Thursday, March 23, 2023
Improving WAR - Finding the replacement level
Bill James responded to a discussion we had:
But if we can agree or more-or-less agree about the wins and losses, then we have isolated the problem of finding the replacement level. Once we reach THAT point, where players have won-lost records like 11-19 (Win Shars) or 3.7-6.3 (WAR), then I BELIEVE that everybody will be able to see that players with .200-.300 winning percentages in a season mostly do NOT lose their jobs, and sometimes continue to play at that level for years--thus, that the replacement level has not been properly assessed. Regardless of what the truth is on that level, the point is that we will be able to see. Every discipline is a matter of solving sequential problems. When we resolve one, we move on to the next.
Well, Bill, glad you asked.
Step 1: Gerardo Parra
After the 2018 season was over, Gerardo Parra was granted free agency. In those last three seasons, 2016-2018, he had a total WAR of -1.4. We can represent his performance as an Individualized Won-Loss record (or what I affectionately call The Indis). Parra had a 4-15 record (which is 19 Individualized Games, or iGames, or iG). That's 4 Individualized Wins, or iW, and 15 Individualized Losses, or iL. A 4-15 record is a win percentage of between .200 and .250.
A full season is about 10 iGames, and an average player would have a 5-5 record. Parra has 19 iG in his last three seasons, so the equivalent of 2 full seasons played over three years. He played in 359 games with 1249 plate appearances in 2016-2018. That's about 2 full seasons, which is how he gets 19 iGames.
Let's continue. Parra signed with the Giants before the 2019 season. Since the Rockies were still on the hook for his guaranteed deal, they just had to sign him for the absolute minimum. The Giants released him after playing a little bit, and the Nationals signed him on a free agent deal for the rest of the 2019 season. He did not play in 2020. He played for the Nationals in 2021, and that's the last time he played in MLB. In total, from 2019-2021, he earned close to the league minimum as a free agent. His iW-iL record was 2-4 (6 iGames), which is a win percentage of .333.
Step 2: Prologue
When we talk about replacement level, this is pretty much what we are talking about. You find a player who had a win percentage around .250 to .300 over the previous three seasons, on the idea that most of that was real, and some of it was bad luck. And we expect that player to put up a .300 win percentage the following season (aging notwithstanding). And, this is the important part: we expect such players to be on their last leg, maybe 1 or 2 more seasons left.
And in this illustration Gerardo Parra is representative of this idea. Now, is this idea REAL? Or just a cherry-picked example?
Step 3: The worst players
I looked at all players between 1982 and 2018, who were in their walk year: at the end of that season, they would enter free agency. It's not technically a walk year for all players, since some players will just be outright released, and granted free agency on that basis. There's going to be a selection bias to consider.
In any case, we have 1256 players (non-pitchers) in our study. Of those, 14 had an Indis win% of under .200. Who are these 14 players? Let me pick out a few with a personal connection. We start with Doug Flynn entering free agency for the 1985 season. In the three previous seasons (1982-84, almost all with my Expos), he had a 2-19 record. When your Indis win% is under .100, you are really really on your last leg. Which was the case for Doug Flynn: 1985 was his last season, with 61 total plate apperances with two teams. Based only on Flynn, we suspect that the win% has to be well over .100 for a player to be able to last more than a season or two.
My next player is Mike Laser Lansing, also a former Expos player. He had a 3-13 record, an Indis win% of just under .200. How did Laser do? Well, he never played in MLB again. So, a .200 win% is too low for a replacement player, if we focus only on Lansing.
How about the anti-saber player Joe Carter, hero to all Canadians? In each of 1995, 1996, 1997, he was a below 0 WAR player, even though those last two seasons, he had 100+ RBI in each. His Indis record is 5-24 for an Indis win% under .200. What happened after he hit free agency? Well, 1998 was his last season, he played for two teams totalling just over 400 PA.
Let's look at a 4th and final player, before we do the summary of all 14 players, my saber-nemesis, Ryan Howard. Those who follow my blog know how much I've written about Howard. In his 2014-16 seasons, his Indis record was 3-20. He never played MLB afterwards, even though a couple of teams signed him to minor league free agent deals.
