-
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
Hey Nomo;
I gotta ask you a question.. After you spend weeks if not months doing the Testing for the rest of us (Fantastic BTW), Do you ever get to just sit back and enjoy the fruits of your labor with your own Franchise? Seems like the season might be over by the time you finish.
Anyway, as I posted back on page 2, I'm doing an Old-School thing with no DH and Classic Stadiums. Therefore I already realize that my stats are going to be off yours by quite a bit.
I also stated that I was going to try last years sliders for a bit.
Turns out that last years sliders are essentially the same as Ralphieboy's this year with a couple of exceptions;
These are just for MY personal taste for Old School Baseball...
But I have both Pitcher Staminas at 7, and Managers Hook at 3.
This is because I wanted about one in five starts to be complete games, and I would like to see my Starters go at least 7 innings, with perhaps 2 relievers coming in, a set-up guy in the 7th and 8th, with a Closer in the ninth.
What I'm seeing so far is my Starters being taken out at around 100 pitches almost exactly. I'm getting roughly one K per half inning, roughly one walk per four Ks.
I was going on default for the first day on schedule, but every game was 1-0, 2-1 with only about 6 Ks per game. I didnt like that. So thats why I went back to last year.
Anyway, I don't know if this information is useful to you because my set-up is so different and I have my own objectives; But the interesting thing was that by increasing strikeouts and stamina, I'm actually getting a lot more offense.
Is it statistically accurate? I'm sure not by todays standards. But hopefully I'm getting something close to late 60's or early 70's standards at least to my taste.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
All of those games were played with the modifications I listed. It was in exhibition mode, so you are not dealing with energy issues or hot streaks, etc.
I'm not sure about the steals. I did have base runner speed to 6 but I thought we determined in years past that it doesn't affect SBs. The only change I made was frequency down to 4.
Maybe I was creating more steal opportunities because walks were more realistic.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
These are exhibition games. I split it up evenly between leagues and ballparks. Starters should be balanced. Usually the first game I played would be 1 vs. 2, then 3 vs 4, then 5 vs. 1, then 2 vs. 3.....
I continued to use that pattern throughout.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
Good stuff! I'm about 30 games with my latest test, I'm working with strike frequency on 4 and my walks are around 2.9 in a smaller sample size but I'm also using franchise and have had all games started by #1 and #2 starters so I'm hoping that number will tick upward as I move into the lesser skilled pitching. My other sliders are similar but not exact to yours but it seems like what you are posting and what Nomo has there are trends developing. I am hoping to be at 75 games in a few days and will post results then.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
It really takes some patience. While those numbers I posted were great, I'm not even sure if 75 games is enough. When I first started I almost bailed on that set because batting average was around .275 and walks were up at 3.3. As time passed those numbers started to come back down. I think next I'm going to try those same sliders in a season mode.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I'm not doing it for anyone. I'm just very interested in how AI works in baseball games (was my childhood obsession... learned computer programming essentially to create text-sim baseball game, haha).... so this is part of my gaming actually. To see where the games does well, where it is failing, etc.
... and it really doesn't take as much effort as it looks. Just let the game start in the back ground, come back in 40 min to see if the game has finished, then spend a couple minutes recording numbers... (If you look at the spreadsheet, the cells are even ordered to make number entry faster off of box score, game log, and pitcher analysis screen...)
I don't think you can fully reproduce that era of baseball... the best you could do (without editing players) is to set Managers Hook at 0 and Staminas at 10, I think. That way, you will see a lot more CGs and less early hooks. But since baseball has changed, and stricter pitch count being implemented, starters getting hooked around 100 pitch count mark when the game is tight probably is hardcoded in the game.Anyway, as I posted back on page 2, I'm doing an Old-School thing with no DH and Classic Stadiums. Therefore I already realize that my stats are going to be off yours by quite a bit.
I also stated that I was going to try last years sliders for a bit.
