Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

[Caulfield]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Thanks Nomo, hardwork is much appreciated. These are the threads SDS are looking for. Good stuff.

R.I.P. Season mode, we hardly knew ye.
2006-2017
May you be reincarnated in another life in 2019.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Thanks Nomo, hardwork is much appreciated. These are the threads SDS are looking for. Good stuff.

R.I.P. Season mode, we hardly knew ye.
2006-2017
May you be reincarnated in another life in 2019.

The memorial is fitting and appropriate given that was the mode used to do all these testing.

I don't think what I do is what SDS is looking for though, haha. From what I have seen, I think there has been some tension between people who consider essential to make the game realistic when the game AI plays against AI (which I consider an important element if AI ever plays realistically against humans), but I think there are also people who do not see the value (one developer uttered to me, "who cares?"). And I think The Show has been leaning toward the latter for a while.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

...

I like to play a lot of games with every team to get decently close to league average, much like you do with CPU, although I can't get near the sample size since that would take up way too much time, actually playing every pitch. I still play enough with different sets to start to get a decent sample size. The goal of getting the stats close to league average is almost as fun to me as when I start playing my actual franchise.

But anyway, have seen the same doubles trend enough to recognize it. I also play a full 162 game season every year, playing every pitch, and zone hitting, so in that case I do get a good sample size.

Being a tuning maniac, I also enjoy accumulating stats while turning difficulty and sliders to make the game play most realistically when I control players, so I understand what you are saying.

I've been trying to come up with (another) spreadsheet for tuning the game for HUM vs CPU play... here is a working template that I was making last year.

I think there are a few people doing something similar with different sets of stats to compare, but looking at what I collect, it might be easier to see where I am coming from as I respond to your concern about the amount of doubles you generate.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

...

If it's because you're a Zone hitter and you're hitting line drives because you're a good hitter...I would do...nothing. The sliders are probably reflecting your own ability as well as your team's ability.

League averages are good benchmarks...but don't forget why there's deviations. Not every team is league average, nor every player. And since we hit for every player and you're using zone where you control the bat itself for every player...you have to think more about your playstyle archetype at least as much. Because if you hit like, say, Altuve, you're going to be better than league average at something and doubles sounds about right if you're a long-term line drive hitter. That's one of the advantages of line drives. Removing that would be more wrong than right, in my opinion. It might also mess with your BABIP.

...

Of course, this is coming from the perspective of a Directional hitter. I don't have control of the bat so playing with solid hits and power would have more implications since I'm counting on event chances (i.e RNG) more than squaring up with the PCI (which is an event chance of it's own impacting all the other chances). Or I would probably have to raise contact and hope that didn't do too much.

The point on the difference between Zone and Directional is indeed a good point to keep in mind in terms of how the hitting interface and your stick skill influence the game.

If your goal is for the game to play realistically in terms of yielding stats mostly inline with the real-life MLB averages, the best interface to use is still probably Directional/Timing, since the game/AI/random number generator is tuned to succeed/fail at the right frequency, and you only have swing timing (and some influence over PCI in Directional) to influence that outcome.

With Zone, you are effectively modulating that balance strongly toward your skills/tendencies by having full control over PCI, so whatever deviations you see from something like CPU vs. CPU games or when playing with Directional are mostly attributed to your skills and style of play. So the name of the game would be to identify what you are doing differently there.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Yeah, I've never been one to try to use league average to measure any one team. But by using every team an equal amount of times, those combined numbers and averages usually fall very close to MLB league average. The game impresses me that it is even somewhat possible. I dont care if any one team is at league average or not, since every team has different strengths and weaknesses. But I do I want each team to be within the pretty standard year to year range of MLB teams. In other words I dont want any one team to set record high or low numbers in any statistic. Especially not consistently. No team hits close to 30% doubles. As it is right now, and last year, etc, I could set league records in doubles with many teams.

Yeah, if you are randomly rotating teams and other environment factors (e.g., ball park, starting pitchers, etc.) you should be comparing your averaged stats to the MLB average. So I don't think anything is wrong there.

As far as my skill level, I can't see that being the issue. I can't hit line drives any more than anyone else. I struggle to hit in many games with some teams. But those teams still hit close to 30% of hits being doubles. The problem seems to be not too many line drives, but too high of a percentage of line drives being gap doubles.. Even if i experiment and get my line drive % down to 14%, way too many of those are gap doubles. But at that rate babip and average drops. When I hover right around league average of 19-20% line drives, all other stats start to fall right in line, except doubles.

