Split issues

[Murcer]

I wrote this in the patch thread but thought I should create a separate post.


I've run a number of 300+ season tests to check splits and still see some issues post-patch. Here are the normalized OPS results I'm getting in OOTP versus real-life MLB numbers from the 1980s (I chose 80s because my leagues roughly replicate that era):

R vs R
Real-life = .684
OOTP = .702
dif = 17.8 points

R vs L
Real-life = .729
OOTP = .730
dif = 1.0 points

L vs R
Real-life = .735
OOTP = .729
dif = -6.4 points

L vs L
Real-life = .663
OOTP = .670
dif = 7.3 points

Summary
Right handed hitters are still doing far too well against right handed pitchers which is skewing overall results. An easy way to check this is to look at your career OPS leaderboard and you'll notice a lot of righty hitters at the top of the list.

Left handed hitters are close but might need to be tweaked as well. I realize 6 or 7 points in OPS is not a lot, but it's consistently off about that much in every test I've run over millions of plate appearances.

[jcard]

Appreciate you sharing. Do you have any observations as to the respective roles played by…?:

A) EngineDesign / Systemic
B) Player ratings / Inputs
C) Roster and Game Management / Strategic

[kidd_05_u2]

There are two problems with your post Murcer.

1) You are comparing against an arbitrary point when OPS fluctuates a lot according to the offensive environment. To check if there are problems, you need to look at the gap between the four scenarios. A righty batter has an OPS 5% higher against lefty pitchers than against righty pitchers. Does this hold? A lefty pitcher has an allowed OPS x% higher against righty batters compared to lefty batters. Does this hold?

2) You compare against the 80s, but the OOTP splits target output in the modern game. Splits were more pronounced in the 80s. There are no historical player creation modifiers for splits (afaik).

[Murcer]

To the points/comments above:
I went out of my way to use a fairly generic set-up and let the computer manage all teams/rosters so I don't think it has much to do with setup. Personally, I believe it has to do with the game-generated player ratings because the dev team tweaked ratings with the latest patch and it greatly improved the split differences.

For the differences in OPS by hand:
1980s MLB
Right handed hitters OPS was 6.37% better against left handed pitchers
Left handed hitters OPS was 10.3% better against right handed pitchers

2022-2024 MLB
Right handed hitters OPS was 5.68% better against left handed pitchers
Left handed hitters OPS was 9.21% better against right handed pitchers

My OOTP Tests
Right handed hitters OPS was 4.03% better against left handed pitchers
Left handed hitters OPS was 8.38% better against right handed pitchers

So while the split differences are slightly lower in recent seasons versus the 1980s, they aren't that much different. I still hold that the split difference for righty hitters needs to be increased in OOTP to better represent MLB.

Again, the issue becomes more apparent when you look at your OPS leaderboard. In my latest test that spanned 340 seasons, 11 of the top 15 hitters in career OPS were right handed, 2 were lefties, and 2 were switch hitters. If you look at the active career leaders right now, the top 15 are:
8 right handed
6 left handed
1 switch hitter

[kidd_05_u2]

Originally Posted by Murcer

Again, the issue becomes more apparent when you look at your OPS leaderboard. In my latest test that spanned 340 seasons, 11 of the top 15 hitters in career OPS were right handed, 2 were lefties, and 2 were switch hitters. If you look at the active career leaders right now, the top 15 are:
8 right handed
6 left handed
1 switch hitter

There are several different things in OOTP that could explain this, including randomness, so it doesn't necessarily help your arguments on splits as much as you seem to think. Maybe the game generates too few lefty batters? Maybe there are too many lefty pitchers? There are other factors, and you are hardly looking at a great benchmark to begin with.

In fact, your other calculations may show that righties don't have enough of a split, but I'm not sure they say enough about righties being better than lefties overall. Can you access what is the OPS of left handed batters and what is the OPS of right handed batters overall?

[kidd_05_u2]

Number of plate appearances vs each side would help too.

[Brad K]

Originally Posted by Murcer

To the points/comments above:
I went out of my way to use a fairly generic set-up and let the computer manage all teams/rosters so I don't think it has much to do with setup.

Testing something like this needs to be done with a replay league setup. Not saying you're wrong. I believe it's likely you're right. Only saying that this isn't the data to prove it.

[Murcer]

I'm open to multi-causal explanations, all I'm saying is that right handed hitters are over represented in the career OPS leaderboards. And given the number of variables involved, I decided to point out the most easily measurable stat - OOTP split differences versus typical MLB.

I've run numerous 300+ season tests and in every one of them, right handed hitters dominate in OPS.

A comparison of the top 100 career OPS, my last OOTP test (covering almost 40 million PAs) versus MLB:
Right handed hitter counts: OOTP 63 / MLB 44
Left handed hitter counts: OOTP 26 / MLB 52
Switch hitter counts: OOTP 11 / MLB 4

Given how all the variables interact, I can't say with certainty what is causing this discrepancy. However, I can say with some confidence that the resulting split differences for right handed hitters in OOTP are off as compared to real-life.

