2016 MORPS projections are finally ready. Unlike the baseline projections published several weeks ago, these projections include all players projected to win a MLB roster spot on opening day. I have also included a number of impact rookies who are projected to join a roster early or mid-season in 2016. Rookie projections use stats generated during either minor league or international play. Factors are applied to adjust the stats to MLB equivalent stats. MORPS projections also account for expected adjustments in personal playing time.
The excel version of the projections include a key tab that defines all headings used in the projections. In short, fantasy baseball players that play in rotisserie leagues should key on the R-ROTO and ROTO columns. ROTO is a point value derived from weights on the categories in a standard 5×5 rotisserie league. R-ROTO is the player ranking based upon the ROTO point values. If you league uses a customer scoring system, you can use the projections in the categories of interest to customize your rankings. Fantasy baseball players that play in a more realistic format like Baseball Manager or a similar simulation league should reorder the pitching based upon OERV and the batting based upon RC. OERV stands for Out Earned Run Value. This stat attempts to value a pitcher by combining ERA with the value of number of innings pitched. This is a way for fantasy managers in simulation leagues to compare the value of a relief pitcher with a starter or a starter who pitches 200 innings with one that pitches 100 with a slightly lower ERA. RC is Runs Created. A league like Baseball Manager uses RC as a basis for the points they generate in their daily games. The more realistic the simulation, the closer the hitting will align with RC.
For those who like to resort the projections for their own fantasy system, make sure you filter out the players with a roster status of “N”. These players will most likely not make an opening day 25 man roster. Those players who were still in competition for a position were included with a roster status of “Y” in most cases. I posted the “N” players for those managers who have keeper leagues or deeper rosters that may want to pull one of these folks onto their list.
Team projections for 2016 will be posted within the next week.
The 2016 MORPS baseline projections are ready. This is the third year we have provided baselines. These projections use all the models we have put together over the years for projecting player performance. This means that the projections are still based on four years of data, positional mean regression, etc. However, they do not account for a player changing positions, reductions in playing time, new players to the big leagues, etc. We entered all MLB player transactions into system since the end of the regular season last year. While this doesn’t guarantee that we have caught every trade, free agent move or player being waived; we are hoping that the majority of these type of transactions were captured in the system.
Some may like the baseline projections more than the final version. I read one review of MORPS in 2014 that criticized the fact that we took the time to model anticipate plate appearances and batters faced for each team before releasing our final projections. They didn’t consider that process “scientific”. Our perspective is that the modeling allows us to adjust the ratios between each stat and plate appearance or batter faced to account for situations that weren’t present the year before. This could be a player being part of a platoon when they played the position full-time the year before. It could be a reduction in playing time due to the appearance of a blockbuster free agent or anticipated rookie hitting the big leagues. It could also be a pitcher coming back from Tommy John surgery after being out of the game for over a year. Regardless of the situation, we believe that the modeling of plate appearances and batters faced for each team adds significant value to MORPS projections. This view is supported by our #1 ranking in 2014 for player projections using root mean square error (RMSE). If you still doubt our ability to accurately model these situations or you have an early fantasy draft and need something now, you’re in luck. You can use our baseline projections.
So…. without further ado, we present the 2016 MORPS Baseline Projections. The Batting and Pitching projections are available in excel and PDF formats. Follow the links below to download your copy.
If you player Roto baseball, you will find the projections already sorted in Roto Rank order. If you play a more realistic version of fantasy baseball, like BBM, you will need to re-sort the XLS spreadsheet in RC order for batters and OERV order for pitchers. Play Ball!
Each year we take a moment to review our projections against the actual season results. First lets look at the team projections. We definitely missed on the Kansas City Royals and their World Series win. On the positive side, we did predict playoff runs for the Mets, Cardinals, Dodgers, Pirates and Blue Jays. Our only miss in the NL was the Cubs over the Nationals. The AL is another story. In addition to the Royals, we also missed on the Astros, Rangers, and Yankees. Predicting 50% of the playoff participants isn’t bad considering the number of roster changes that happen during the course of the season.
