Stats-based sliders for CPU

[Heroesandvillains]

Re: Stats-based sliders for CPU

Nomo, I was about ready to start a thread on this...and I still might, because I play most of my games (so I have to factor in user input)...

But I like what I see.

This may or may not be relevant, but I've been looking to decrease strikes in hitters counts. Lowering consistency has made an impact, but moderately. My walks are up, but usually generate from 3-2 counts. I have yet to touch strike frequency, by the way.

Two questions:

A. Before you made any slider changes, what was your impression of strike percentages in hitter's counts only?

B. Since the description of the strike frequency slider states that it only impacts logic for early counts...though I realize, less strikes equals less confidence, equals more balls in general, etc....have you noticed any impact on strikes in hitters counts since lowering strike frequency? Or are pitchers just falling behind more often? Thus, no noticeable impact to balls in hitter's counts?

I sure hope I can raise consistency for my played games, and focus mainly of on strike frequency...or maybe a combination of both. I realize that playing and simming are two different beasts, but the CPU logic for this should be relatively the same.

[nomo17k]

Re: Stats-based sliders for CPU

Two questions:

A. Before you made any slider changes, what was your impression of strike percentages in hitter's counts only?

B. Since the description of the strike frequency slider states that it only impacts logic for early counts...though I realize, less strikes equals less confidence, equals more balls in general, etc....have you noticed any impact on strikes in hitters counts since lowering strike frequency? Or are pitchers just falling behind more often? Thus, no noticeable impact to balls in hitter's counts?

I'm not keeping track of that specific stats so these are pure speculations, but:

I think the strike % in hitters counts must be slightly higher than the overall ave, reasons being (1) pitchers need strikes in those counts, so AI would try to get strikes and (2) at early-count (and also in pitchers count) the strike % is lower than the overall strike % ave, which means pitchers must have a slightly higher strike % to the number to balance out.

I'm not sure of the "confidence" aspect of things because I'm not taking into account in my experiment (yes it had very significant impact in MLB 10, but also heard they turned it down this year quite a bit, because some players thought the game was too driven by confidence... not for sure but I read something of that sort in the CD impression thread).

But if the confidence has any effect at all, it should work the way you described, though I think the effect won't be noticeable unless you sample many, many at bats... And we may be seeing slightly more walks purely by forcing pitchers to pitch from behind more often, the effect perhaps enhanced by their reduced confidence (and therefore their pitch command).

As far as sliders are concerned, the only thing we could do to induce more balls in hitters' counts is to lower control/consistency slider, I think. (Now I'm aware of the "upping the sliders for more walk" theory, but I cannot believe it till I see the stats.) I attach pitcher analysis screen with control/consistency both maxed out, control max/consistency min, and control min/consistency max (in order). Can only see 25 pitches, but you see how the locations scatters more when you reduce these sliders.

[Heroesandvillains]

Re: Stats-based sliders for CPU

The only variable producing anything even resembling a ball is with a lower consistency. Whoa...that's crazy.

What do you want to bet that the 'secret sauce' will be a combo of reduced strike frequency and consistency?...for overall walks I mean, not count specific strikes/balls.

For my played games, I'm going to try something like you are currently, with frequency at 2 and consistency at 4. It won't help you much, but it will help me truly understand how these interact/don't interact with one another.

I'm also a little surprised the WP are THAT high...Dai's are too, obviously, but with consistency at default in your tests, I would have figured it would be moderately lower than his. It looks like strike frequency must impact that too, huh?

[nomo17k]

Re: Stats-based sliders for CPU

I tend to think things are actually simpler than it appears, so my bet is that the strike % and pitch consistency are not necessarily influencing each other.

My guess as to why the strike frequency and pitch consistency are the two main sliders influencing the walk number is that these two are designed to force pitchers to nibble (strike freq) or widely miss targets (consistency). Just direct ways to increase balls. The pitch control only increases balls if a pitcher is nibbling but at the same time could increase meatballs, just flipping a coin, so the effect on walks is not very strong.

