LOOGYS

Anyone else frustrated that the game engine will not pull your LH specialist when they face RH hitters? HAL brings in your LOOGY to face a LH hitter and then leaves him in to get torched by the next 4 or 5 righties in the line up.

I know they have stated that they only look at the first hitter, but come on, how hard can this be to set the program up to pull him? They way it is now, these guys are more ham than good.

*harm

Although ham isn't bad either.....

Are you setting your loogy's to quick hook, avoid righties, max of 1 to 2 innings?

I have noticed that unless you also have a designated righty specialist Hal will leave the lefty specialist in to face righties. This works both ways he will also leave a designated righty specialist in against lefty hitters if no pitcher is assigned to pitch against lefty hitters.

Yes, set to avoid RH, quick hook and 1-2 innings max.

Interesting idea / point about RH specialists. Haven't tried that. Does HAL leave the second guy in too long?

I will say I have had less heartburn about relief pitching decisions since I started reducing the number of static business rules I give to HAL.
I have experimented with nearly identical players on different teams and found much better outcomes by leaving most settings unset.

In fact, in a recent case, HAL removed Dan Wheeler after facing Altuve, Delahanty, and Stenzel across two innings, and while still at F9, HAL then inserted Grant Jackson (home park Forbes09) to face Mark McGwire (10.82). On the whole as a best scenario in a close game for the combined pitcher-hitter combos, it was probably the right move. But the key was HAL had the freedom to do his matchup calculations without caveats. I actually get very favorable results by tweaking less.

What I do tweak:
*Avoids (but only in very extreme cases like Mark Clear, Joe Sambito etc)
*Quick/slow hook
*Set MopUp player

At least for my mini experiments, the HAL black box of business rules in the online game is different than the CD ROM I believe, and I have verifiably better RP results in my small sample by setting less settings.

In your case, without knowing all the settings and all the fatigue states of all pitchers, and the score, and the state of the bench if non DH, it is impossible to know with certainty what HAL was thinking...but I guarantee he was thinking something, and you may have planted a thought in his head so to speak, by giving him a decision tree of bad choices for the situation. It sucks, because we are used to making the face to face decisions, but it is what it is--unknowable.

Note--this would not work well for extreme long inning reliever stuff--this is strictly conventional reliever corps.

In fact here at 80M is the latest experiment, this reliever corp's only setting: Ben Wade as mopup.
https://365.strat-o-matic.com/team/1534061

And here at 140M, only setting is Billy Wagner as slow hook:
https://365.strat-o-matic.com/team/1534069

No settings here:
https://365.strat-o-matic.com/team/1536438

In other cases I have had a lot of settings, not so good results so far.

Interesting theory, that less bullpen settings achieve better results. While you might be on to something, it is far from proven by three examples.

Those three teams might be doing quite well, but I do not believe you are getting the most out of what you have.

For instance, Billy Wagner should have the most innings pitched, and he doesn't.

Can you elaborate on why leaving the settings blank helps your bullpen perform better? Or is just an experiment based on a hunch? ... Either way your findings are of some interest.

Sure...and valid challenge. I have other examples. Still, I am aware enough of what constitutes a valid sample statistically such that I make these claims based only on anecdotal results. I am not claiming these as proven truths. Until I see otherwise, I am going with it, but it's only my hypothesis.

Based on the following.
Problem 1: The bullpen v2+ algorithm is a black box. That is we don't know what all the business rules are. We don't know what decisions HAL weighs.

Problem 2: We don't know how HAL prioritizes decision choices. When decision points arise, one thing HAL does do is check against static instructions we have given him, i.e., our pitcher prefs and team strategy settings. In some cases, it affects from the very start of the game (pitcher game availability due to advanced POW algorithms as one example). But what takes precedence? What is the primary instruction and what is secondary?

Problem 3: When we insert static instructions, ie we cut off branches of decision trees, we may be railroading HAL into a nonsensical scenario. My point is there is no way to know what we have railroaded him into.
Practical example. Let's say I always want Mark Clear facing a righty and not removed against a righty unless fatigued. Well, he is facing a team full of reverse righties in Minute Maid. He is at F9 fatigue and facing Raul Mondesi. Ahead by a run. Man on first. two out. 7th inning. Next guy is Lazzeri. You have a rested Randy Myers in the pen. HAL listens to you and keeps Clear in. Bad move? I think so. Now add to it--I have Myers set as Closer, or do not use before the 8th. Well now what have I done?
So the problem is this: Since we don't have knowledge of the black box, we don't have knowledge of the range of options HAL considers nor do we know how our choices and his are prioritized. So, when you remove degrees of freedom from the machine's decision space, you may have some unintended consequences. Machine decision logic works best with maximum degrees of freedom when it is trying to approximate human decisionmaking. Won't make a thesis on that here.

As for Wagner--maybe. But in a 140M league it is less important to squeeze the 5.78M of value out of him than at say 80- or 100. And I don't think you can definitely say the outcomes would have been better because he had more innings. He would have likely pitched in more situations with higher fatigue, or not have been rested/available for the optimal times when he was most needed to be the deciding factor. I don't think innings pitched is automatically the usage optimization strategy for him on a 140M team unless there are weak starters.

Yes, set to avoid RH, quick hook and 1-2 innings max.

You also need to set your team strategy to aggressive bullpen.

If set on normal or conservative bullpen, you basically tell Hal to follow your instructions only from times to times. Aggressive bullpen means: the instructions are orders.

And you gotta make sure your other relievers are available. I noticed that Hal (sometimes, not always) begins to ignore my instructions when there is only one reliever left available in the bullpen. By available, I don't simply mean "not-fatigued", I mean who can come to pitch based on your instructions.