Stats-based sliders for CPU

[Heroesandvillains]

Re: Stats-based sliders for CPU

I haven't started real runs on any game mode, but for sliders I just removed the games they pitched from the stats. If I control over their appearances then I just skip Dickey's start and force Wakefield not used out of the pen. But DFA is a good idea.

Hopefully SCEA patches the knuckleball glitch before starting real runs.

What's the knuckleball glitch?

I've pitched with Wakefield, and didn't notice anything unusual.

????

[nomo17k]

Re: Stats-based sliders for CPU

What's the knuckleball glitch?

I've pitched with Wakefield, and didn't notice anything unusual.

????

http://www.operationsports.com/forum...uckleball.html

[nomo17k]

Re: Stats-based sliders for CPU

Here's some update. I've used a couple different slider sets and here's what I get.

The one I used in my last update is everything at default (5) except:

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

The stats after 40 games for this set is the first attachment. The goal was to set strike frequency at 2 to generate more walks which it did but not as much as I like; with HPB way down, I wouldn't mind having walks around 3.5. Also reducing that slider by one didn't appear to have such a strong effect on offense stats. Offense actually is a tad weaker than I'd like to be.

Since most people in the CPU thread has power at 4 and generate good results, I wanted to try going back to power at 4 and use other ways to reduce offense a tiny bit. (Maybe the reason why I tend to have slightly higher power number is due to the fact that I've been using wind at default when others have been turning off while testing.)

That got me trying the solid hit slider. Increasing hit variety is cool as well if it actually does so. But I wasn't sure how much effect the slider has on offense, I thought I was going to test the slider. For that set, everything at default (5) except:

Solid Hit: 0
Fielding Error: 10
Throwing Error: 4
BR Steal Ability: 10
BR Steal Frequency: 8
Pitcher Hook: 0 (I just wanted to see if this reduces hooks significantly)

The stats after 15 games using this set is the second one attached.

Clearly the offense wasn't reduced that much compared to everything set at default. The only significant differences that the solid hit slider caused appeared to be (1) the reduced number of line drives (down from 17.5% to 10.8% of balls in play) which in turn increased the number of fly balls and ground balls and (2) slightly increased the numbers of pop flys and choppers.

So this limited testing shows that the solid hit sliders doesn't have such a strong effect. If it ever affects the batting average, I think it all does is by converting some solid hits into poor hits and vice versa. And we know this can act in both ways. Looking at replays I didn't see increased occurrence of bloopers for hits and all, so I'm kinda disappointed.

[nomo17k]

Re: Stats-based sliders for CPU

On wind effects...

I've been setting wind to default at 5 just because I forgot to turn it off. But I've been keeping track of the weather info (thru 16 days into April, so what I write here certainly discounts long-term weather effects), and I'm thinking of reducing the wind slider. Question is how much.

I plotted the histogram for wind speed (attached). From this article:

http://www.hardballtimes.com/main/article/when-the-wind-blows/

(Thanks to DaiYoung for this reference) I can get rough ideas about how frequent those windy days are IRL and compare that to in-game stats that I've been accumulating.

The above article indicates there are (1) about 16% games with wind < 3 mph in any directions, (2) 25% games with wind stronger than 10 mph blowing either in or out, and (3) 59% games with neither.

So I wanted to reduce the wind slider until the games with wind > 10 mph blowing in or out becomes roughly 25% of total games.

With the wind slider at defaut, here is the stats: (1) 9%, (2) 34%, (3) 57%.

At 4: (1) 15%, (2) 27%, (3), 58%

At 3: (1) 25%, (2) 12%, (3), 63%

So it seems like setting the slider to 4 gets the number closer to real life.

However, I decided to go with 3 since I feel the wind effect is too strong in general.

[nomo17k]

Re: Stats-based sliders for CPU

I hope I'm coming closer to the final version of sliders, and here's what I'm testing right now.

Everything at default (5) except:

Power: 4
Foul Frequency: 4
Pitcher Consistency: 3
Strike Frequency: 3
Pitcher Hook: 0
Fielding Error: 10
Throwing Error: 5
BR Steal Ability: 10
BR Steal Frequency: 8
Wind: 3

Comment: Power is back at 4, hope reduced wind effect remove some HRs that I saw earlier in this setting.

Strike frequency is back at 3; I didn't like artificially forcing AI to pitch from behind. But I'll lose some walks so lowered pitcher consistency in hope that I can generate about 0.3 walks per game per team by this; PC at 1 roughly added 0.6 walks so it is not an unreasonable expectation...

Pitcher hook at 0 (as well as BR ability at 0) is just for experiment. So far I don't see much effect... CPU still making unnecessary moves.