[kidd_05_u2]

You've done a great job in identifying two issues.

To more conclusively draw the arrow between the two, we need the PA data. Otherwise, we can't say if the leaderboard problem is caused by something different but plausible, like right-handed batters being overrepresented in the major leagues.

[Brad K]

Originally Posted by kidd_05_u2

You've done a great job in identifying two issues.

To more conclusively draw the arrow between the two, we need the PA data. Otherwise, we can't say if the leaderboard problem is caused by something different but plausible, like right-handed batters being overrepresented in the major leagues.

If the tests were done with a replay league this possibility would be greatly reduced.

As I posted before, I expect he's right but doesn't have enough of the right stuff to prove it.

[kq76]

I'd just like to say I appreciate you looking into this and getting others to think about it. Like others, I'm not sure you're going about it the best way, but I've also had a feeling for a little while that left-handed players weren't getting their due so it's interesting seeing someone actually put up some numbers towards that and starting the conversation.

[Murcer]

I've never used replay leagues and didn't use them to test this because I assumed the splits were based on real MLB numbers. Therefore, it wouldn't help prove my point about game-generated players.

PA stats, latest test league vs 2022-2024 MLB:
Total PA by righty hitters: OOTP 69%, MLB 59%
Total PA by lefty hitters: OOTP 31%, MLB 41%

Right handed hitters get more PAs in OOTP than they do in MLB. I saw this in earlier tests but couldn't explain why given all the variables. That is - are the righties getting more PAs because their ratings were too high (especially vs righty pitchers) leading the game engine to give them more PAs? Or is there something else going on - too many right handed hitters, too many good righties, etc?

[Brad K]

It seems to me the starting point is to see if the game engine gets splits right with real players and as many variables as possible removed. That points to replay. I never do replay as a save as for me there is no thrill in seeing if OOTP can duplicate history but it's a great setup for tests.

[Murcer]

Originally Posted by Brad K

It seems to me the starting point is to see if the game engine gets splits right with real players and as many variables as possible removed. That points to replay. I never do replay as a save as for me there is no thrill in seeing if OOTP can duplicate history but it's a great setup for tests.

If anyone has a long-term replay league, I'd be happy to look at the results. I'd just need the csv extracted files:
players_career_batting_stats
players

[kidd_05_u2]

Originally Posted by Murcer

PA stats, latest test league vs 2022-2024 MLB:
Total PA by righty hitters: OOTP 69%, MLB 59%
Total PA by lefty hitters: OOTP 31%, MLB 41%

Right handed hitters get more PAs in OOTP than they do in MLB. I saw this in earlier tests but couldn't explain why given all the variables. That is - are the righties getting more PAs because their ratings were too high (especially vs righty pitchers) leading the game engine to give them more PAs? Or is there something else going on - too many right handed hitters, too many good righties, etc?

Well, if you want to keep being awesome, you could compare the share of RHB and LHB produced by OOTP in the draft and IAFA vs what we see in real life.

Just as a reminder, we have three plausible explanations, all of which could be true to a degree:
1) The ratings splits are helping right-handed batters get more ABs and perhaps perform better (I need to check your earlier numbers)
2) The game is producing too many RHBs vis-a-vis LHBs in player creation.
3) OOTP managerial decisions are leading to RHBs getting more PAs than they do in real life.

[Murcer]

I thought about comparing OOTP drafts with real-life drafts but that would require a lot of manual work.

At this point, I'm tapping out. I started this thread hoping the dev team would take an interest as they can probably get a better/faster grasp on what's going on given their access to the game code. For example, I'm sure there is an algorithm controlling the number and quality of R/L/S hitters when new players are generated.

If anyone has questions about the data from my test leagues, I'm happy to help.

[kidd_05_u2]

Originally Posted by Murcer

I thought about comparing OOTP drafts with real-life drafts but that would require a lot of manual work.

At this point, I'm tapping out. I started this thread hoping the dev team would take an interest as they can probably get a better/faster grasp on what's going on given their access to the game code. For example, I'm sure there is an algorithm controlling the number and quality of R/L/S hitters when new players are generated.

If anyone has questions about the data from my test leagues, I'm happy to help.

Thanks for what you've tested and shared. Matt adjusted ratings for splits twice in the last year, so I'm sure he will read and react if he feels more refinement is needed.

[Lukas Berger]

Originally Posted by Murcer

I thought about comparing OOTP drafts with real-life drafts but that would require a lot of manual work.

At this point, I'm tapping out. I started this thread hoping the dev team would take an interest as they can probably get a better/faster grasp on what's going on given their access to the game code. For example, I'm sure there is an algorithm controlling the number and quality of R/L/S hitters when new players are generated.

If anyone has questions about the data from my test leagues, I'm happy to help.

Originally Posted by kidd_05_u2

Thanks for what you've tested and shared. Matt adjusted ratings for splits twice in the last year, so I'm sure he will read and react if he feels more refinement is needed.

Yes, this is good stuff, thanks so much for posting. I know we have some split related things in the roster set on the list to potentially work on for 27, so can likely take a closer look here as well.