Player Progression Mod v1 Released

[billharris44]

I just uploaded it and it will hopefully be available shortly.

There is one disappointment, which is that walks are going to decline roughly 25% in simmed games over 20-30 years, and there doesn't appear to be anything that can be done to fix it. The A.I. doesn't appear to factor in plate discipline when evaluating players. So as a Dynasty progresses, you'll see 1-2 guys a team with plate discipline below 30. Plate discipline is tied very speficially to walks in simmed games, so these guys will be drawing 15-20 walks in a season.

It won't affect your team, because you'll be using that rating as part of your evaluation. And it doesn't seem to have a significant effect during played games. But it definitely, significantly affects simmed games.

Plate discipline is a unique rating because it usually peaks late in someone's career, as opposed to almost all other ratings peaking in the 25-27 range. But there's only one probability table to decide if a rating goes up or down at a particular age. So just as plate discipline should be going up, it's actually going down with the other ratings.

I feel pretty badly about this, because everything else looks stellar. And I'm still hoping to find a fix, but so far I've drawn nothing but blanks. The discipline.csv table, which seemingly would affect the distribution of the plate discipline rating for created players--doesn't. I'll keep looking and I'm open to ideas.

Here's an excerpt from the read me with statistical comparisons.

****

6. STATISTICAL COMPARISON

I evaluate a player progression system in the following areas:

A. League totals should remain relatively accurate over time.

B. Team ranges. It's important that there is an appropriate spread between the league leading team in that category and the worst team. Accurate league totals are great, but if the league leading team hits 400 home runs and the worst team hits 100, it's wrong.

C. League leaders. The Top 30 in the most popular statistical categories (BA, HR, ERA) should be in a range that's consistent with real major league statistics, and they should be stable over time.

D. Age distribution. Age distributions for position players/pitchers should be as close as possible to real major league distributions.

We tested with the following conditions: Difficulty Level--All Star, Default Roster--Yes, Injuries--Off, Budgets--On, Energy effect--On. Those are the only settings that should have any possible effect on stats. As I mentioned earlier, difficulty level affected simmed stats last year, increasing offensive stats by about 5% if using Pro level instead of All-Star.

Here's a statistical comparison.

A. League totals

After one season:

STAT MLB2004 DEFAULT2005 MOD2005

Hits; 44,500; 45,786; 45,783

Runs; 23,376; 21,675; 22,759

HR; 5431, 5064; 5533

2B; 8918; 7924; 9558

3B; 898; 918; 943

BB; 16,222; 18,035; 18,064

SO; 31,828; 32,392; 32,691

SB; 2589; 2223; 2050

What you'll immediately see in the first season is that the CPU doesn't score runs efficiently. Even with more total bases than MLB2004, the CPU scores less runs.

As for accuracy, it looks like this:

MLB2004 DEFVAR MODVAR

Hits; 44,500; 2.89%; 2.88%

Runs; 23,376; -7.28%; -2.64%

HR; 5431; -6.76%; 1.84%

2B; 8918; -11.15%; 7.18%

3B; 898; 2.23%; 5.01%

BB; 16,222; 11.18%; 11.35%

SO; 31,828; 1.77%; 2.52%

SB; 2589; -14.14%; -20.82%

The mod is more accurate, but the default settings are not bad--far better than usual for a non-text sim. By the way, the one anomaly in all this data that I can't explain are Year One steals. I didn't touch anything related to stealing frequency, so the numbers should be very similar (like BB and SO, which can't be adjusted with park factors), but steals in the mod are almost 10% lower.

After 20 seasons:

STAT MLB2004 DEF2024 MOD2024

Hits; 44,500; 49,329; 44,520

Runs; 23,376; 23,969; 20,542

HR; 5431, 7035; 6086

2B; 8918; 8492; 9296

3B; 898; 932; 819

BB; 16,222; 15,729; 13,164

SO; 31,828; 32,454; 33,530

SB; 2589; 2754; 1842

And the variance (from MLB2004):

STAT MLB2004 DEFVAR MODVAR

Hits; 44,500 10.85% 0.04%

Runs; 23,376 2.54% -12.12%

HR; 5431 29.53% 12.06%

2B; 8918 -4.78% 4.24%

3B; 898 3.79% -8.80%

BB; 16,222 -3.04% -18.85%

SO; 31,828 1.97% 5.35%

SB; 2589 6.37% -28.85%

Home runs at the default settings go through the roof--up over 35% from season one. It will usually take 60+ home runs to lead the league after about 15 seasons. I looked at the player ratings and it's easy to see why. Most players have at least 80+ for power on at least one side, and there are a ton of 99/99 (contact/power) guys. So the ratings distribution itself is way, way off after 20 seasons. Almost every team has eight players with 20+ home runs.

