CPU plate discipline logic has been re-written

[Brian SCEA]

Re: CPU plate discipline logic has been re-written

Exactly. I feel like this point needs to be driven home. Its not the strikeout rate that needs work, its how those at bats end up in a strike out.

Yes, just getting the number of strikeouts right is aiming low. To list some examples..Does the CPU strikeout looking enough? Does the CPU strikeout chasing or in the zone? Does the CPU get to strike 2 (all four counts) correctly? Does the CPU walk enough? Do the different types of hitters strikeout and walk correctly? How about the different pitch types and their break/location? Does the CPU adapt/anticipate to gain an advantage, or be fooled because of it? If the Human pitches differently than a CPU, how does this affect the CPU hitter?

When evaluating tuning, it's quite common for one question to be correct with all other questions wrong. I would rate tuning more on how close everything is, weighted by priority, rather than how close just one factor is.

Also, we don't want the CPU hitter to work the count right just because the CPU pitcher happens to pitch "normally" most of the time for him. It's important that if you decide to throw more strikes (maybe it's a blowout and you're saving your bullpen for next game), that the CPU hitter gets a hitting advantage in exchange for an increase in strikeouts and decrease in walks. Same vice versa. This is as important as for things to be right on average.

In other words, shouldn't what you do matter a lot in terms of what happens? For many people, what happens in a HUM vs CPU game when you try different things is arguably more relevant than what happens in passive observation in a CPU vs CPU game. The latter by itself is almost trivial, and this is the difference between an interactive game versus merely a simulation that breaks the moment you do something a CPU wouldn't usually do. Every decision and action matters.

Certainly, it's a lot easier to verify that CPU vs CPU behaves correctly than HUM vs CPU, but that doesn't mean we should make single player or even two player a lower priority in tuning. If one seems to work but the other doesn't, it's a good sign that both aren't working as well as they could.

A lot of things besides CPU discipline were looked at and adjusted, same as every year. In any case, I only wanted to clarify what the bullet point was referring to. When the game is out, you can decide for yourself.

[tessl]

Re: CPU plate discipline logic has been re-written

No one is even questioning strikeout totals. The post that you quoted had no mention of them. We are talking about chase rates. The strikeout totals in this game are pretty accurate, but that doesn't mean the AI is achieving them in a realistic manner. The end does not justify the means. Those realistic strikeout rates are being complimented by unnatural strike percentages, diminished chase rates, and other such oddities.

I have confidence in Brian to figure it all out but it seems like if you increase the chase rate and decrease the cpu ability to know pitch location you increase strikeout totals.

If it is all being done to increase realism I'm 100% in favor of it but if it is being done to placate joystick users who want to strike out cpu batters with pitches a foot off the plate then I can see that causing problems for manage mode where every batter is a cpu batter.

[Brian SCEA]

Re: CPU plate discipline logic has been re-written

I have confidence in Brian to figure it all out but it seems like if you increase the chase rate and decrease the cpu ability to know pitch location you increase strikeout totals.

A lot of other things were tuned, and they affect strikeouts as well. If we started listing these, they would probably cause more confusion than answer questions. Don't read too much into marketing bullet points, because they exist to raise awareness and excite consumers rather than document the whole development process.

[Heroesandvillains]

Re: CPU plate discipline logic has been re-written

I have confidence in Brian to figure it all out but it seems like if you increase the chase rate and decrease the cpu ability to know pitch location you increase strikeout totals.

If it is all being done to increase realism I'm 100% in favor of it but if it is being done to placate joystick users who want to strike out cpu batters with pitches a foot off the plate then I can see that causing problems for manage mode where every batter is a cpu batter.

To go along with Brian's posts Tessl, like I was saying earlier (and as Brian reiterated just now), the game IS tuned for CPU vs. CPU play. He looks at these numbers (I know this from experience) so I wouldn't worry too much if I were you.

As I said earlier and as Brian again just reiterated, the game is also tuned and refined based on Human players as well. It has to be.

Last year at CD, their was a specific issue caught from the QA team. They recognized this and it was forwarded to Development and it was thereafter addressed. Human interaction is equally as important to the strength of the code and instances like this prove that.

For the sake of a playable baseball simulation, both sides need to be working together. The AI is adaptive, and is programmed to recognize patterns.

Also keep in mind that the hitting and pitching AI is blind (meaning that neither side knows what the other is doing). This is an important point as it means CPU tests are a reliable way to gauge the statistical accuracy of the system. But it obvious can't be the only way (most people play the game).