Turns out that last years sliders are essentially the same as Ralphieboy's this year with a couple of exceptions;
These are just for MY personal taste for Old School Baseball...
But I have both Pitcher Staminas at 7, and Managers Hook at 3.
This is because I wanted about one in five starts to be complete games, and I would like to see my Starters go at least 7 innings, with perhaps 2 relievers coming in, a set-up guy in the 7th and 8th, with a Closer in the ninth.
What I'm seeing so far is my Starters being taken out at around 100 pitches almost exactly. I'm getting roughly one K per half inning, roughly one walk per four Ks.
I was going on default for the first day on schedule, but every game was 1-0, 2-1 with only about 6 Ks per game. I didnt like that. So thats why I went back to last year.
Anyway, I don't know if this information is useful to you because my set-up is so different and I have my own objectives; But the interesting thing was that by increasing strikeouts and stamina, I'm actually getting a lot more offense.
Is it statistically accurate? I'm sure not by todays standards. But hopefully I'm getting something close to late 60's or early 70's standards at least to my taste.
I hope this part of the game gets expanded though... (AI variation to capture different styles of baseball, that is)
I don't know how you are rotating the games for stats, but those numbers do really fluctuate from game to game, players to players, so unless you have rotated No. 1 - 5 starters a few times, I wouldn't jump to conclusions. It's that variation that makes the game very dynamic, but makes it harder to do stats tracking.
But I have found this year's game a tad strikeout happy and it probably won't be as straightforward to bring walks numbers up. Last year, all I needed to do was to drop Pitcher Consistency by one. It seems a little more is necessary this year...Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I think increasing BR Speed helps base stealers a little bit in success rate. I remember seeing some effect in MLB 11.... but I think it's the number of steal attempts that seem to be a bit low in other people's numbers, so BR Speed probably doesn't matter...
Necessary sample size really depends on what you are looking for... but mostly depend on how often the event happens and/or how small of difference you are trying to see.
if you are trying to see a rate at which relatively rare event happens (like triples and steal success/failures which happen less than once per game per team), then you will need a larger sample size to collect large enough evidence till the number stabilizes. Something like pitch count per game is easier, since in a game we collect 140 pitches or so and evidence quickly builds up.
However, if the effect you want to see is small (like overall strike % for Strike Frequency at 5 and at 4... which probably changes strike % by 1% or so... very small), then you do still need an even larger sample size even though you can collect 140 pitches (and about 90 strikes per game) or so.
I think 75 games (when settings are sufficiently varied) is a decent sample size for pitcher/analysis related stats and for a majority of box score type stats.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I'm trying to nail down reliever stamina but the opinions are so varied. I notice myself by watching games with relievers set to 5 for stamina and manager hook at 4, that the manager keeps relievers in for too long, even in game situations where he should be playing matchups. Would a lowered reliever stamina slider actually make a difference in how long the manager would keep him in, with manager hook being lowered to 4 and all?T.K.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I think lowering Reliever Stamina might make a difference in how long the relievers stay in the game, but the adjustment will probably appear for wrong reasons... if you lower Reliever Stamina too much, I think what will happen is that when a reliever stays in the game a bit too long, then he will perform poorly and gets pulled because of the poor performance, and not because of pitch count. (I think Brian@SCEA wrote before that the thresholds for pitchers becoming much less effective than their natural abilities are set for the energy levels below 25%... and if you watch many CPU vs. CPU games, relievers get knocked around quite often, especially when they are left in the game too long.)
If you like quick hooks, then Manager Hook slider might be the best bet... but I think we really need different managerial tendencies implemented in AI to satisfy the needs here. You can only do so much with sliders right now.Last edited by nomo17k; 03-29-2013, 04:31 AM. Reason: not sure about the exact threashold for relievers... couldn't find the postComment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I agree with what Nomo wrote. Another concern for me in lowering reliever stamina is how it plays out over the course of a season or franchise.