Home runs go up with the mod as well, but less than 10% from season one, and it's planned, to help compensate for the inefficency of the CPU in scoring runs. 2B are down about 3% from season one, 3B are down about 12%. Runs are down about 10%. Again, you have to pick your poison on runs. It's possible to get them up to MLB2004 levels, but Hits, HR, 2B, and 3B would have to be cranked up well beyond real levels to compensate for scoring inefficiency, as well as to compensate for the decline in walks. The default settings are very close on runs, but you'll have 30 guys in the league hitting 40+ home runs every year, which is crazy.

You can see the devastating effect of walks going down. The default settings get more walks because all offensive ratings are cranked up to heroic proportions, including plate patience, but even then they decline slightly. The A.I. should factor in plate patience when evaluating players, but as far as I can tell, it doesn't. And with no separate table for plate discipline advance/decline, there's no way to increase it without increasing everything else, which will unbalance everything else.

One note about steals. I could adjust team's steal tendencies to generate 2500 steals leaguewide. The problem is that I can't adjust the success rate. The steal success rate is correct in the first season, because with hand-made ratings, base stealing aggressiveness can be tied to speed, but the player creation doesn't make that association, and the ratings tables don't allow linkage. So the success rate goes down steadily and is in the 55% range. That means that steals are not improving run production--they're actually decrementing it. So I decided to leave it alone.

After 40 seasons (I didn't run the EA default for another 20 years)

STAT MLB2004 DEFAULT2024 MOD2024

Hits; 44,500; x; 45,007

Runs; 23,376; x; 20,358

HR; 5431, x; 6149

2B; 8918; x; 9359

3B; 898; x; 901

BB; 16,222; x; 12,503

SO; 31,828; x; 33,645

SB; 2589; x; 1883

After 25 seasons, all default players should be cleared through the system, and I believe that stats should be stable from around season 30 onward. So what you see at season 40 should be true for seasons 30-100 (or as long as you wanted to sim). For everything but walks, anyway. I went ahead and checked BB in year 50 of a dynasty and the total was 12,150. So, amazingly, walks are still dropping slowly. What happens is that an average of two starters a team will have plate discipline of 35 or lower. Those two guys will draw 15-20 walks in a season instead of 75-80. So that's 3,6000 lost walks. They don't strike out excessively, though, because plate discipline seems to directly affect walks only for simulated games. It's driving me crazy, but without the A.I. considering plate discipline in player evaluation, and without being able to find the creation table for plate discipline, I can't think of a fix. There is a discipline.csv file, but I modified it and it didn't affect the plate discipline of created players.

The good news is that you, as human manager, can use plate discipline in evaluating players, so you should still get the normal number of walks in simmed games. And people on the test team who have played games haven't reported any problems in on-field playability, although it's a small sample.

B. Team Ranges

STAT MLB2004 DEF2005 MOD2005

Hits; 1614-1358; 1690-1444

Runs; 949-615; 1016-542; 948-626

HR; 242-135, 242-120; 231-134

2B; 373-226; 312-192; 397-282

3B; 54-15; 50-14; 47-16

BB; 705-415; 786-462; 806-515

SO; 1335-874; 1245-999; 1187-995

SB; 143-44; 157-57; 117-39

Those huge team ranges (particularly with runs and home runs) with the default settings are due to park factors. I modified them by taking park factors from ESPN.com and translated them into the settings the table uses to create more realistic team ranges.

Here's a look in 2024:

STAT MLB2004 DEFAULT2024 MOD2024

Hits; 1614-1358; 1889-1392; 1578-1309

Runs; 949-615; 1148-657; 826-561

HR; 242-135, 339-157; 274-123

2B; 373-226; 344-229; 395-238

3B; 54-15; 65-11; 41-15

BB; 705-415; 776-387; 603-309

SO; 1335-874; 1189-847; 1326-971

SB; 143-44; 157-57; 83-37

C. League Leaders

Look at this comparison of the range of the Top 30 leaders after 2005 and 2024.