...

My first impression is that I'd really want to check the inflation of doubles is really your issue. While there is absolutely no reason for me to doubt you have the issue there, it's not a typical issue people report, so I wonder what is going on.

For example, in CPU vs CPU games, traditionally the relative lack of doubles has been more of an issue, and not the reverse. So I can think of quite a few factors that could lead the game to produce less doubles, but the reverse is rather peculiar to me.

Now, I see the revamped ball physics for MLB 18 changed the gameplay enough that I recognize that my old intuition of the game hasn't really translated as well as it used to in this year's game, but you say that an inflated number of doubles has been your consistent issue even prior to this year. There might really be some particular tendency of yours that leads to the issue then.

First of all, no matter how you feel, if you slow down Fielder Run Speed, then you are definitely making the game easier to produce doubles (and triples). So there is at least a possibility that you may be slowing them down too much. (Also, I'd check if there is any issue with the roster you are using; if its outfielders have substantially different Speed/Reaction attributes from the default roster, that would have similar issues as changing related sliders.)

Second, since your issue seems very peculiar to me, I'd collect a little more data that make easier to see what really is going on. You say that your issue may be your tendency to hit gap doubles, while your line drive % is reasonable. Is that *really* true? Since the game has batter analysis tool, the stats are tedious but fairly easy to collect on this. Spending a little more time collecting data mindlessly is often much easier and time efficient than speculating over what is going on without really knowing what goes on. So I'd take note of a little more data to nail this issue down.

But after all that and you are sure that it's solely the inflation of gap doubles that's the problem, what could be done?

[KBLover]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

From what I have seen, I think there has been some tension between people who consider essential to make the game realistic when the game AI plays against AI (which I consider an important element if AI ever plays realistically against humans)...

+1

I remember when testing AI mods in RTS, I would prefer to watch the AI play against itself and observe behaviors, reactions, builds, etc.

Otherwise, testing becomes much more difficult because I'm shooting in the dark between game sessions and trying to play and observe at the same time.

It's too bad SDS doesn't seem to share this sentiment.

I've been trying to come up with (another) spreadsheet for tuning the game for HUM vs CPU play... here is a working template that I was making last year.

One of the best things I ever did was ask to use your CPU vs CPU sheet. Since then, I've added some stats I wanted to look at in my HUM vs CPU gameplay and have gotten objective basis to make decisions and judgments on. Thank you again for making the foundation of that work and introducing me to the idea when I happened to get curious and wander into your thread for the first time.

I still use it for my HUM vs CPU games and have tabs for the slider combos I'm testing.

I will certainly be taking a look at your template.

For example, in CPU vs CPU games, traditionally the relative lack of doubles has been more of an issue, and not the reverse. So I can think of quite a few factors that could lead the game to produce less doubles, but the reverse is rather peculiar to me.

Yeah this is usually my issue - a lack of doubles often brought on by a lack of line drives and then line drives into the gaps (they seem to either be flies into the gaps or line drives that are not too far from the OF so they get tracked down either for singles or outs, and this is with fielder speed already at 0).

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

...

I remember when testing AI mods in RTS, I would prefer to watch the AI play against itself and observe behaviors, reactions, builds, etc.

Otherwise, testing becomes much more difficult because I'm shooting in the dark between game sessions and trying to play and observe at the same time.

It's too bad SDS doesn't seem to share this sentiment.

...

Just to be clear since I don't wish to misrepresent the developers (especially when I am not one of them and this involves speculation on my part)... it's not that they do not care; I think that they still clearly do care (at least the people who tune the game), since the game is still a great simulation game when used as such.

What I meant to say was that I have felt there has been (what I think is) misunderstanding between those who value simulation authenticity and those who think it still needs to be a game (i.e., fun factor, influence of user input, etc.).

I think it's an interesting (but long) story when you think about what new features had been touted for recent iterations of The Show... but the key is that the devs had put whole a lot of effort into making the game accessible to a wider audience, and also had tried to address frequent complaint about how the game feels "scripted," where users did not feel their (good) user input was rewarded appropriately at times.