I haven't tested in those modes much yet this year. I've mostly been running a lot of exhibition tests. Last year you would find situations where a bullpen was completely gassed after a stretch with no day off or after a long extra inning game.
In those situations the AI would bring in a reliever, but almost immediately pull him because he was too tired. Occasionally all bullpen options had been used and the AI would turn to starters.
To avoid these situations, I actually bumped up reliever stamina last year. Avoiding those situations was more important to me than how the AI dealt with game by game situations.
Again, I'm not sure if this is the same as last year, but most of the AI bullpen strategy does seem to be similar from the games I've played.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
Yah I figured that would happen but I ran some simulations and league ERA's, batting averages, and strikeout rates remained all relatively close through 4 different season simulations. 2 sims had starter stamina at 7, hook at 4 and relievers at 7. 1 sim had starter stamina at 6 and relievers at 2. The last sim had starter stamina at 5, relievers at 2. Hook was still set at 4 for the latter two. I didn't take the time to check individual statistics for relievers to see if they worsened but the league averages suggest they dont. Also disparities in complete games were apparent. At 7 stamina team league leaders were around 25, 6 stamina were around 22 and 5 stamina was pretty well dead on with league averages for 2011 (2012 seemed to be an off year in regared to CG) and the high was only 18.
I hate quick hooks and I noticed a big difference when I lowered it to 4. Starters had a longer leash but relievers did too.Last edited by Threeebs; 03-29-2013, 11:23 AM.T.K.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
Although Stamina and Manager Hook sliders affect both gameplay and simmed games, I (usually) find they affect the player usage slightly differently... it's kinda unfortunate things are this way, but these two games run on different game engines, so not everything is tuned to play out the same way.
For example, when I track pitcher performance through out the season (a) in the regular CPU vs. CPU gameplay and (b) in simmed games for a year worth of playing time, usually he pitches more innings in (b)... the league leaders usually have 220 IP in simmed stats but guys from CPU vs. CPU games tend to pitch less (unless you jack up Stamina and/or lower MH sliders). I don't exactly know why this happens, but I presume it's because gameplay engine is obviously more elaborate and simulate each pitch based on the player's ability, and that may be giving the pitcher opportunities to perform slightly worse for a few reasons (pitch commands for individual pitches might not be as good and pitch count has more effects on outcomes over that specific game, whereas in simmed games I think the simulation isn't based on per-pitch basis, but likely by per-at-bat basis, for example... so if this really is the case, pitch count and therefore stamina/energy change isn't as well simulated in simmed games).
The discrepancies between gameplay and simmed games are not very apparent, but that's only because people sim most of the games played in their franchise, so the stats are dominated by those from simmed games.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I was thinking this was the case but I'm glad someone else with some experience doing evaluations said what I was thinking lol. I've found that comparing the simulation stats to MLB averages, the two are very close except for k/9. About 200 k's less in simulation. I see poster's mentioning bb/9 but from my evalutaions I think it's right where it should be. If we tried to adjust bb/9 OBP would be way to high. I'm going to comb through pitcher's algorithms for k/9 and adjust those that need be when OSFM rosters come out.
It looks as though manager hook and injuries really have the most impact on simulated games. I think I'll try leaving stamina's at 5 for both pitchers for both HUM/CPU for when I manage or watch games. That way those games reflect those of "real life". When do you think you'll have your sliders "ready"?T.K.Comment
-
Re: Stats-Based CPU Sliders [MLB 13 Version]
I've seen this answered before but I didn't take note of it. Do we know which sliders affect simmed games? I know Injury Frequency does. I didnt realize Starter Stamina, Reliever Stamina, CPU Manager Hook did but its important to know that.
I have a separate slider I load for when I sim games. I like to play every 5 games of franchise and sim the others typically. I am going to put the above sliders to 5 in my Simulation sliders along with Injury Frequency 10 which I tested and found was the most accurate compared to MLB averages of injuries. But any other sliders that affect sims besides those ones?Comment
Comment