At this point in the fantasy baseball cycle—in between your draft and the start of the regular season—it is easy to be confident. You know exactly how much you value your players because you know exactly what you expect from them in the upcoming season. A few players, like Chase Utley, might have more questions than certainty around them, but for the most part if you were to receive a trade offer today, you would know fairly easily whether you would accept or reject this offer.
Only in rare moments of immense humbleness and self-honesty will I accept a preseason trade offer in which the players I receive were drafted later the players I am giving. Otherwise, the other 95 percent of time, I simply reply to the sending owner, “If I wanted those players I would have drafted them first.” And usually do so in some manner of disgust.
As confident as you might be right now, in a very short time—two days to be specific—all that confidence will be eroded by the tidal wave that is the start of the regular season and all that you thought you knew will be washed away. All of a sudden new stats are being generated every day, and every day your opinions of players must be adjusted.
Now, a month into the season you receive a trade offer; not as easy to evaluate it now, is it? Before the season you thought Player A was better than Player B, but so far Player B has outproduced Player A and shows no signs of slowing down. Objectively speaking, you have Tom Tango shouting “Regress to the mean!” in one ear.
Midseason this means weighting the player’s current performance by the number of plate appearances he’s had (as a fraction of the total number of plate appearances you project him to receive over the whole year), then weighting your preseason expectations of that player by his projected plate appearances for the rest of the season (again as a fraction) and then adding the products together. So for example, if you pegged Nick Swisher for a .260 average in 600 PAs this season and through 100 PAs he’s batting .200, you would project him for a [(.260*(500/600)) + (.200*(100/600))] = .250 batting average the rest of the way. It might not be the most statistically sound method of simulating regression to the mean (it is probably smart to over-weight the player’s actual statistics a little compared to your true talent estimate) but I understand it and it works well enough.
Sometimes though, the specific circumstances surrounding a player overwhelm what the objective numbers tell you. Some people take this as an excuse to ignore doing even a rough mental estimate of the exercise I performed above, or ignore what a ZiPS or Oliver rest-of-season projection tells you. Even though a player is now playing through an injury or is facing the possibility of losing playing time, you can simply reflect these changes quantitatively by adjusting your “true talent” or plate appearance estimation of this player.
Used correctly, regressing to the mean works most of the time. Most of the time does mean, however, that a number of players will defy the rules. Every year there are Trevor Cahills who continue pitching shutouts, James Shieldses who continue to get shelled, and Jose Bautistas who continue to blast home runs. Plenty of players also obey the regression rule, but you tend not to hear about them as much.
The unfortunate truth is that because of the small samples we work with in partial seasons, our lack of perfect information, and the small number of roster decisions we make midseason, the edge gained by staying faithful to correctly calculated rest-of-season stats is slight over the person who over-indulges in hot streaks. And that is the overall point to take away here.
Although you should follow what a statistically sound rest-of-season projection tells you when deciding on a trade or roster addition, if the numbers are close enough and you have a gut feeling for one player, go with what your intuition tells you. And on the other hand, if you have no idea what to do, go with what the numbers tell you.