Stats-Based CPU Sliders [MLB 14 Version]

[BrianU]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Jeez...

A 75 game test and a PM get Brian to do an all nighter!

I know I've said it before but it's an absolute privilege to be able to play a game that receives so much Developer dedication. I know a lot in your industry are just as dedicated, but I DON'T KNOW how often you guys get thanked for it. So yeah. Thank you.

This aesthetic change you mentioned; do you feel that it's possible that steals can be improved on your end without compromising the reason behind the visual change? Or is it too early to tell?

I feel guilt bring up issues but as SCEA are the most responsive dev team I've ever seen it's hard not to, I know they will follow every lead and improve the game as much as they possibly are levied by the people above them. I spend money on so much crap over the year but my $60 for The Show is a no brainier and never regretted.

[ralphieboy11]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Thanks a lot, Brian.

I have noticed cpu pitchers doing sort of a slide step from time to time with runners on first.

Plus there's that animation I mentioned before where the 2B or SS really gets the tag down quick to get the runner on a bounced throw from the catcher.

I can't remember a bounced throw last year resulting in anything but a safe call at 2nd.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

If somebody has PMed the dev because of what I'm doing in this thread (I never PM a dev myself), I actually do not appreciate that. There is a bug report form at theshownation.com, and that's a good enough venue to get developer's attention.

Some people seem to take the access and interaction with the developers here in the forum for granted, and I am not sure if that will be to a good effect over a long term. The developers here must be bombarded with PMs if people casually try to reach them.

I'm sorry, Brian, but thanks for the detailed update. We of course appreciate your time interacting with us. I keep saying this but we are indeed extremely lucky to have you work on our favorite sport game.

[Brian SCEA]

Re: Stats-Based CPU Sliders [MLB 14 Version]

This aesthetic change you mentioned; do you feel that it's possible that steals can be improved on your end without compromising the reason behind the visual change? Or is it too early to tell?

Personally, I think last year's method is better. One thing we looked at yesterday were the types of throws you can make at 2nd. The 2nd best type of throw to 2nd was almost perfect, and the best throw available is "always perfect" (as opposed to any variation or delay). This in turn means we need to give runners slightly better jumps than we might otherwise give in order to stay realistic. So it's a bit counter-productive to put in aesthetic caps on the jump. There's a better way.

In order to resolve this the better way, I'm also looking at the throw types to compensate for the good throws being too good. Especially with the 2nd best throw being a bit too close to the best throw in results. Improving this in turn would mean less need for better jumps and any kind of capping that would scale with pitcher delivery speed.

[Brian SCEA]

Re: Stats-Based CPU Sliders [MLB 14 Version]

If somebody has PMed the dev because of what I'm doing in this thread (I never PM a dev myself), I actually do not appreciate that. There is a bug report form at theshownation.com, and that's a good enough venue to get developer's attention.

Yes we saw the report on theshownation and someone on our team told me more info. It's true that we want to be impartial and objective in prioritizing bugs, but we don't want to miss out on accurate info either. I think that's what makes having one centralized location like theshownation work well.

[Heroesandvillains]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Personally, I think last year's method is better. One thing we looked at yesterday were the types of throws you can make at 2nd. The 2nd best type of throw to 2nd was almost perfect, and the best throw available is "always perfect" (as opposed to any variation or delay). This in turn means we need to give runners slightly better jumps than we might otherwise give in order to stay realistic. So it's a bit counter-productive to put in aesthetic caps on the jump. There's a better way.

In order to resolve this the better way, I'm also looking at the throw types to compensate for the good throws being too good. Especially with the 2nd best throw being a bit too close to the best throw in results. Improving this in turn would mean less need for better jumps and any kind of capping that would scale with pitcher delivery speed.

The results last year certainly reinforce your opinion on last year's baserunning.

If this cannot be fixed for 14 to your liking, is there anyway you'd consider trying to increase the frequency of the lower tiered throws in order to balance out the success percentages in this year's version?

I'd imagine this could make the games less aesthetically pleasing than your team might like though. It's a band-aid approach, for sure, but I figured I'd ask it (or wish for it, anyway) nonetheless.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Quick update. I look at this yesterday and this morning, and also ran some deeper tests overnight.

