2063 - Marcel Projections: 2063 Batters Edition

[Bob Breum]

In BBA Today #104, Kurt expressed a desire to incorporate a stats-based projection system within OOTP. He mentioned ZIPS as his favorite. This sent me on a mission to see if that was feasible to implement.

First I surveyed the current projection systems. These include proprietary systems such as ZIPS (Baseball Think Factory), PECOTA (Baseball Prospectus), Steamer (FanGraphs), and Oliver (The Hardball Times). These are all closed systems that do not reveal their algorithms.

Tom Tango, the best sabermetrics guy not named Bill James, wrote his own system and made it open source. He named it the Marcel the Monkey Forecasting System. In concept, it is most similar to Steamer. In independent testing, it has compared well with the commercial systems despite its rudimentary calculations.

I have used Marcel to create these projections. It uses a weighted average of the last three seasons with some regression to the mean and an adjustment for age.

Note that rookies with no previous BBA exposure are omitted. Only BBA-level stats are used.

If you want to understand the methodology, this is the best source that I have found:

https://www.beyondtheboxscore.com/2016/ ... ath-is-fun

Since I'm the GM of hitting, I did batters first. I'll be releasing the pitching version soon.

Here are the projections in alphabetic order:

Marcel.pdf

(168.53 KiB) Downloaded 8 times

And here they are by team:

Marcel by Team.pdf

(201.08 KiB) Downloaded 15 times

Here it is in .csv form:

Marcel.csv

(51.78 KiB) Downloaded 4 times

It is possible, even likely, that the dataset has an error or two, especially with regard to players with identical names. Please bring them to my attention.

[BaseClogger]

Really cool of you to take this idea and run with it—thank you. I didn’t realize there’s an open source option. I do wonder if league equivalents could be calculated and allow us to do a similar thing with minor league stats. Sadly, those are often the most interesting projections.

I’ll spend some time looking at the files in more detail as I have more time… busy weekend.

[Bob Breum]

I didn’t realize there’s an open source option.

It's open source in the respect that he published the calculations. You still have to implement those in some fashion. I used Excel.

[woods]

This is great, but Richard Vernon hitting 5 home runs? I can only suspend my disbelief so far.

[Bob Breum]

This is great, but Richard Vernon hitting 5 home runs? I can only suspend my disbelief so far.

At first I thought you were suggesting that the projection was too low but then I pulled up Mr. Vernon's stats. TBF, he is now 27 and the engine awards slight bumps until a player hits 29. He did hit four in 2061.

[CTBrewCrew]

Cool thanks for posting this.

[RonCo]

Much fun. Just getting to really look at this. I'd say Marcel does not like Bikini much.

[ae37jr]

This is fun.I don't get why the game doesn't do this. It obviously figures out preseason predictions and lists top 10 pitchers and hitters. I always wished they would have a page where you could see player predictions by team.

[Bob Breum]

I implemented the calculations in Excel. Since then, I discovered that there is Python source code available from Jeff Sackmann on GitHub if you want to play with it.

https://gist.github.com/JeffSackmann

You'll want generateBattingMarcels and generatePitchingMarcels.

Jeff Sackmann is an author and software developer who has worked in the fields of sports statistics and test preparation.