For individual projections, we were pleasantly surprised that MORPS was the number one overall projection system in terms of Root Mean Square Error (RMSE) according to The Baseball Projection Project. The 2014 results were published in March on Fangraphs – click here. Based upon improvements implemented for 2015, our hope is that those results are replicated when 2015 results are published later this year. For fantasy owners, this means that MORPS has the lowest error rate of all published projection systems on the market. A lower error rate means that you can rely on the order that players are ranked within the MORPS projection system.
Overall predictive capability was another rating category tackled by The Baseball Projection Project. MORPS didn’t do as well in this category. Upon analysis, this is due to the regression to the mean built into the MORPS engine. This doesn’t have a huge impact on established players. It does have an impact on players with three or less years of experience or players returning from lengthy injuries. We’ll be looking into this further to determine if there is a way to compete effectively in both categories effectively in the future.
Going into 2016, we are confident that our free projection system stacks up quite well with all of the systems out there. This includes those that cost quite a bit of money to access.
MORPS projections are late coming this year. I’ve delivered a set of baseline projections several weeks ago. However, you’ll find that the actual projections have some drastic differences. I’m always amazed by the amount of player movement during the off season.
The Major-League Obie Role-Based Projection System (MORPS) uses four years of player performance data for all hitters. Since I started playing with Sabermetrics using Tango’s Marcel system, the first iteration of MORPS four years ago used the same formulas. After learning the basics, the batter formulas were adjusted to include the most recent four years of performance data. Adjustments were also made for player age, home ballpark data and expected playing time. The most complicated part of the system is the regression formulas. Tango provided formulas for his three year model. I had to crack open the math books to figure out how to transition the formulas to a four year model.
One of the most time consuming tasks in developing the system was determining the proper mean for player regression. If the goal was to ensure that the mean of all the projections competed favorably with end of year player means, the task would have been straight forward. However, my goal was to make the actual player projections as accurate as possible. “Role-Based” means that the player projections are regressed to position specific means. National League means are also separated from American League means.
While conducting research, I noticed that most projection systems used minor league stats as well as any available major league stats to project the future performance of young players. There are even formulas that anticipate player regression when entering the majors. The interesting thing is that Tango’s Marcel system does just as good at predicting rookie performance as other projection systems and he doesn’t use any minor league stats. Some players are great in the minors and simply can’t make the jump to the major leagues. Some players start out great, but find that major league pitchers start exploiting weaknesses they never knew they had. Others outperform all expectations. By calculating the reliability of a player’s projection using only major league data, MORPS adds a proportional dosage of a player’s positional mean to complete a rookie’s player projection. Since we are focused on individual player performance, I didn’t see the point of including all minor league stats when the results don’t seem to provide significant value. The last year of a rookies minor league or international season is included, with appropriate adjustments for competition, if no major league experience exists. While efforts have been made to adjust projections to reflect anticipated playing time, players who have a roster flag of “N” are projected using baseline projections only.
The formulas used to create pitcher projections are very similar to those that we have already discussed with batters. MORPS uses four years of data to create a pitcher projection. Adjustments are made for age, home field and anticipated role. The reliability of a projection is calculated based upon the amount of data available for a particular player. Someone with low reliability will regress more to a position specific mean than someone that has faced a lot of major league batters over the last four years.
The big difference between projecting pitchers and batters is the usage disparity between relief pitchers and starting pitchers. A good relief pitcher may face 350 batters in a season. A top end starting pitcher may pitch to 900 batters in a season. The plate appearances for position players are typically not dependent on role. A first baseman and shortstop may both have 600 plate appearances over the course of a year. Their position means will be different. First basement will typically have higher power stats while shortstops have higher speed stats. But, they are similar enough that their projections can be calculated using the same basic formulas. The disparity between relief and starting pitchers forces them to be calculated very differently. For months I struggled with pitching projections. When I finally figured out that starting pitchers and relief pitchers had to be calculated separately, everything fell in place.