I think fielding error at 10 is a must to generate enough fielding errors, but still slightly short of the MLB ave. Set throwing error back at default; I think this will roughly match IRL, though increasing this perhaps by one might produce enough total (fielding + throwing) errors.

Wind at 3, hoping not to change this from now on.

[DaiYoung]

Re: Stats-based sliders for CPU

Informative stuff Nomo. Love the histogram!

The error sliders are causing me headaches. I've gone to fielding errors at 10 for the past 45 games and seen only a slight increase in fielding errors (in those 45 games I have seen 21 fielding errors and 33 throwing errors).

My perception is that raising the fielding error slider has only impacted on outfielders and, more specifically, outfielders misjudging fly balls. Unfortunately I haven't recorded the breakdown of errors by position so I can't be 100% certain of this, but it seems that infielders are still bobbling balls/misplaying grounders too infrequently. Perhaps it is not possible to get this just right without editing infielders fielding ratings, and I'm not doing any more editing (probably).

[nomo17k]

Re: Stats-based sliders for CPU

Informative stuff Nomo. Love the histogram!

The error sliders are causing me headaches. I've gone to fielding errors at 10 for the past 45 games and seen only a slight increase in fielding errors (in those 45 games I have seen 21 fielding errors and 33 throwing errors).

My perception is that raising the fielding error slider has only impacted on outfielders and, more specifically, outfielders misjudging fly balls. Unfortunately I haven't recorded the breakdown of errors by position so I can't be 100% certain of this, but it seems that infielders are still bobbling balls/misplaying grounders too infrequently. Perhaps it is not possible to get this just right without editing infielders fielding ratings, and I'm not doing any more editing (probably).

I was concerned about the same thing about (poor quality of) outfielder errors, but we don't have separate sliders for in/outfielders so there's no simple way to get that proportion right. Certainly not without editing...

I do see quite a few fielding errors by infielders but it's just not quite enough even at fielding error at 10. So as far as impressions go, we are seeing the same issue.

I think it's not very easy to get accurate stats for something like SB and errors because they are quite dependent on situations; even if we have 100 attempts, only some would occur in similar situations so we are not strictly seeing the effect of slider itself. I think that's why we don't always see agreements in numbers for those stats.

But we need to make a compromise at some point and right now I think FE at 10 and TE at 5 appears the best combo from what people are doing so far.

[DaiYoung]

Re: Stats-based sliders for CPU

I was concerned about the same thing about (poor quality of) outfielder errors, but we don't have separate sliders for in/outfielders so there's no simple way to get that proportion right. Certainly not without editing...

I do see quite a few fielding errors by infielders but it's just not quite enough even at fielding error at 10. So as far as impressions go, we are seeing the same issue.

I think it's not very easy to get accurate stats for something like SB and errors because they are quite dependent on situations; even if we have 100 attempts, only some would occur in similar situations so we are not strictly seeing the effect of slider itself. I think that's why we don't always see agreements in numbers for those stats.

But we need to make a compromise at some point and right now I think FE at 10 and TE at 5 appears the best combo from what people are doing so far.

One part of the problem is the way errors are charged on infielders. The animation where the throw to first is almost but not quite on target and the first baseman drops it would probably get scored a fielding error on the 1B more often than not in MLB, but in the game it is always scored as a throwing error. I'd guess that raising the fielding error slider increases the likelihood of 1B fumbles on such plays.

The other animation I've seen a few times is the ground ball that takes a last second big hop and fools the infielder. I've not seen this scored as an error in the game but it's a borderline call that would probably draw an error call some of the time in MLB.

There's probably other plays that I haven't witnessed that also effect the balance between throwing and fielding errors, I think that's why raising the fielding error slider also increases the number of throwing errors. I don't think they are independent sliders.

As you say, FE at 10, TE at 5 is probably the best compromise. At least it produces approximately the correct amount of total errors, albeit with an inaccurate distribution.

[DaiYoung]

Re: Stats-based sliders for CPU

Nomo have you noticed anything odd about games played at Yankee Stadium? I need to investigate this further but just as my last game was loading they showed the HR allowed stats for the whole league and the Yankees had given up about 20 more than any other team!

EDIT:
Hmm, probably a case of small sample size/park factor but there have been 50 homers in 13 games at Yankee Stadium in my franchise.

[nomo17k]

Re: Stats-based sliders for CPU

Nomo have you noticed anything odd about games played at Yankee Stadium? I need to investigate this further but just as my last game was loading they showed the HR allowed stats for the whole league and the Yankees had given up about 20 more than any other team!

EDIT:
Hmm, probably a case of small sample size/park factor but there have been 50 homers in 13 games at Yankee Stadium in my franchise.

