CPU vs CPU Testing

[dp68]

This applies to the MVP 05 PS2 release.

I hope this will help some people who like me are interested in developing a good set of sliders for CPU vs CPU, whether you are managing or just watching the ole ballgame.

This thread may eventually produce a full set of sliders. Hopefully it will. In the meantime I want you to see the changes I am making to certain sliders and their effect on the gameplay.

Here are some stats to shoot for. These are AL 2006 statistics. I don't intend to hit these perfectly as stats fluctuate year to year anyhow. But it at least tells us what is realistic in modern baseball:

(Plate Appearances=At Bats + Walks)

Batting Average: .275
3.0 Home Runs per 100 Plate Appearances
11.8 Home Runs per 100 Hits
8.5 Walks per 100 Plate Appearances
16.6 K's per 100 Plate Appearances
13.1 Runs per 100 Plate Appearances

[dp68]

The first test uses the default Pro sliders plus the following alterations for both the User and CPU:

Pitch Control -5
Bat Contact -20
Bat Power -20
Bunt Ability 5

I only changed Bunt Ability because it seemed a lot of bunt attempts were being botched. Otherwise I am only concerned about the first 3 slider changes.

At Bats: 405
Plate Appearances: 437

Batting Average .331
Home Runs per 100 Plate Appearances: 5.0
Home Runs per 100 Hits: 16.4
Walks per 100 Plate Appearances: 7.3
K's per 100 Plate Appearances: 9.4
Runs per 100 Plate Appearances: 16.0

The Offensive Output is far too high based on this sample of 437 PA's. Batting Average needs to be reduced by a good 50-60 points. Home Runs occur about 25% too often per hit. Both of these stats explain the high Run output. Strikeouts are about 40% too low also.

On the plus side Walks are reasonable in this sample set.

[bcruise]

Yeah, Strikeouts being too low was always a common problem with this game, be it for the user or the CPU. Most aces usually can't manage more than 5 or so, even guys Like Johnson, Schilling, Oswalt...you get the idea. I think you'll find that the biggest problem is the CPU hitters swinging too early in the count way too often, and putting the ball in play. It pretty much kills all things realistic when that happens, because you don't get K's, walks, and often times a higher number of singles than there should be. You'll probably find yourself getting into the Ball Rate and Swing Rate sliders to fix that problem, and good luck with that - I never did find one combo that actually worked the way I wanted it to.

Good luck with the testing anyway - I just picked up the PS2 version again a few weeks ago (before the Show came out) and still play it from time to time, just like HH2004.

[dp68]

Thanks bcruise. I'll probably integrate those sliders too, but I'll break them in one at a time to see if it is worthwhile.

[dp68]

The second test again uses the default sliders plus these changes:

Pitch Control 0
Bat Contact -50
Bat Power -50
Bunt Ability 8
OF Speed 10


If I wasn't getting enough walks, why increase the pitch control? Well I am concentrating on getting the batting averages in check, so I wanted to give the pitchers a little better control than -5 for now. I also noticed a lot of balls dropping in to the outfield which I thought should have been caught, so I tweaked the OF speed higher.


At Bats: 505
Plate Appearances: 537

Batting Average: .275
Home Runs Per 100 Plate Appearances: 2.4
Home Runs Per 100 Hits: 9.4
Walks Per 100 Plate Appearances: 6.0
K's Per 100 Plate Appearances: 7.4
11.5 Runs Per 100 Plate Appearances


You know, I could actually live with this output. Walks are still reasonable though they need to come up some. K's are again way too low but my guess is I won't ever get those right. But Batting Average is excellent, and Home Runs and Runs came in very realistic if a bit on the low side. My feeling is those numbers were the result of a few well pitched games, and that I don't need to bring them up. Will see.

[dp68]

The third test uses the default sliders plus these changes:

Pitch Control -15
Bat Contact -50
Bat Power -50
Bunt Ability 8
OF Speed 10
Foul Ball Frequency 15


Another small tweak, this time adding Foul Ball Frequency to potentially add more strikeouts.


At Bats: 349
Plate Appearances: 370

Batting Average: .295
Home Runs Per 100 Plate Appearances: 3.2
Home Runs Per 100 Hits: 11.7
Walks Per 100 Plate Appearances: 5.7
K's Per 100 Plate Appearances: 12.4
10.5 Runs Per 100 Plate Appearances


The Batting Averages perked up too high again. Obviously 350-450 Plate Appearances is a nice place to start, but we probably need a least 4x this to get a highly reliable sample. So more work is needed on Batting Averages. Though with Contact at -50 I'm not sure what else I can do. Maybe add some more Outfield Speed?