You will notice a new stat category has been introduced that is unique to MORPS – OERV. OERV stands for out earned run value. This new stat attempts to rank pitchers based upon a combination of performance (earned runs allowed) and the number of outs generated for their team. For example, Aroldis Chapman is expected to have a slightly better ERA than Matt Cain. However, Cain is projected to pitch 211 innings compared to Chapman’s 175. As a result, Cain’s OERV is better than Chapman. For those that play rotisserie baseball, a combination stat like this may not have value. You just want to get the best players in each of X specific categories. Head to head simulation leagues, like baseball manager (BBM), use sabermetric calculations to determine daily winners. These leagues will probably find this new stat very useful. This stat attempts to answer the old question that every fantasy manager in these leagues ask on draft day – “When should I opt for a pitcher that eats innings over the pitcher with a lower ERA”.
Every year you see tons of websites predicting which MLB teams have made the right moves to get their team to the playoffs. Some make their predictions based upon their inane baseball IQ. Others use a popularity approach, which teams are getting the most press or the teams that have landed the big name free agents. Perhaps some sites use dart boards or drawing names from a hat. How else can you explain sites that predict the Cubs or Astros getting to the playoffs! Well, we are going to take a little different approach.
Bill James, the pioneer of baseball sabermetrics, created a formula to predict a team’s winning percentage called “The Pythagorean Expectation”. Without boring everyone with the fine details, this formula models the winning percentage of a team based upon runs scored and runs allowed. With anticipated starting and reserve lineups, MORPS has already projected runs created and runs allowed for each team in order to create individual projections. By feeding this data into a refined version of Bill James’ formula created by David Smyth, MORPS team win/loss records can be projected. The records are then adjusted slightly to show the number of games played within each division, league and inter-league matchups. This adds an element of anticipated strength of schedule to a set of formulas created to model the past rather than predict the future.
In the next few days/weeks, I will plan on releasing projected wins and losses for the teams in MLB. I was hoping that all the big name free agents would be off the board at this point, but progress can’t wait on hard-headed agents or budget conscious General Managers. Since my main goal leading up to spring training is to continue updating MORPS for upcoming fantasy drafts, I will begin releasing team projections very soon. When final free agents sign contracts, projected wins and losses may adjust slightly. I may go back and update team projections prior to the start of the season if time permits.
For those that are interested in the detail, I have outlined several the formulas below that are used in team win/loss projections.
The Pythagorean Expectation (developed by Bill James)
Pythagenpat formula (developed by David Smyth)
Exponent = ((r + ra)/g)0.287
Runs Created (developed by Bill James) – calculated for each individual player
The pre-season version of the 2013 MORPS Pitching projections are also now available in excel and pdf formats. Follow the links below to download your copy.
The pitching projections took a little more time to get ready. I’ve added some logic this year that adjusts the historical numbers when a player moves from a relief to a starting role or vice versa. A good relief pitcher and good starting pitcher can have 500 to 600 difference in batters faced over the course of a season. This makes it impossible to directly translate a relief pitchers performance to a starting role or a starter into a relief role. I’m hoping that my new logic solved this puzzle, at least from the point of view of sabermetric projections.
The pre-season version of the 2013 MORPS Batting projections are now available in excel and pdf formats. Follow the links below to download your copy.
MORPS Pitching projections should be available within the next day.
I’ve completed the projection modeling process of 26 teams at this point. Only a few more left. As I was reflecting on this tedious modeling process, I though some might be interested in how it is done.