Both HUM and CPU testing work in harmony. If one's 'off key,' so to speak, adjustments get made. And yearly, these refinements have made the CPU side closer to real life.

Again, I expect this year to be no exception to that.

[Bobhead]

Re: CPU plate discipline logic has been re-written

I have confidence in Brian to figure it all out but it seems like if you increase the chase rate and decrease the cpu ability to know pitch location you increase strikeout totals.

If the game is designed by a gerbil, sure. But I'm pretty sure we have a team of very competent humans, and I'm going to go ahead and assume these guys are capable of changing things without breaking the game.

Besides, if the AI is made to chase more balls, it would follow that they would chase more strikes, too (after all, one swings at a ball only because they perceive it to be a strike).

Thus, yes, increasing swings equates to strikeouts on chases increasing.
but increasing swings ALSO equates to strikeouts on called strikes decreasing, by roughly the same number.

So if you really want to be technical, the net change in strikeouts, even without any other changes or re-balancing, would not be significant.

[tessl]

Re: CPU plate discipline logic has been re-written

A lot of other things were tuned, and they affect strikeouts as well. If we started listing these, they would probably cause more confusion than answer questions. Don't read too much into marketing bullet points, because they exist to raise awareness and excite consumers rather than document the whole development process.

Thanks for the replies, that explains it and thanks for fixing the problems with manage mode for mlbts 13. Looking forward to 6 May.

[tessl]

Re: CPU plate discipline logic has been re-written

If you are curious, you can click on "CPU vs. CPU games" link in my signature and see them for yourself. Games are all played out by CPU in season mode. I vary sliders but always do a set of about 75 games at default sliders. The column BA (O-Swing %) is the chase %, the first two rows are the MLB ave in blue and measured in black.

I understand you only seem to care about what Brian says, but he and you are not the only participants here, and other people are also here for sharing conversations. Knowing that the developers are partly here to gather feedback, there is absolutely nothing wrong with having constructive feedback.

I wouldn't say I only care what Brian says but he is the lead AI programmer for the game and the only person in this thread who has both played and has the ability to adjust mlbts 14 so his comments do carry significant weight. His posts in the past regarding which attributes most impact the game are the most informative posts I've seen on this forum.

I commend you for your research. I looked at your chart and I have a question. Are those only for games with default sliders?

[rjackson]

Re: CPU plate discipline logic has been re-written

It's important that if you decide to throw more strikes (maybe it's a blowout and you're saving your bullpen for next game), that the CPU hitter gets a hitting advantage in exchange for an increase in strikeouts and decrease in walks. Same vice versa. This is as important as for things to be right on average.

This is a relationship that a more sabermetrically inclined baseball fan understands and then using this data to fine tune this game...add in a couple of initiative driven individuals with a vision of what franchise mode will look like in a few more years and Bam! Best baseball sim EVER.

I'm genuinely impressed and want to give give you a sincere thank you.
BRAVO! BRAVO!

[Armor and Sword]

Re: CPU plate discipline logic has been re-written

Another thing that for me plays into CPU batter chasing etc is their plate vision and discipline.

I feel this plays right into my hands for certain batters with certain pithcers and certain locations based on hot and cold zones etc.


Again....last night playing game three in my ALCS I was able to get Ken Camaniti to chase at a 1-2 change almost in the dirt down and away. It was beautiful and something I never saw with any regularity on 11 or 12 but in 13 I have no doubt seen and felt the difference.


It's all about the set up, the hitter, the situation, the pitch type and truly what pitching interface you use as well! My goodness so many varibles.....and most importantly the human imput. I don't play cpu vs cpu.....so hum vs cpu is about patterns, habits, or being able to truly mix it up...work pitches....does my guy have his best stuff today? Is the hitter on a cold streak? ratings vs Lefty or righty, so many variables in the hitter/pitcher showdown.


The Show does it better than any baseball game.....and really....it's not even close.


Thank you so much to Brian and the team for giving us such an incredible baseball game. I have waited my entire video game life (dude...I made Intellevision Baseball fun for myself by keeping stats, that is how much I love to play and watch the game of baseball) for this kind of console baseball game.....and you guys keep striving higher and higher.

[tessl]

Re: CPU plate discipline logic has been re-written

I don't have the game but I'm wondering if you guys who wanted more batters chasing bad pitches are seeing that.

[nomo17k]

Re: CPU plate discipline logic has been re-written

I have only run a couple games so far, but I'm just starting to look at this. I'll be using a couple of pitchers I've always used in the past few years to see how

I think it's pretty clear CPU plate discipline has been tweaked quite a bit. How CPU reacts to pitches has been getting more and more natural every year. It's very beautiful the way CPU hitters get fooled by breaking/off-speed pitches now.