Of these 14 players, the most successful was Billy Hatcher. He had a 4-18 record, but pulled a rabbit out of his hat in 1993, where he was an average player in 136 games. Alas, 1994-95 had him come to bat a total of 340 plate appearances.
So, we can conclude that under a .200 win% is a clear signal that your career is just about over. These 14 players averaged 2 iGames in their first season in free agency, which is 20% of a season. Their total career after entering free agency was an Indis record of 8-38. A .174 win% is not what a team has any tolerance for.
Step 4: The .200 to .250 players
Now, how about the next group of players, those with a .200 to .250 win%? We have 17 players in this group, including Parra. Remember, we are looking at a total group of 1256 players, so these 14+17 players represent less than 3% of all players. They are the really worst players of this time period. These players had an Indis record of .226. How did their careers unwind?
In their first free agent year, none of them were full-time players. The most successful player was probably Mike Matheny, as he followed up his 3-12 record before free agency with a 16-30 record, since 1999. Matheny has the distinction of being one of the worst batters ever, but as a catcher, that can be forgiven with Gold Gloves, which Matheny claimed 4 during this time period. Matheny is probably the exception that proves the rule.
Another player among these 17 is Willie Bloomquist, who is the very face of replacement level. If you play worse than Willie, you are not going to play in MLB. Willie really has had a remarkably long career, but always for a very low salary each year. I believe one time he was able to sign a two-year deal. That's another staple of replacement level players: they only sign one year deals.
Anyway, 9 of the 17 played less than one full season's worth of games. Another 3 played about exactly one full season (Bloomquist, Chris James, Tony Pena), another 4 played 2 or 3. And then Matheny.
Step 5: Interlude
Now, if you want to argue that the replacement level is somewhere between .200 and .250, you could make that argument, based on these 17 players. These 17 players averaged 12 iGames, which is just a shade over 1 full season.
Step 6: The .250 to .300 players
But before we make that proclamation, let's look at the next group, those with a .250 to .300 Indis win%. How did THEY do? Well, we have 43 players. So, now we are up to 74 total players with an Indis record under .300, which is 6% of all the players. These 43 players with the .250-.300 win% actually had a similar rest-of-career as the 17 players with the .200-.250 win%. They averaged 11 iGames, which is just a shade over a full season. They averaged a 4-7 record during free agency.
Step 7: The .300 to .400 players
Indeed, even the NEXT group were very similar. We have 78 players with a .300 to .350 Indis win%. We are now at the bottom 12% of all the players. They averaged 12.5 iGames for the rest of their career, with a .360 win%.
Even the NEXT group after THAT also had a similar rest-of-career. We have 121 players with a .350 to .400 Indis win%. They averaged 13.3 iGames for the rest of their career, for a .361 win%.
As you can see, teams are not really giving much playing time to players at this low a performance level.
Step 8: Epilogue
Remember, earlier I said this:
Now, if you want to argue that the replacement level is somewhere between .200 and .250, you could make that argument, based on these 17 players. These 17 players averaged 12 iGames, which is just a shade over 1 full season.
I will now amend that to say this:
Now, if you want to argue that the replacement level is somewhere between .200 and .400, you could make that argument, based on these 259 players (20% of all players). These 259 players averaged 12.6 iGames, which is just a shade over 1 full season.
Including the 14 really bad players, we have 273 players out of 1256, or 22% of all players. And in the three years prior to entering free agency, these players averaged a .327 Indis win%.
And that's why we use .300 as our baseline level to represent the zero-point of replacement level.
Recent comments
Older comments
Page 1 of 150 pages 1 2 3 > Last ›Complete Archive – By Category
Complete Archive – By Date
FORUM TOPICS
Jul 12 15:22 MarcelsApr 16 14:31 Pitch Count Estimators
Mar 12 16:30 Appendix to THE BOOK - THE GORY DETAILS
Jan 29 09:41 NFL Overtime Idea
Jan 22 14:48 Weighting Years for NFL Player Projections
Jan 21 09:18 positional runs in pythagenpat
Oct 20 15:57 DRS: FG vs. BB-Ref
Apr 12 09:43 What if baseball was like survivor? You are eliminated ...
Nov 24 09:57 Win Attribution to offense, pitching, and fielding at the game level (prototype method)
Jul 13 10:20 How to watch great past games without spoilers