I’ve already shared how batter and pitcher projections are created. The role based portion of MORPS is the regression to the mean based upon the primary position of a player. Player roles are further refined during the team modeling process by evaluating each player that has been projected to make the 25 man team roster. Free agent acquisitions, retirements, changes in team strategy and a variety of other things will impact the positions of a player during the upcoming season. The art of MORPS projections is analyzing all these movements, team-by-team, and shifting position regression and anticipated plate appearances or batters faced. If I was simply concerned about how MORPS stacks up against other projection systems, I wouldn’t worry about these tweaks. MARCEL projections have shown us that simple projections will be very close to actual when taken as a whole without these adjustments. However, this doesn’t help a fantasy owner who wants to know if they should select player A on draft day. The team-by-team analysis allows MORPS projections to reflect the anticipated roles of a player for the upcoming season. This is also why the projections are updated several times leading up to the season. Roles change based upon Spring Training decisions. Those change then impact individual player projections.
So… How does team-by-team modeling work? The first step is to calculate the total number of plate appearances and batters faced for each team position, batting order position, etc. Although one can argue that a team will never have the same number of plate appearances or batters faced from year to year, we are assuming that the variation in these numbers from year to year will not be dramatic. I then take projected lineups, including backups, and assign them to the position or positions they are anticipated to play. Plate appearances and batters faced are adjusted to reflect anticipated playing time and primary positions are adjusted for those players that are moving to a new role in the upcoming year. Projections are then recalculated for each player on that team using these adjustments. Unlike last year, I’ve written a windows program that simplifies this process. That being said, it is still tedious and time consuming.
When all the team projections are completed, I will then put all the players together into batting and pitching lists in preparation for publication. The last check is to scan position totals for abnormal variances that would indicated mistakes in the adjustments. Those things are then fixed and we’re ready for publication. With only a handful of teams left, I am hoping to have the pre spring training projections out within the next several days.
Let me know if anyone has questions. I figure the more open I am with the process, the more people will be able to trust the end result.
It is generally accepted that the first fantasy baseball format that used live performance results was Rotisserie. It was started in 1980 by Daniel Okrent. The name comes from the place where the league met – a New York City restaurant called “La Rotisserie Francaise”. Rotisserie baseball is typically referred to by the number of batting and pitching categories that are monitored. A 4×4 roto league tracks 4 batting categories and 4 pitching categories. This format became popular because one could easily add up their player’s stats using weekly data information that was published in the USA today. In a 12 team roto league, the top manager would get 12 points in a category they lead, 11 points for second, etc. By adding all the category points together for those players you started that week, you derive the leaders (and eventual winners) of your league. Simplicity is the biggest strength of this format. You can run it just like a fantasy football league. Waivers are processed on one day. Weekly standings are published at the end of the weak. Etc. Etc. That is probably why this has become the largest fantasy baseball format. The negatives are the game does not reflect real individual player performance. Team stats like RBIs, Wins, Losses, etc. are dramatically impacted by the strength of a player’s team in lieu of their individual performance. Roto also doesn’t emphasize the daily nature of baseball. Everything is done on a weekly basis. In other words, the simplification that makes it popular to casual fans has also made it unrealistic for baseball purists. As real-time stats have become more accessible, many rotisserie variations have popped up. Head-to-Head leagues use rotisserie categories to award team wins, losses or ties on a weekly basis. Some leagues have even replaced traditional roto categories with sabermetric calculations.
Before rotisserie, a game called Strat-o-Matic (created in 1963) was the baseball diehard’s game of choice. Many didn’t consider this real fantasy baseball because live stats were not used. It was actually a tabletop game that used historical player data recorded on player baseball cards to simulate games and even seasons. Some sites still exist today that are built upon either historical or fictional player performance. All of these games have their roots in Strat-o-Matic. The pro of this format is that mangers often play a full 162 game season that has the daily feel of major league baseball. The obvious negative is that the game does not use live performance data. Thus, you are not really playing “fantasy” baseball.