HR numbers look pretty good. Walks are still too low, but I haven't been mucking with this too much. I can still decrease Pitcher Control and add Ball Rate.

K's look much better. Still low, but we are in the ballpark so to speak.

[bcruise]

I meant to ask, what level are you using for your testing? I noticed that I sometimes got much better looking CPU games at the Pro level, due to the higher number of CPU swing-throughs (on MVP they hit almost anything over the plate, often hard). It still swings early in the count too often, but the swings and misses do help to extend some at bats.

It's easy to tell that the CPU has different AI for each level, try this:

Put CPU Ball Rate at 50 and Swing rate at -50 for both Pro and MVP. Go into a game and observe the CPU's pitching patterns. On MVP you should see FAR more balls, and walks in general, especially of the 4-pitch variety. On Pro, things are very different - you'll still see walks, of course, but not nearly as many - and the CPU swinging more often (even with swing rate at -50)because so many more pitches are over the plate.

I don't know exactly what makes the CPU so much different and tougher on MVP, whether it's juiced ratings or something else. But I do know the AI's decision making does change as you lower the difficulty.

[dp68]

Bcruise -

I started all my sliders at Pro Level, and adjusted from there. Those are good tips. I appreciate it.


The last few tests I have had all kinds of difficulty getting the CPU Batting averages under .300. Even with the Batting Contact at -50, guys are still mashing. So I've been looking for other ways to push the averages down.

[dp68]

Another Test at Pro Level using the default sliders plus:

Ball Rate 25
Bat Contact -50
Bat Power -50
Foul Ball Freq 35
OF Speed 35
IF Speed 25
Catch Effort 10


Honestly I have no idea what Catch Effort is. I am going to assume it is an aggressiveness slider of some sort. Again the the goal here is to lower batting averages. I am forcing more swings to go for fouls, and increasing the defensive speed to track balls down. Also I put Pitch Control back to its default setting to hopefully give pitchers an edge, but increased Ball Rate to compensate with more BB's.


At Bats: 189
Plate Appearances: 206

Batting Average: .286
Home Runs Per 100 Plate Appearances: 2.4
Home Runs Per 100 Hits: 9.3
Walks Per 100 Plate Appearances: 8.3
K's Per 100 Plate Appearances: 8.7
12.6 Runs Per 100 Plate Appearances


This is a smaller sample rate than I prefer. However it was obvious through 3 games that K's were still way too low and Batting Average wasn't budging downward. More changes necessary.

[dp68]

Pro Setting with Default Sliders plus:

Ball Rate 10
Bat Contact -50
Bat Power -50
Foul Ball Freq 50
Swing Freq -15
OF Speed 50
IF Speed 50
Catch Effort 25


I am starting to get desperate. If I cannot bring Batting Average under control now, I have to assume the MVP Engine is just flat out too unrealistic with offensive output. I could of course try a lower setting than Pro to see how the AI changes.

Maxed Foul Ball Freq should put batters in the hole more often. There are still plenty of balls dropping into the outfield which I think "could" have been caught, so let's max out fielder speed and tweak Catch Effort. Also Decreased Swing Freq as everyone seems prepared to mash the first pitch.


At Bats: 407
Plate Appearances: 440

Batting Average: .265
Home Runs Per 100 Plate Appearances: 2.5
Home Runs Per 100 Hits: 10.2
Walks Per 100 Plate Appearances: 7.5
K's Per 100 Plate Appearances: 13.4
10.2 Runs Per 100 Plate Appearances


Now this is more like it! The slider changes did exactly what I intended them to do, so I don't think this is a statistical anomaly. There were more fouls putting batters in the hole, and slightly lowering the ball rate while decreasing swing frequency led to more batters being "frozen" on a well pitched third strike.

All of this got the batting averages under control. They are a little lower than modern era, which itself is higher than historical. So that is a good thing. I doubt I can get K's up to 16.6 per plate appearance, but this latest test at least gets us a reasonable amount. BB's look good, albeit a little low. I doubt I can increase them without blowing up batting avg and K's. HR's and AVG look darn good.

I think I have a set of sliders I can run with. If anyone wishes to add more plate appearances to the database by testing these sliders, please post. Also please share any tweaks as I am most interested.

FYI - I have not looked at stolen bases yet. There may be work to do there.