I have 12 games done at Yankee Stadium and the Yanks have 26 HRs and 27 HRs by the opposing teams.

I did feel there were quite a few powerfests in Bronx but never knew it was *this* much. However, I have had the wind slider at default, so I should see a more exaggerated HR effect, which I actually do here.

The park size (and wind) would definitely affect HRs, but I've never heard that there's some intrinsic park factor tied to each ballpark... Or maybe there are park factors.

[nomo17k]

Re: Stats-based sliders for CPU

One part of the problem is the way errors are charged on infielders. The animation where the throw to first is almost but not quite on target and the first baseman drops it would probably get scored a fielding error on the 1B more often than not in MLB, but in the game it is always scored as a throwing error. I'd guess that raising the fielding error slider increases the likelihood of 1B fumbles on such plays.

The other animation I've seen a few times is the ground ball that takes a last second big hop and fools the infielder. I've not seen this scored as an error in the game but it's a borderline call that would probably draw an error call some of the time in MLB.

Yes, these I observe as well. IRL the discretion by the scorers does affect the numbers, so this is something I can overlook (i.e., SECA scorers have certain tendencies...).

There's probably other plays that I haven't witnessed that also effect the balance between throwing and fielding errors, I think that's why raising the fielding error slider also increases the number of throwing errors. I don't think they are independent sliders.

As you say, FE at 10, TE at 5 is probably the best compromise. At least it produces approximately the correct amount of total errors, albeit with an inaccurate distribution.

I don't know if I see the increase in TE by increasing FE, but given most (probably all) errant throws are charged to the throwers regardless of how bad really the throws were, it's reasonable to think it does.

[DaiYoung]

Re: Stats-based sliders for CPU

I have 12 games done at Yankee Stadium and the Yanks have 26 HRs and 27 HRs by the opposing teams.

I did feel there were quite a few powerfests in Bronx but never knew it was *this* much. However, I have had the wind slider at default, so I should see a more exaggerated HR effect, which I actually do here.

The park size (and wind) would definitely affect HRs, but I've never heard that there's some intrinsic park factor tied to each ballpark... Or maybe there are park factors.

I'm not sure if they have park factors either. Probably because the dimensions of each ballpark are the same as their real life counterparts, it's not something that needs adding (although maybe there is some altitude adjustment to replicate Coors Field).

Maybe the power surge we are seeing in the Bronx isn't that unlikely. Interesting piece (albeit from 2009) on Yankee Stadium here:

http://sports.espn.go.com/fantasy/ba...ory?id=4297107

[nomo17k]

Re: Stats-based sliders for CPU

I'm not sure if they have park factors either. Probably because the dimensions of each ballpark are the same as their real life counterparts, it's not something that needs adding (although maybe there is some altitude adjustment to replicate Coors Field).

Maybe the power surge we are seeing in the Bronx isn't that unlikely. Interesting piece (albeit from 2009) on Yankee Stadium here:

http://sports.espn.go.com/fantasy/ba...ory?id=4297107

Right, I read a similar story about HR friendliness of the new Yankee Stadium, and how they don't know where it came.

Putting the park dimensions and winds aside, most of the time people talk about HR park factor, it appears to be tied with the "air" conditions, being it humidity or thinness due to altitude, etc. But to make that sort of thing affect the game they need to hard code the physics and I don't know if SCEA has gone that far yet.

And the wind effect isn't actually as simple as something blowing in one direction (which the game implements already). The ballparks with complicated structures can generate unusual air flow, so I'd be surprised (but impressed) if that sort of park factor is already implemented in the game.

[Heroesandvillains]

Re: Stats-based sliders for CPU

The only park that I'm aware of SCEA 'coding' is Colorado's.

Breaking balls have less room for break (think of it like what happens to the 'break' arrows as pitcher tires during played games). This effect should be noticable from pitch one. As a result, more balls get hit with more authority.

I'm unaware of anything like this at Yankee Stadium (in the game). I wouldn't be surprised, as I just gave up the most homers I've ever given up in this year's game over a 27 inning stretch. Nine. The wind was between 2-11 MPH, and only was blowing out in one of the games.

Guess where it was.

I have no idea how SCEA made even poorly hit balls jump there. Yankee stadium is nuts this year.

[doncoryell]

Re: Stats-based sliders for CPU

I don't know if it means anything or not, but the park factors for Yankee Stadium for 2010 were: Batting - 113, Pitching - 111 according to baseball-reference.com. (www.baseball-reference.com/teams/NYY/2010.shtml) There were also 223 total home runs hit at Yankee Stadium in 2010, with 8 being the most in one game and 0 in several games. The number of home runs combined in one game that appeared the most was 4.

For 2009, it's first season, it was Batting - 97, Pitching - 96.