Tuesday, February 10, 2009
Predicting replacement levelPosted by Paul Singman at 1:01am
During a draft, many people experience this typical dilemma: Should I take the best player available, or should I take the most valuable player available?
Drafting a catcher in the third round is never easy, especially when there is a slugging first baseman available; just as taking that first basemen early is tough, knowing that good first basemen could be available later in the draft and all of the good catchers could be taken by the time you get around to selecting one. There is no right answer here; every situation is unique, but there are certain adjustments we can make to players' stats that can help us make the right decision.
One of those is the replacement level adjustment.
Back in November I wrote an article on how replacement level is applied to fantasy baseball. Let's review an except from it to make sure we are on the same page:
A replacement player is expected to produce at the replacement player level. The best way I can explain replacement level is by creating the following hypothetical situation: There are 30 teams and only one shortstop per team. (That makes 30 starting shortstops.) No team has a bench shortstop and all non-starting shortstops are placed in a pool from which any team can sign them, but only if a team's starting shortstop cannot play due to injury or some other reason.
One starting shortstop does get injured, so that team signs the 31st-best shortstop to "replace" its starter. His expected production is replacement level. It is the baseline from which all other production or value should be judged. The difference in production between the starting shortstop and his replacement is the starting shortstop's value over a replacement player.
Then I continued:
In a fantasy baseball league, replacement level players are those you can simply add from waivers or the free agent pool. They are in abundance and cost nothing to acquire.
In retrospect the part that says, "They are in abundance" is not exactly true. In real baseball it is true that replacement level players are abundant. However, in fantasy leagues, if you are looking to add someone from the free agent pool, there will only be a few free agents worthy of considering. Even though all free agents cost the same, some are much better than others, meaning "good" replacement level players are not so abundant.
With that in mind, let me explain how I am predicting position scarcity for 2009.
When a lot of "fantasy baseballers" (props to Razzball for the term) try to incorporate positional scarcity into their rankings, they make one huge mistake; They use last year's numbers. When determining position scarcity for the 2009 season, I would want to use 2009 season numbers. Problem is, the 2009 season has not yet been played.
To get around this problem, I use the next best thing, which is projections for all players for 2009. Conveniently, we have our own projection system here at THT, courtesy of David Gassko. While the projections are currently available only to me in an Excel spreadsheet, they can also be available to you if you purchase the 2009 THT Season Preview Book.
Calculating replacement level
In a standard 12-team league, the replacement level player for each position will be on average the 18th catcher (30th in two-catcher leagues), 24th first baseman, 19th second baseman, 21st third baseman, 19th shortstop and 50th outfielder. To find the replacement level for each position, I averaged the projected OPS for the first replacement level player, and then the players projected before and after him. So for the catcher position, I average the 17th, 18th and 19th best catchers to end up with one expected projection for the replacement level catcher.
Below is a chart of the results with the 2009 numbers being the projected replacement level for each position, thus the dotted line.
Note: C1 represents a one-catcher league, and C2 represents a multi-catcher league
Sure it is a nice looking graph, but what does it mean? The projected drop in replacement level from the outfield, second base and catcher positions means that your replacement level adjustment should be more than it was for these positions last year. Conversely, your shortstop and first base adjustments should lessen because the replacement level first basemen—the guy you can add for nothing—is projected to be better.
To make the adjustment you simply subtract the player's OPS from his position's replacement level OPS. For 2009 the projected values are (these are the values in the graph):
+-----+-------+ | Pos | OPS | +-----+-------+ | C1 | 0.735 | | C2 | 0.705 | | 1B | 0.791 | | 2B | 0.728 | | 3B | 0.776 | | SS | 0.727 | | OF | 0.778 | +-----+-------+
So, if you have a first baseman with an OPS of 1.000, his value is equal to that of a catcher with a .944 OPS.
Hmmm, that does not seem right. Remember that this is just the replacement level adjustment. There is another adjustment for position scarcity and it is the talent distribution adjustment. I will get into that adjustment in another article.
Deciding what replacement level exactly is and determining how to calculate it are always sticky subjects, so now I am going to point out of the possible flaws of my system.
Using OPS will bother a lot of people. I doubt using using wOBA or another stat of the sort would change the results around a lot, and I could have used wOBA only if I used Marcel instead of the THT projections. Nothing against Marcel—it always holds its own—but I cringe at the projections when it comes to younger players. In the future I might go through the same methodology using Marcel's projected wOBA and then we can compare results.
One of the hardest things tro decide was how many players from each position are drafted. The minimum was 12, but then I had to account for batters in the UTIL spot and on the bench. Admittedly there was no real scientific approach to designation of which players are replacement level, but I did get my numbers through some logical reasoning.
To be specific, I started by saying 108 batters must be drafted. That is, nine starting batting spots multiplied by 12 teams to get 108 starting batters. Then, I decided that on average eight teams will have three bench players, with the remaining four teams holding four batters on their bench. That is 108 + (8 X 3) + (4 X 4) which equals 148. So, in the average league 148 batters are drafted. If you sum up the totals of replacement level rank for each position minus one (because the last player drafted is the player right before the first replacement level player) you get 17 + 23 + 18 + 20 + 18 + 49 which equals 145. If you account for three DHs being drafted, you get the previously deduced 148.
In calculating the replacement level production, I did not want to just take the OPS of the 18th catcher so I used a three-player average. Again, no real scientific reason behind the decision to do that, but I feel it worked out well.
Obviously my methodology is not the tightest, so I am open to suggestions to improve the way I went about this. You can leave those in the comments.
Paul has been managing fantasy baseball teams for many seasons and writing for THT Fantasy over the past three years. He is currently a student at UPenn welcomes readers' thoughts at his email here or in the comments below.