I actually am willing to bet that we can produce decent number of walks with pitch consistency very high, as long as strike frequency is way down. I might try that experiment once most other stats stabilizes. That might be an idea to reduce WP (though HBP is already way low... I think we should ask SCEA to produce roughly equal number of HBP and WP in future...).

[Heroesandvillains]

Re: Stats-based sliders for CPU

I tend to think things are actually simpler than it appears, so my bet is that the strike % and pitch consistency are not necessarily influencing each other.

My guess as to why the strike frequency and pitch consistency are the two main sliders influencing the walk number is that these two are designed to force pitchers to nibble (strike freq) or widely miss targets (consistency). Just direct ways to increase balls. The pitch control only increases balls if a pitcher is nibbling but at the same time could increase meatballs, just flipping a coin, so the effect on walks is not very strong.

I actually am willing to bet that we can produce decent number of walks with pitch consistency very high, as long as strike frequency is way down. I might try that experiment once most other stats stabilizes. That might be an idea to reduce WP (though HBP is already way low... I think we should ask SCEA to produce roughly equal number of HBP and WP in future...).

Yeah...you're not kidding. SCEA did a wonderful job reducing HBP, but they may have gone a tad overboard. And don't even get me started on the WP issue. But, progress is progress, and I respect them for that.

So, I see where the tests are headed! Sweet. You may be right. Nailing down BB with strike frequency is a good idea...for MOM/CPU (Players may need consistency lowered in order to judge balls depending on their ability). Your test yielded very extreme results, which I never would have guessed. I would have bet money on consistency playing the larger impact between the two. I guess I was wrong. Both played an impact, but it's clear who the champion is: Strike Frequency.

[ParisB]

Re: Stats-based sliders for CPU

One thing that has always annoyed me in this game is the 2-0 and 3-0 counts. It's almost impossible to draw a 4 or 5 pitch walk.

As soon as it gets to 2-0 or 3-0, I can almost guarantee 2 consecutive nasty pitches that barely hit the zone. It's frustrating, because this happens no matter how good/bad the pitcher is.

I can draw 3 walks, but it feels like I have to completely zone in just to get that and it usually takes me out of my game plan because once the CPU is on Ball 3 or down 2-0 they just turn into a completely different animal with their precision.

[nomo17k]

Re: Stats-based sliders for CPU

Which difficulty level are you talking about, ParisB?

[Heroesandvillains]

Re: Stats-based sliders for CPU

One thing that has always annoyed me in this game is the 2-0 and 3-0 counts. It's almost impossible to draw a 4 or 5 pitch walk.

As soon as it gets to 2-0 or 3-0, I can almost guarantee 2 consecutive nasty pitches that barely hit the zone. It's frustrating, because this happens no matter how good/bad the pitcher is.

I can draw 3 walks, but it feels like I have to completely zone in just to get that and it usually takes me out of my game plan because once the CPU is on Ball 3 or down 2-0 they just turn into a completely different animal with their precision.

I...COULDN'T...AGREE...MORE!!!

Walks are there, no doubt. But the above is the feeling I went to bed with last night.

Hang on. I'm PMing you.

[Heroesandvillains]

Re: Stats-based sliders for CPU

Huh..? Looks like Paris logged off.

He just did a drive-by 'observation-ing' and took off.

Shame. I would have liked to compare notes with him.

[nomo17k]

Re: Stats-based sliders for CPU

Okay here is the stats after 15 games with the current slider set.

There is nothing to hate about the BB number (!), and the other stats are quite good after only 15 games.

I have two issues now:

(1) K is a bit high due to a very high swing & miss % (23.3% compared to 19.3% in real life). I think this is not good. In games I can really see hitters miss too much. I increased the contact slider by one, but decreasing the foul frequency slider by two appears to have a very significant effect on the whiff rate. I never knew it would be this big! The number of fouls have been reduced but still not enough to match real life.