A few observations: Below average stealers (around 25 ability, 50 speed) fail more than they should, and it's partly hidden because they don't steal often to begin with. For decent stealers (50 ability, 75 speed) - let's call them "2nd-tier" stealers - a lot of steals are being decided by a very small margin (i.e. a few frames or <50-100ms), and this clustering is denser than it should be. Pitch speed can only account for a fraction of the difference, but still more than it would on paper because of the clustering. Something else must explain the rest of the gap. The most important lead I found was that certain pitchers appear to unrealistically hard to steal against.

I did a search and found one significant change from last year that would account for some of this. It's intended to aesthetically prevent runners from getting "too good" a jump against pitchers, but the actual result is that for pitchers with fast stretch deliveries and also 2nd and 3rd tier runners the steal balance is affected. It causes clustering where they're frequently "almost safe" because of the cap, and even a small shift either way will change steal balance more than it normally should. In addition, I'm looking at other factors related to this that make it hard for lower tiered runners to steal. Unfortunately, since the issue is caused by a cap, the base running slider will only help whenever the cap is not happening (ironically).

I'm working on other bugs as well, but when all things are done I'd be happy to go into more detail.

I have been taking notes of who attempt to steal in what situations and such, and in about 20 games my observations are consistent with this. (I have set the Steal Ability slider to 10 during this run.)

20 games is still not that many games for this sort of thing, but contrary to what I noted earlier, it looks like the players who do steal are not particularly off, meaning, those who steal are at least somewhat decent in Speed and BR Steal categories (at least "average" or "decent" as you described in the quoted post). It does indeed seem like the steal success rate is low, across the board.

I think there was a one or two failed hit & run attempts with a slow runner, but that's about it. It might still be slightly more steal attempts made than what the MLB numbers suggest, but I think that itself is within the adjustable range by Steal Frequency slider.

The SB success % changed from 36.2% with default sliders (75 games) to 46.8% with the BR Steal Ability slider set at 10 (33 games). I am not entirely convinced if the Steal Ability slider is still reversed, but either way this doesn't seem like correctable with it.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

After 25 games with the most recent set (labeled 4/9), the changes I made didn't really seem to increase the BABIP to the extent that I wanted. Increasing Solid Hit by two does seem to improve BABIP a bit, but I need a .033 increase in BABIP, which is rather big, and increasing Solid Hit by two probably isn't enough.

The low batting average and BABIP could be fixed by increasing Contact, but most other indicators (like those pertaining to plate disciplines) appear to be fine, so I don't wish to change drastically things that could have potential side effects. But my feeling is I probably need to touch Contact at some point to improve BABIP.

I made a few more changes. I lowered Pitcher Control and Consistency in hope of generating more offense, and also induce more walks. The rest of the changes are minor.

Foul Frequency: 3
Solid Hit: 7
Pitcher Control: 3
Pitcher Consistency: 3
Pitch Speed: 10
Fielding Errors: 8
Throwing Errors: 4
BR Steal Ability: 10

This is the 4/11 slider set in the spreadsheet.

[Heroesandvillains]

Re: Stats-Based CPU Sliders [MLB 14 Version]

How about reducing run and reaction speed, Nomo.

This is the least invasive way to increase BABIP without messing with AI hitting approach, in my opinion.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

How about reducing run and reaction speed, Nomo.

This is the least invasive way to increase BABIP without messing with AI hitting approach, in my opinion.

Fielder Reaction/Speed could be used to adjust BABIP, but I'm at this point not trying to vary too many sliders at once. BABIP out of the box seems quite low, so if I were to just use Fielder Reaction/Speed, then I would probably have to set them to something extreme. In terms of fraction of hits, however, doubles and triples are not that much lower than the MLB ave. Using extreme settings for these slider may break that balance.

Also, it usually is better to isolate a smaller set of variables that explain most of the change we see than to have more variables explain the same change. The latter is a recipe for getting confused by confounding relations.

In that regard, it might actually be better to try increasing Contact first, after all it's one slider, but the particular slider seems to change a lot of different things at once (PCI size, timing, discipline, check swing ability, etc.), so it's effectively similar to adjusting a few sliders at once, which I am reluctant to do.

So I use the Contact slider only when I really need to... but this might be that case when I need it though. I just want to get other things closer to where I want them to be (like BB%, which always needs adjustment), before going Contact route.