Do you remember the introduction of "precision input" stuff? I don't remember which year it was used as a marketing term, but if you think about the implication of the wording, it is basically saying that the game did not used to consider user input precisely as it was given by the user.

It wasn't really discussed a lot in the forum at the time. That is understandable since it is kind of under-the-hood stuff that should not really be in the realm of those who are just there to enjoy the game.

But I wonder what people think about this: user input not being precise. Why would a game not treat user input precisely?

I think this is a very interesting question.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

...

But after all that and you are sure that it's solely the inflation of gap doubles that's the problem, what could be done?

Picking up on this NolanRyansSnowmonkey's issue of inflated doubles to the gaps...

*If* what is really happening is that you are hitting to the gap(s) more than you should, then I'm inclined to make some adjustment in Pitch Speed and/or Timing to make it tougher for you to time pitches so well.

I don't know if you tend to hit line drives to the pull or away side of the gap, but depending on which side your hits tend to cluster I'd adjust Pitch Speed differently.

One possibility is that you tend to hit gappers to the pull side. That says you are pretty good at timing pitches optimally. In that case, I think you might want to increase Pitch Speed slider (start with just one click) so as to make it tougher to hit with that optimal timing. Make it a little tougher to catch up and stay disciplined a bit, and your hits may have more variety due to swing timing variation caused by the ever so slight increase in timing them accurately.

Another possibility is that you tend to hit to the gap in the away side... I think this is harder to correct naturally, since (regardless of your intention) you are tracking pitches well and also time well to go the other way. I don't think there is anything wrong with this approach (in fact it is a good one if your intention is to hit for a high batting average).

In that case the challenge is that by controlling all your batters, you are effectively forcing your (good) tendency to propagate to all players under control who in real life do not share that approach... so the issue might be in a category where things are not really correctable with things like sliders.

I think one thing you could try, however, is still to change Pitch Speed either way, in order to break you out of the "comfort zone" where you can hit so comfortably going away to be successful so frequently. Effectively I'm trying to introduce a different kind of stimulation to which you need to respond with changing your strategy a bit.

Should you increase Pitch Speed? That might force you to rethink your strategy to commit a bit earlier, making you pull a little more often. Or that might actually end up making you totally a slash hitter going away too often... and you might change your strategy seeing that's not how you want to hit.

Decrease Pitch Speed? You might start pulling more naturally, even if your swing timing does not change, which might be good. Or you might start hitting too well... in which case you have a number of ways to tune hitting to become harder with all the standard sliders.

That's the first thing I'd try in your particular situation. One reason why I collect swing timing info in my CPU vs. HUM spreadsheet is to gain some information about my own tendency in timing pitches. If I'm consistently early/late, then I might not be using a good Pitch Speed slider setting for my own playing.

[ericjwm]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Nomo, With the current slider settings you're using are you seeing the top K pitchers get 12+ strikeouts sometimes (cpu vs cpu)? That's been one of the difficult things for me to nail down this year while still having enough offense. I'm talking pitchers like Scherzer, Kluber, Sale, etc..

I'm planning to try your sliders soon but I'm still testing this mixed difficulty thing for now.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Nomo, With the current slider settings you're using are you seeing the top K pitchers get 12+ strikeouts sometimes (cpu vs cpu)? That's been one of the difficult things for me to nail down this year while still having enough offense. I'm talking pitchers like Scherzer, Kluber, Sale, etc..

I'm planning to try your sliders soon but I'm still testing this mixed difficulty thing for now.

Strikeouts have not been an issue. Even at the default, the game is already very strikeouts happy... which is fine given that the real-life MLB has become so.

If you want to increase swinging strikes (and therefore more swinging strikeouts), decreasing Timing slider would probably be the least intrusive, if you do not want to touch others.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

In terms of the stats-based slider set, the 4/21 set looks pretty good and, unless something really changes with the current experiment) that's what I am planning to use as the first recommended set.

I'm currently mostly trying to see how much I could lower Fielder Run Speed slider without inflating XBHs too much, for better aesthetic.

[NolanRyansSnowmonkey]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Thank you Nomo for the great input. It's a busy day so I dont have a lot of time to respond yet, but I will later.

