Elephant in the Room: AI Bullpen Usage

[lvann4x]

Also, the way the AI chooses who will fill the closer role in Franchise Mode gets worse as seasons go by.
Teams with RP's rated in the 80's will have a Closer rated in the low 70's or even high 60's just because he's the best pitcher who happens to have CP next to his name. The AI needs to be tuned to use one of at least it's top 2 relievers as it's closer, regardless of whether his position is RP or CP, or even in some cases where they may have an SP with low stamina but great stuff who is head and shoulders better than their best reliever, but not a good fit in the starting rotation.

[Uncle Elton John]

Quote: Originally Posted by lvann4x

Also, the way the AI chooses who will fill the closer role in Franchise Mode gets worse as seasons go by. Teams with RP's rated in the 80's will have a Closer rated in the low 70's or even high 60's just because he's the best pitcher who happens to have CP next to his name. The AI needs to be tuned to use one of at least it's top 2 relievers as it's closer, regardless of whether his position is RP or CP, or even in some cases where they may have an SP with low stamina but great stuff who is head and shoulders better than their best reliever, but not a good fit in the starting rotation.

yes, it is a debacle. let us hope they incorporate a OOTP like sim engine for MLB 17.

[Mike Lowe]

I have not found a way to reduce the number of complete games and shutouts pitched in simulated games. Nothing that doesn't require a ton of work.

[oldtimey]

Quote: Originally Posted by Mike Lowe

I have not found a way to reduce the number of complete games and shutouts pitched in simulated games. Nothing that doesn't require a ton of work.

I put in a lot of work reducing the stamina rating on all starters in my 30-team franchise and I am very happy that I did. I sim all games daily, except one featured game that I watch live. I rotate the games that I feature to allow me to see at least one series in all 30 ballparks. I am in September and my leaders in CG are

AL: Keuchel, HOU 6, Carrasco, CLE and Minor, KC 5;
NL: Lester, CHC 5, Wainwright, STL, Grienke, AZ, Samardzija, SF, Arrieta, CHC and deGrom NYM 4.

I checked the leaders in the real MLB and saw that Sale leads the AL in CG with 5, and Bumgarner and Cueto are tied for the NL lead with 4. So I am pretty satisfied with my results. I have also had two no-hitters (Bundy of BAL and Nelson of MIL). And today, Cobb of TB took a no-hitter into the ninth in Yankee Stadium before Headley broke it up with 2 outs to go. I play cpu-cpu, with starter stamina slider set at 4, reliever stamina 0 and manager hook at 5, in addition to the above mentioned adjustments.

[Mike Lowe]

Quote: Originally Posted by oldtimey

I put in a lot of work reducing the stamina rating on all starters in my 30-team franchise and I am very happy that I did. I sim all games daily, except one featured game that I watch live. I rotate the games that I feature to allow me to see at least one series in all 30 ballparks. I am in September and my leaders in CG are AL: Keuchel, HOU 6, Carrasco, CLE and Minor, KC 5; NL: Lester, CHC 5, Wainwright, STL, Grienke, AZ, Samardzija, SF, Arrieta, CHC and deGrom NYM 4. I checked the leaders in the real MLB and saw that Sale leads the AL in CG with 5, and Bumgarner and Cueto are tied for the NL lead with 4. So I am pretty satisfied with my results. I have also had two no-hitters (Bundy of BAL and Nelson of MIL). And today, Cobb of TB took a no-hitter into the ninth in Yankee Stadium before Headley broke it up with 2 outs to go. I play cpu-cpu, with starter stamina slider set at 4, reliever stamina 0 and manager hook at 5, in addition to the above mentioned adjustments.

What edits did you make exactly to SP stamina?

How do your RP innings/game look? Are they still pitching 1+ inning a lot of times, or are you *actually* (huge asterisk) seeing situational bullpen usage by AI teams?

[oldtimey]

Quote: Originally Posted by Mike Lowe

What edits did you make exactly to SP stamina? How do your RP innings/game look? Are they still pitching 1+ inning a lot of times, or are you *actually* (huge asterisk) seeing situational bullpen usage by AI teams?

Check out this thread from a couple of months ago, http://www.operationsports.com/forum...na-rating.html. Also, attached are some spreadsheets from a game that I watched between CHW and TB. You can see examples of the bullpen usage, as well as all the other statistical results that have been achieved with the adjustments described in the link, my sliders with Bacon96's roster from spring training. The only thing that has been off compared to the real MLB has been stolen bases. I recently just Increased the SB frequency and I am happier now with that. As far as situational relief is concerned, the AI seems to go with performance in the franchise games rather than lh/rh ratings or which way the batter hits. And it looks at the next three hitters due. So, if the next three hitters cumulatively are hitting worse against LHP, then the AI will bring in a LHP. When the game is late and close, with reliever stamina set at 0, if that LHP comes in and permits a baserunner, he will get the hook and the AI manager will go to the bullpen until someone can get him outs or he has no one left in the bullpen to go to. At this time, that's the best this game can do as far as situational bullpen usage. Check out the thread and the spreadsheets and I will be happy to answer any other questions you might have.
https://docs.google.com/spreadsheets...?usp=drive_web
https://docs.google.com/spreadsheets...?usp=drive_web

[KBLover]

Quote: Originally Posted by oldtimey

As far as situational relief is concerned, the AI seems to go with performance in the franchise games rather than lh/rh ratings or which way the batter hits. And it looks at the next three hitters due.

I've noticed it will go by handedness or at least matchup for extreme differences.

For example, I can be assured to see a LHP RP for Gallo, etc.

But in the case of guys with no real split or not a big one, it has a harder time, as expected since pitchers have no splits.

[oldtimey]

Quote: Originally Posted by KBLover

I've noticed it will go by handedness or at least matchup for extreme differences. For example, I can be assured to see a LHP RP for Gallo, etc. But in the case of guys with no real split or not a big one, it has a harder time, as expected since pitchers have no splits.

Agreed. Guys who are platoon players have had there split versus same-handed pitchers adjusted way down to create poor stats in my franchise to encourage these situations. It would help a whole lot if the pitchers had split ratings. Then, the AI could lean on the pitchers strength in those situations when it's neutral or reverse.