Page 1 of 1

The AI is only human...

Posted: Sat Mar 01, 2008 4:30 pm
by Frozen Stiffer
I'm giving Puresim's AI the benefit of the doubt and assuming it was a simple error/oversight, but then- does AI really make errors per se? Is an AI "error" just a calculation or decision that didn't pan out as expected? For that matter, does the AI even expect things?

I bring this up because I had a pitcher in my rotation that had suffered a minor injury. It was an 8-day arm thing that wasn't worth sending him to even the 15-day DL. Thus, I chose to leave him in the rotation with the mental plan to replace him 'for the day' when it was his turn to pitch, instead of replacing him outright. Well, soon enough his turn came around and I, as I suppose (read: hope) several others of you have done, forgot to swap him. He suffered an aggravation of his existing injury resulting in an 8-day injury evolving into a 5-week disaster. Frankly, that was no one's fault but my own. I should have paid more attention and I didn't.

Well, it seems that the AI is vulnerable to this same thinking. I just finished a game where the AI pitcher suffered an aggravation of an existing injury. This means the AI deliberately chose to send an injured pitcher out to play. So, was this a decision gone wrong, or was this a mistake by the AI? At first you'd jump to "mistake", but... again, does the AI make mistakes? Or, did it choose to go with the pitcher (because a random slew of calculations suggested he was the best option) and then one unforseen event later, it all went to crap? The AI is all about numbers and calculations- there is no mind or memory, so he didn't "forget" to swap out the pitcher.

I suppose this is all an extension of other personnel management problems the AI has exhibited; such as one case where I found a team had no backup catcher and was playing the poor starter into the 'bright red' EXHAUSTED fatigue level. So are these mistakes, oversights, bad decisions, or deliberate choices based on calculations falling within the values that make those decisions?

With that said, I look at this post, read it and then re-read it and I realize. What am I asking? Am I even ASKING anything? Am I reqesting input from others? Ouch. Damn it. I aggravatated an existing mental injury. I'm going to go take the next 15 days off, see if that makes things better.

RE: The AI is only human...

Posted: Sat Mar 01, 2008 4:48 pm
by Bustoff
I have also seen the AI send an injured pitcher to the mound and aggravate an existing injury. This has happened maybe 3 or 4 times in literaly thousands of games. At the time, I wrote it off to some freakin' brilliant piece of code that Shaun has in the program that causes AI "managers" to make bad judgment calls every now and then...just like we do, right FS? I am hoping that's the case, but then again...
The exhausted catcher problem is again something I have seen--again, not very often--and curiously, it seems to be specific to catchers. This baffles me. Anyone else want to weigh in?
Take care of your brain, FS, it'll be back on its feet again in no time. [;)]

RE: The AI is only human...

Posted: Sat Mar 01, 2008 6:55 pm
by Nukester
ORIGINAL: Frozen Stiffer
Or, did it choose to go with the pitcher (because a random slew of calculations suggested he was the best option) and then one unforseen event later, it all went to crap? The AI is all about numbers and calculations- there is no mind or memory, so he didn't "forget" to swap out the pitcher.

I would guess this is the case with the pitcher. When all the calculations were done, the code still said he was the best choice. Thats just my guess though. Im not sure but I would hope that there is some kind of random chance built-in so that the AI doesnt make "the perfect move" all of the time. Seeing some of the moves the AI sometimes makes, it certainly doesnt always make what I would consider the perfect move, and once in awhile the moves baffle me as to why they were made.

RE: The AI is only human...

Posted: Sat Mar 01, 2008 8:22 pm
by KG Erwin
Nukester, I think you've hit upon how the AI "thinks". However, there must be a random factor that comes into play when a series of alternatives are available. This is why it can be so maddening -- you really cannot predict what it will do. It doesn't necessarily follow the rules of "expected behavior", just as we humans don't.

Take for example "calling for a pitchout". Each time I want to steal a base, I've gotta pause for a moment and consider the possibilities. Of course, I take considerable pride when certain players (Jackie Robinson, for example) can sometimes beat the throw and make the steal anyway.

It all boils down to a virtual roll of the dice.

As for the human manager's decision-making, I can only offer my own POV: I love playing PS when I'm drinking a few beers (or many). This can provide frequent opportunities to mutter under your breath, "why in the f*ck did I do THAT?" [:D]