STAT MLB2004 DEF2024 MOD2024

BA; .372-.305; .375-.331; .352-.302

HR; 48-31; 51 to 38; 50-34

ERA; 2.27-3.93; 2.07-3.78; 1.88-3.18

SO; 290-153; 210-160; 213-178

SB; 70-19; 44 to 21; 34-16

You will see the league leaders cycle in 3-5 year periods. The ranges you can expect to see are as follows:

BA: Leader .340-.360, 30th .302 to .310

HR: Leader 44-52, 30th 32-34

ERA: Leader all over the place (exceptional performances in certain years), 30th 2.97 to 3.24

Like I said, these cycle. You may think the league is going down in HR over a 2-3 year period, for example, then you'll see several years with 50+ league leaders. Averaged out, they are very reliable over time. You will very rarely see a 60+ HR season or a .380 BA, but they happen.

D. Age Distribution

POSITION PLAYERS

AGE MLB2004 DELT2029 MOD2029

19-22; 4.97%; 3.33%; 3.25%

23-25; 15.94%; 20.00%; 19.25%

26-28; 28.65%; 30.00%; 29.75%

29-31; 23.39%; 19.52%; 24.25%

32-34; 15.06%; 18.81%; 15.00%

35-37; 8.48%; 6.43%; 7.25%

38-40; 2.92%; 1.43%; 1.25%

40+; 0.58%; 0.00%; 0.00%

Nothing to complain about with the default definitions here. The mod is better, but both are relatively accurate. One thing I've noticed is that the 19-22 age group will tend to have a higher percentage of players (8-9%) until all the default roster plays work their way out of the system. Distributions also move around in the first 25 years, for the same reason. Finally, distributions will change each year. There's an average percentage for each age group over time, but there will be 2-3% variations above and below that as the percentages cycle around. I'm guessing that with the player creation variables allowing some degree of latitude, there may be Superstar/Joe Shlabotnik classes.

PITCHERS

AGE MLB2004 DEFAULT2029 MOD2029

19-22; 3.76%; 12.43%; 5.38%

23-25; 20.06%; 25.44%; 27.20%

26-28; 31.48%; 24.26%; 30.59%

29-31; 20.06%; 15.68%; 16.43%

32-34; 12.95%; 11.83%; 13.31%

35-37; 8.22%; 7.99%; 5.38%

38-40 2.51%; 1.78%; 1.70%

40+; 0.97%; 0.00%; 0.00%

The pitcher distributions are not as good for either the default or the mod, and that's tied to the 1-15 ratings increase tables not being hooked up. Since pitches go up 1-5 points a year and no more, pitcher progression is much less sophisticated and much more linear. This was exceptionally difficult to adjust. After twenty seasons, I had a significantly better distribution:

MLB2004 Mod2024

19-22; 3.76%; 5.38%

23-25; 20.06%; 27.20%

26-28; 31.48%; 30.59%

29-31; 20.06%; 16.43%

32-34; 12.95%; 13.31%

35-37; 8.22%; 5.38%

38-40; 2.51%; 1.70%

40+; 0.97%; 0.00%

However, over the next five years as the last default players go out of the system, it tilts younger. And with only the UP/DOWN/SAME rolls to adjust to control the change, the only way to fix this would require me to rework all the UP/DOWN/SAME rolls for position players as well (because overall pitching ratings would have to go down). As it is probably not even noticeable during the course of a Dynasty, I'm going to leave it alone for now.

*****

[The Show]

I'm confused. In all your posts you say that it's to be used in Dynasty Mode but can we use it in Owner Mode?

Edit: I read the readme and it says that you tested with the default rosters. If I use the Ultimate Rosters, will the progression system change?

[stahr]

yea I did, ooops

[poonani]

says you added justin Sherrod(boston) in your readme, but I cant find him.

i think you're posting in the wrong thread.

[jhath]

i gotta start my dynasty over for this to work... correct?

[jhath]

anyone?

[Goldensuitcase]

I'm confused. In all your posts you say that it's to be used in Dynasty Mode but can we use it in Owner Mode?

Edit: I read the readme and it says that you tested with the default rosters. If I use the Ultimate Rosters, will the progression system change?

Shouldnt. This fixes stats internally, not where you can see and play with in in-game play.

[billharris44]

1. Yes, you can use it in owner mode. It should work fine.

2. It should also work fine with other roster sets.

3. You do not need to start your Dynasty over. The simulated stats would be more accurate if you started with a new dynasty, but half a season or something makes very little difference in the first year.

[cubby]

Amazing work. Thanks so much for this. EA should hire you on the spot~