There are basically two types of baseball projection systems that exist: human projections and scientific projections. The human projections would be the type you see at ESPN or CBS where they have their staff make their best guess at what a player will do next season. The scientific projections are based off of formulas and weighting systems taking into account various factors related to past player performance. These would be systems like Marcel, ZiPS or Steamer.
When relying on the scientific projections, we get a good sense of what we should expect out of players for the upcoming season. These systems generally follow a method of taking data from that player's recent history and weighting the most recent years more heavily than older years. The projection systems differ from each other by the amount of weighting used and the different factors used for the weighting. They also differ in the stats they use and that some systems take advantage of minor league or foreign league data while others do not. But, in the end, the basic concept of taking history from recent years and weighting it based on a number of factors is the same between all projection systems.
When using these projections for fantasy baseball purposes, there are generally a few caveats to keep in mind as a result:
- They are a conservative estimate. Since the projections look at the past statistics, an unexpected breakout season is generally not going to be predicted. Maybe you think that John Mayberry will break out this year. However, projection systems will not predict that because of his limited playing time (and thus limited production) during his career thus far. If past data doesn't show someone to be a 20/20 player then projection systems won't bend out of their way to make that prediction. So, if you're browsing through the projections to see your favorite sleeper then don't worry if he's not projected to take a large step forward based on data from ZiPS or Marcel.
- Following along that same train of thought, rookie projections aren't as reliable. These systems are based on using past data and the lack of that data for rookies muddles things. Some of the projection systems take minor league data into account which does improve their projections of rookies. But, even with those systems, you may be surprised to see a lower projection for a rookie than you expected. That shouldn't stop you from drafting that player if you really believe in them though.
- On the flipside, projections for veterans are more reliable. When a guy has played in the majors for 10 years or so, the margin of error for a projection is a lot lower because there's so much data to use in making a prediction for them. Almost all systems take into account an aging curve so they recognize that performance for a 35 year old will be similar to his past three years but worse to a certain degree. More data to use for projections means more reliable data.