[ralphieboy11]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Here are some stats with sliders I've been testing out. They've been more about adjusting the pitching and fielders than the hitting. Maybe they can help someone out there.

Everything default except...

Pitcher Control 2
Pitcher Consistency 4
Foul Frequency 3
Fielder Run Speed 3
Fielder Reaction 4
Fielder Arm Strength 1
Baserunner Speed 4
Fielding Errors 8
Throwing Errors 4

This is through 57 games...
Mine(MLB Avg.)

Batting Average .254(.256)
Runs 3.98(4.33)
BB 2.48(3.12)
SO 6.71(7.22)
HR 0.93(0.97)
2B 1.44(1.73)
3B 0.25(0.18)
GIDP 0.70(0.75)
Errors 0.62(0.64)
HBP 0.13(0.32)
WP 0.32(0.34)

I tried to bring down pitcher control and fielder speeds to increase batting average and walks. Seemed to work somewhat, but still not seeing quite enough doubles or walks. Strike outs also took a dip with the move down in pitcher control.

I left stolen base stuff out because it's sort of a mess right now.

Not sure where to go next. Interested to see Nomo's numbers with his latest set.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Here are some stats with sliders I've been testing out. They've been more about adjusting the pitching and fielders than the hitting. Maybe they can help someone out there.

Everything default except...

Pitcher Control 2
Pitcher Consistency 4
Foul Frequency 3
Fielder Run Speed 3
Fielder Reaction 4
Fielder Arm Strength 1
Baserunner Speed 4
Fielding Errors 8
Throwing Errors 4

This is through 57 games...
Mine(MLB Avg.)

Batting Average .254(.256)
Runs 3.98(4.33)
BB 2.48(3.12)
SO 6.71(7.22)
HR 0.93(0.97)
2B 1.44(1.73)
3B 0.25(0.18)
GIDP 0.70(0.75)
Errors 0.62(0.64)
HBP 0.13(0.32)
WP 0.32(0.34)

I tried to bring down pitcher control and fielder speeds to increase batting average and walks. Seemed to work somewhat, but still not seeing quite enough doubles or walks. Strike outs also took a dip with the move down in pitcher control.

I left stolen base stuff out because it's sort of a mess right now.

Not sure where to go next. Interested to see Nomo's numbers with his latest set.

Interesting you could get the batting average up within the .250 range just by Pitcher Control 2. I'm using 3, but I'm not very optimistic that mine goes beyond .240, even.

I cannot calculate BABIP from just those stats, so I don't know what the difference really is. You are using one of SCEA rosters, aren't you?

[ralphieboy11]

Re: Stats-Based CPU Sliders [MLB 14 Version]

I am using SCEA rosters. I think that .250 range might be deceiving as well. Maybe I just got a run of good luck. In a similar set that I didn't post the results, I was more around the .230s.

I'm going to keep at it.

I just calculated BABIP for those games. If I did it correctly I came up with .291.

[nomo17k]

Re: Stats-Based CPU Sliders [MLB 14 Version]

I am using SCEA rosters. I think that .250 range might be deceiving as well. Maybe I just got a run of good luck. In a similar set that I didn't post the results, I was more around the .230s.

I'm going to keep at it.

I just calculated BABIP for those games. If I did it correctly I came up with .291.

It's actually interesting. I think I have had a run of either lucky offensive games or just mid-tier starters pitching and the batting average has come up to .250 range (with BABIP .291).

That's after about 25 games with Pitcher Control and Consistency both at 3. I somehow doubt making pitcher command poorer just a bit has this much effect on offense, so this might just be luck.

But if I don't have to drastically change Contact then that's a good thing in my book...

[bcruise]

Re: Stats-Based CPU Sliders [MLB 14 Version]

Just wanted to mention those pitch control/consistency tweaks make a huge difference for HUM vs CPU as well. I've drawn 4 walks off Gerritt Cole in 3 innings when before I've been lucky to get 1 against him. And he's heading towards 100 pitches by the 4th inning. Confidence is such a slippery slope.

Might be too much since he's never walked more than 3 guys in a game. One start means nothing in this kind of testing of course, though.

Edit: Done after 4 without recording an out in the 5th. Walked 5, Struck out 3. First truly "wild" bad start I've seen in this year's game (I've seen several bad ones due to getting shelled).