In short, the doubles are definitely happening at a high rate. In fact, not only does it happen in larger sample sizes, I rarely go even a single game without hitting at minimum 2 doubles out of 8 hits, which is not unheard of by any stretch, but percentage-wise its high. I know some games you should hit 4 doubles, and other games zero doubles, but by the time my 3rd hit (in most games) comes, one is as almost invariably a double. I know one game is an entirely too tiny of a sample size to even mention, but when it happens basically every single game, you dont really see much fluctuation at all, except between 25 to 30%, over time, all of which is high.
I've gotten to where I involuntarily keep track of the percentage in my head, which annoys me lol, with every base hit. And it rarely drops below 25% of hits for even one game. (I think the very top of the league in large sample sizes is usually around 23%, give or take, with MLB average around 19% or so).

I will definitely try pitch speed. For a while I was also tracking whether I was late or early etc on all hits. It wasnt exaggerated in either direction, so it was about even. But with all my new testing I switched how I track everything and left that out. So much has changed and that is one slider I havent touched since the first patch. Thank you for bringing pitch speed to my attention.

I've moved timing around quite a bit already. I've had to move it up because of too many strikeouts on legend and im hovering around 2 to 4 for good K numbers. I didnt see any difference on the doubles rate, but I'll take a closer look to see if there were some small percentage point differences. But with pitch speed changing I may have to adjust that as well, like you said.

Like I said my thoughts aren't fully together because of real life at the moment. But I'll try to get more specific when i can. Thanks again for such detail and care in your responses. It helps a ton.

The bad thing about testing playing every pitch, is it just takes too long to gather decent sample sizes, so it forces me to go by feel more than I would like, before changing a slider. Otherwise I would be playing half a season every time I changed a hitting slider. So im basically forced to rely on feel, and not so big sample sizes. So I definitely dont expect it to be perfect. But if I can get it somewhat close, I can tweak as I go.

[NolanRyansSnowmonkey]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Regarding what you mentioned earlier about myself hitting for every player, I did something interesting last season that seemed to help some, maybe (probably not but I enjoyed it)

I realized that I basically had the exact same approach for every hitter, so I looked up my team on baseball reference and fangraphs, and took nerdy notes on their swing rate tendencies, even on different counts. Then I tried to give them all roles, as I played. If a player was a free swinger first pitch, I would (without tracking it exactly) try to come up with an approach that was similar, i.e. what am I looking for on first pitch with this hitter to have a similar swing rate. Same with guys that have green light on 3-0. It's amazing how many I saw that rarely even swung on 3-1. So by doing this in many different areas, I gave my team a little personality.

I have not done that this year yet because I play with every team in testing.

It's still going to be my timing and skill on every pitch, but at least it gives some resemblance of variance. Probably wont change much overall though. Still fun.

[NolanRyansSnowmonkey]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Ok one final thing. Then back to real life for real.

Is there some sort of weird buffering to power by moving the Contact slider up that I dont know about?

I was playing with hitting sliders around 55433, and getting a normal amount of home runs, well over one per game. Doubles were high but babip and average were a bit low, but home runs were never an issue at all.

Experimenting with another set, in increasing batting average, I kept the same sliders but bumped Contact to a ridiculous amount (8) just to see how much it would change things. I thought it would be just crazy offensive numbers, but to my surprise over 15 games or so (I never intended to play more than one game on this setting) my babip is right at .300 and my average is around .275.

The craziest part is that home runs have dropped to .50 home runs per game now, and doubles have dropped to a reasonable level. This makes no sense to me whatsoever. Is it just a weird fluctuation because of sample size (15 games 15 different teams) or is it something else?

The only two things I can see wrong with this set is less than half the home runs, and 20 points high on batting average. (Also a large increase to the percentage of ground balls going through the infield (about 35 to 40% as opposed to the 19% I was seeing before, and the 23%(?) or so MLB average.)

What the heck? Lol.

Edit: forgot to mention that i also had increased fielder run speed to 5 or 6 to compensate for raised Contact that much, which could make doubles come down, but the home run drop, etc still baffles me.

[KBLover]

Re: Stats-Based CPU Sliders [MLB 18 Version Season-Mode Memorial Edition]

Edit: forgot to mention that i also had increased fielder run speed to 5 or 6 to compensate for raised Contact that much, which could make doubles come down, but the home run drop, etc still baffles me.

Only thing I can think of is that you got more LD and GB so your HR dropped even as your other hits/outs stayed the same.