I haven't accumulated enough data to see if this has any negative effect on the overall balance, but so far I don't see any obvious signs of flaws.

Very nice!

[tessl]

Re: CPU plate discipline logic has been re-written

I have only run a couple games so far, but I'm just starting to look at this. I'll be using a couple of pitchers I've always used in the past few years to see how

I think it's pretty clear CPU plate discipline has been tweaked quite a bit. How CPU reacts to pitches has been getting more and more natural every year. It's very beautiful the way CPU hitters get fooled by breaking/off-speed pitches now.

I haven't accumulated enough data to see if this has any negative effect on the overall balance, but so far I don't see any obvious signs of flaws.

Very nice!

Since you were a leading advocate of this change I'm curious what your results are now. For the ps4 version with an admittedly small sample size of around 50 games played - not simmed- in manage mode on default sliders, umpires off, non-quick counts I'm seeing a serious lack of runs and hits which I attribute to the high number of chased pitches and outs made on chased pitches.

It's as if the cpu batters have no idea where the pitch is located.

[nomo17k]

Re: CPU plate discipline logic has been re-written

Since you were a leading advocate of this change I'm curious what your results are now. For the ps4 version with an admittedly small sample size of around 50 games played - not simmed- in manage mode on default sliders, umpires off, non-quick counts I'm seeing a serious lack of runs and hits which I attribute to the high number of chased pitches and outs made on chased pitches.

It's as if the cpu batters have no idea where the pitch is located.

In CPU vs. CPU games, the lack of offense is mostly caused by the fact that, for some reason, the BABIP is very low on default setting. Basically the ball put in play doesn't become a safe hit as often as it should given what we see in real life. That can be alleviated by things like increasing Solid Hit or making pitchers' command a bit worse (so that there are more hittable pitches that become line drives) by lowering Pitcher Control sliders. I did exactly that and you can bring up the BABIP and batting average in the respectable range.

There is really nothing really wrong with plate discipline. If has only gotten much more realistic now, though there are still things that could be improved (called third strikes are a bit too many I think).

[tessl]

Re: CPU plate discipline logic has been re-written

In CPU vs. CPU games, the lack of offense is mostly caused by the fact that, for some reason, the BABIP is very low on default setting. Basically the ball put in play doesn't become a safe hit as often as it should given what we see in real life. That can be alleviated by things like increasing Solid Hit or making pitchers' command a bit worse (so that there are more hittable pitches that become line drives) by lowering Pitcher Control sliders. I did exactly that and you can bring up the BABIP and batting average in the respectable range.

There is really nothing really wrong with plate discipline. If has only gotten much more realistic now, though there are still things that could be improved (called third strikes are a bit too many I think).

Thanks for the reply.

The reason I assign blame to the change in discipline is because it was noted in the "features" pre-release and we know when you change one thing in the game it has a ripple effect.

Do you have any idea how many chased pitches per game there are in real life and how many outs are made on chased pitches per game? Again, small sample size but I'm seeing around 40 chases pitches and 15 outs per game on chased pitches.

I hate to start tweaking sliders to fix what wasn't broken last year. I wonder if there is any chance the devs will fix this in the final patch.

[nomo17k]

Re: CPU plate discipline logic has been re-written

Thanks for the reply.

The reason I assign blame to the change in discipline is because it was noted in the "features" pre-release and we know when you change one thing in the game it has a ripple effect.

Do you have any idea how many chased pitches per game there are in real life and how many outs are made on chased pitches per game? Again, small sample size but I'm seeing around 40 chases pitches and 15 outs per game on chased pitches.

I hate to start tweaking sliders to fix what wasn't broken last year. I wonder if there is any chance the devs will fix this in the final patch.

If you are really curious, I document all the changes I made to sliders and also all the data I gather in order to make decisions, and when people ask why some changes have been made, and I elaborate on why I made that decision personally:

http://www.operationsports.com/forum...4-version.html

It has the chase % (O-Swing %) from MLB and in the game. If anything, CPU hitters still don't chase nearly as much as MLB hitters do (they swing at about 27% vs. 30% of pitches outside the strike zone).

All the data I have gathered indicate that your suspicion about the perceived negative effect by the plate discipline change is unfounded. The fact that the game was a bit unbalanced at default is not attributable to the CPU discipline improvement. My guess is that it's more likely due to how the game now produce a lot more poor hits that go in between fielders, whereas these hits were very hard to generated in previous iterations of The Show.