(2) I like BB stats obviously, but I wish to increase the strike frequency by one to match real-life better. It's good as is, but watching some pitchers pitch, I felt below average pitchers pitch a bit poorly than I'd like. Increase the slider by one I think will match pretty well the real life number. Everything else kept the same way, that obviously will reduce walks so that may be an issue.

I'm now thinking of ways to do this both at once. Possibilities include:

* increasing contact by one. This would probably act to increase walk total via improved discipline. I'm afraid offense would be inflated again, but I still have a way to reduce via power, solid hit, and maybe timing.

* set foul frequency back to default (I can take a few extra fouls, no problem...)

* fiddling with timing slider for the first time to see if it only affects the whiff rate substantially.

I'll re-read the CPU vs CPU thread and see what I should do next.

[Heroesandvillains]

Re: Stats-based sliders for CPU

With numbers that close, raising contact could do more harm than good. Especially in the Hr department (which is already too high).

Take the fouls.

Timing seemed to be very sensitive last year...almost likened to this: User Contact = CPU Timing...in the sense that offense drastically changed, even with one click, in user vs CPU games. The pci would be all over the place.

That said, it could have very well changed this year. Definitely worth a test run.

[nomo17k]

Re: Stats-based sliders for CPU

Okay, so since the last stats update, I decided to try the following sliders:

Everything at default (5) except:

Power: 4
Foul Frequency: 3
Strike Frequency: 3
Fielding Error: 9
Throwing Error: 4
BR Steal Ability: 3
BR Steal Frequency: 8

The stats for 15 games using this set is attached to this post. I decided to set contact back to default and increased strike frequency to 3, and they have intended effects. Batters swing and miss less often, while pitchers throws more strikes while BB/K isn't so much off from the last set.

Quite encouraged that I'm getting closer to getting strike %, swing & miss % and that sort of things to real life; they are important to me so pitcher vs batter match up look closer to real life!

Now it's getting closer to where I wanted to be, and the next set I'll be trying is the following sliders.

Everything at default (5) except:

Power: 3
Foul Frequency: 4
Strike Frequency: 3
Fielding Error: 10
Throwing Error: 4
BR Steal Ability: 2
BR Steal Frequency: 8

Comments: Power is decreased for obvious reasons -- I need much less HRs. Not sure this is enough to tone down the offense, but I can still lower this and also I could make use of solid hit slider, which has been untouched; looking at pop ups and choppers, I can increase their occurrences even just for more variety... let's see.

I increase foul frequency hoping that would make batters swing and miss less, strike out less, and increase walks a bit by working counts better. Hope this works as intended!

Fielding error is increased. I still see a bit less fielding errors to match real life.

Steal ability is decreased. At 3, the steal success rate is 76.7% (66/86) so far. At 5, it was 82.5% (80/97). Clicking this down by one hopefully makes the number closer to real life, 73.2%.

Loving how the game is playing!

[Heroesandvillains]

Re: Stats-based sliders for CPU

I think you may need to consider dropping the solid hits slider.

Wait, are you testing with wind at default or zero?

[nomo17k]

Re: Stats-based sliders for CPU

I think you may need to consider dropping the solid hits slider.

Wait, are you testing with wind at default or zero?

Wind is at default. Yes, I should've turned it off for slider making purposes, but at least I'm keeping track of it in my game log, so I can remove games in which wind is too strong later if needed.

Maybe this is a good thing after all; I could do how many more HRs get hit when wind is blowing out at > 20 MPH, for example, if necessary.

[Heroesandvillains]

Re: Stats-based sliders for CPU

Wind is at default. Yes, I should've turned it off for slider making purposes, but at least I'm keeping track of it in my game log, so I can remove games in which wind is too strong later if needed.

Maybe this is a good thing after all; I could do how many more HRs get hit when wind is blowing out at > 20 MPH, for example, if necessary.

If you're tracking it anyway, what the heck? Right?

Okay, I played a CPU vs. CPU game today, while hanging out with family. We all loved it.

I vote that you and Daiyoung get together, like TNK and Psychobulk, and make a CPU/MOM set...

For completely selfish reasons, ofcourse. I LOVE watching games!