In the early 1990s, a new game called Baseball Manager teamed with prodigy to create the first online fantasy baseball game. In order to recreate the real baseball feel of Strat-o-Matic and combine it with the live performance data associated with rotisserie, Baseball Manager chose to use newly created sabermetric projection systems to simulate daily head-to-head baseball games. The game engine combined live performances of players that played that day with banked performances that had not already been used in a head to head game. A fantasy manager’s team offensive performance was calculated by running their players thru the Bill James Runs Created formula. Defensive performances were a combination of pitcher ERAs and player’s fielding performance. A daily head-to-head game would be decided by comparing the offensive and defensive performance of each team. A daily sports page was published each evening for participating managers that read just like your local newspaper. Managers chose their daily lineups based upon pitching matchups (LH/RH); established their pitching rotations three days in advance like a major league manager and even simulated travel to an opposing team’s park to play a 3 game series. Although the game has obviously matured from the days of prodigy (web based), the basic idea of the game is still the same. The major positive of a simulation format is baseball realism on multiple levels. That is also the negative. Managers looking for simplicity or a draft and go type of league should not play simulation leagues. It isn’t a huge time commitment. But, you do have to spend some time on your team every few days. Over the last several years other simulation games have popped up that compete with Baseball Manager. Based upon second hand information, these other formats have resolution formulas that are not based upon sabermetric theory. However, I have not played these leagues so I don’t feel qualified to evaluate them within this post.
The last format that began emerging over the last 10 years is the head-to-head point leagues. These are very similar to roto in the sense that you draft your team and play baseball on a weekly basis. The difference is that the games use a resolution formula to award points for individual player statistical categories in lieu of ranking them from 1 to X. In a sense, they have attempted to use resolution formulas like the simulation leagues while maintaining the weekly format of roto. The positives and negatives of this format are really the same as roto. Simplicity, weekly formats, team based stats, etc. It is just a different way to represent the results.
You’ll notice that I did not discuss the various draft formats. I’ll attempt to cover that subject in another post since it seems that I have been typing on this topic for a while. Once you pick a format, this may be a differentiator for some managers.
So…, without further delay, the questions that you need to ask yourself when you want to choose a fantasy game are as follows:
1. How important is being part of the “main stream” of fantasy baseball?
a. Important – You should play rotisserie
b. Doesn’t Matter – Go to the next question.
2. What is more important to you – Simplicity or Realism?
a. Simplicity favors weekly formats like rotisserie and point leagues.
b. Realism favors simulation leagues.
c. Doesn’t Matter – Go to the next question.
3. Do you get bored with fantasy baseball within a month of finishing your draft?
a. Yes – You may want to try a more realistic version of fantasy baseball.
b. No – Go to the next question.
4. How long do you want to play – Full Season or Partial Season?
a. Full Season fantasy baseball is the standard for all fantasy baseball formats. Go to the next question.
b. There is only one format that offers a partial season – Simulation Leagues. Baseball Manager offers a lightning version that is 54 games in lieu of the standard 162.
5. How important is the level of competition that you play?
a. If you want to play the very best competition, you have three choices.
(1). Rotisserie winner leagues – Some sites allow managers who win within a public league to play in a winner league the next year.
(2). Money leagues – Managers typically only put big money into a league when they are serious about winning. Almost every format has money leagues.
(3). Progression Leagues – This is unique to the Baseball Manager Simulation leagues. In essence this is a fantasy pyramid scheme. There is one Tier 1 league, two Tier 2 leagues, four Tier 3 leagues, etc. If a manager wins a lower level league they are promoted to a higher tier the next year. The two bottom managers at the end of each year in each league are demoted to a lower tier. This is the only format I have played that has consequences for winners and losers. As a result, I have never seen an abandoned team in a progression league.
b. If you just want to win, public rotisserie and point leagues may be your best choice. Many of these sites suffer from abandoned teams and only a handful of quality managers per league. But, winning is certainly much easier in these formats.
c. If you really don’t care or you just want to have fun, you may want to go back and consider one of the other questions when choosing your type of league. You can always find a group of friends to play any format if fun is your goal.
I’m hoping these questions help you choose the fantasy baseball format that best fits your interests. Feel free to offer comments if there are other things that you found important when making your choice of formats.