benefits of any system. It's probably not the ideal system but it's a suitable fallback option if you can't make up your mind. As you would suspect, this option generates an average of the available projections that are shared between each of the systems above.

For that same ol' 12-team fantasy league and 2011 projections, here's what an average fantasy hitter and pitcher would produce with this system:

Hitter: .273 avg, 17.5 HR, 72.3 R, 70.1 RBI, 11.0 SB 
Pitcher: 9.35 W, 3.67 ERA, 1.26 WHIP, 8.18 K/9

Comparison

For your purposes, you might find one system fits your taste better than another or you might find that's best to play around with them each.  However, if you want a basic idea of what type of league each operates within, here is a breakdown for each system and how the average fantasy player breaks down for each:

Statistic
Marcel
CAIRO
ZiPS
Fans
Comb.
Batting Avg
.273
.272
.273
.281
.273
Homeruns
16.6
18.3
18.4
20.0
17.5
Runs
68.2
75.3
74.7
82.3
72.3
RBI's
65.7
72.7
73.0
85.2
70.1
Stolen Bases
10.7
10.8
11.7
12.3
11.0
ERA
3.62
3.86
3.57
3.48
3.67
WHIP
1.25
1.26
1.25
1.23
1.26
K/9
7.98
7.81
8.56
8.34
8.18

Filed Under:

0 comments:

Post a Comment