AI Intentional Walk Bug

[Delgadodawg]

My bad, I read Washington Blue Legs 11-9 and thought he had a 2 run lead.

[Bobbo]

Washington was protecting a 9-8 lead, gave up a leadoff double in the ninth, then promptly intentionally walked the next batter.

With NO outs, that's definately not a situation where you'd want to put the go-ahead run on base intentionally.

[Eckstein 4 Prez]

Originally Posted by Bobbo

With NO outs, that's definately not a situation where you'd want to put the go-ahead run on base intentionally.

I know. I wish we didn't have to go through this argument constantly.

No major league manager EVER would put the winning run on base in the ninth inning of a game and nobody out. EVER. EVER. EVER.

(All right, maybe you can find me one time in the past ten years where it's happened - and meanwhile in OOTP it happens every single time. Every time.)

This is a bug. We don't need a slider or a "some AI teams do it and some don't." This is the equivalent of stealing third with two outs in the bottom of the ninth and your team down a run. It just doesn't happen, and for very good reason.

[Crapshoot]

Originally Posted by Eckstein 4 Prez

I know. I wish we didn't have to go through this argument constantly.

No major league manager EVER would put the winning run on base in the ninth inning of a game and nobody out. EVER. EVER. EVER.

(All right, maybe you can find me one time in the past ten years where it's happened - and meanwhile in OOTP it happens every single time. Every time.)

This is a bug. We don't need a slider or a "some AI teams do it and some don't." This is the equivalent of stealing third with two outs in the bottom of the ninth and your team down a run. It just doesn't happen, and for very good reason.

There's this guy named Barry Bonds.- perhaps you've heard of him ?

i can see this happening if the percieved drop off after the first hitter is so large that the percieved run expectancy is lowerd- with the Giants for example, Its happened before and will happen again. as a general rule though, it is a bad idea.

[Eckstein 4 Prez]

August 23, 1873 - Washington Blue Legs 5, Philadelphia Athletics 3 (11)

Fortunately, the bug didn't change the outcome of the game this time, but get a load of this "feature."

Bottom of the 11th. Athletics down by two. The leadoff man, Ezra Sutton, hits a double. The next hitter is one of Philly's best, John Radcliff. Despite the fact that there were no outs and the runner on second didn't matter, the AI in its infinite wisdom intentionally walked the tying run. I challenge anyone to explain that AI choice as a feature rather than a bug.

And before you ask, Radcliff has 3 career home runs in 464 at bats.

[Eckstein 4 Prez]

dola

I'm wondering if the oft-mentioned "comeback code" isn't at least partially due to the AI's abysmal handling of intentional walks, which set up the losing team to have huge innings late in the game. Just a thought.

[Vezna31]

Originally Posted by Eckstein 4 Prez

August 23, 1873 - Washington Blue Legs 5, Philadelphia Athletics 3 (11)

Fortunately, the bug didn't change the outcome of the game this time, but get a load of this "feature."

Bottom of the 11th. Athletics down by two. The leadoff man, Ezra Sutton, hits a double. The next hitter is one of Philly's best, John Radcliff. Despite the fact that there were no outs and the runner on second didn't matter, the AI in its infinite wisdom intentionally walked the tying run. I challenge anyone to explain that AI choice as a feature rather than a bug.

And before you ask, Radcliff has 3 career home runs in 464 at bats.

How did the rest of the inning play out? What happened?

[JDOldSchool]

Originally Posted by Eckstein 4 Prez

August 23, 1873 - Washington Blue Legs 5, Philadelphia Athletics 3 (11)

Fortunately, the bug didn't change the outcome of the game this time, but get a load of this "feature."

Bottom of the 11th. Athletics down by two. The leadoff man, Ezra Sutton, hits a double. The next hitter is one of Philly's best, John Radcliff. Despite the fact that there were no outs and the runner on second didn't matter, the AI in its infinite wisdom intentionally walked the tying run. I challenge anyone to explain that AI choice as a feature rather than a bug.

And before you ask, Radcliff has 3 career home runs in 464 at bats.

This may seem like backwards logic given my stance on this before, but I actually have more of a problem with this instance than most of the others. Perhaps because in a 1-run game, I see walking the go-ahead/winning run as a risky strategy that does make some sense (setting up for the double play). It may be risky. It may be a low-percentage risk (though I dont know if anybody has evidence to support either side). But still, it is based in some kind of logic.

In this case, that runner on second really doesn't make a difference. I'm with you here, unless the hitter is Barry Bonds (or someone of equal value) or some other extraordinary circumstance, putting the tying run on base here is just something that wouldn't happen.

This is up there with my "runner gets thrown out trying to extend a double into a triple with two outs in the 9th of a 2 run game" issue. I see it all the time, when in real life, that player probably wouldn't play for about a week if he did that. In fact, Bret Boone was thrown out at 3rd to end a game recently in a similar way and announcers were saying it's something you see once every 30 years. I see it once every 30 games (at least).

Some things just seem to happen way too frequently.

[Eckstein 4 Prez]

Originally Posted by Vezna31

How did the rest of the inning play out? What happened?

Fly out and double play.

[Eckstein 4 Prez]

Originally Posted by JDOldSchool

This may seem like backwards logic given my stance on this before, but I actually have more of a problem with this instance than most of the others. Perhaps because in a 1-run game, I see walking the go-ahead/winning run as a risky strategy that does make some sense (setting up for the double play). It may be risky. It may be a low-percentage risk (though I dont know if anybody has evidence to support either side). But still, it is based in some kind of logic.

In this case, that runner on second really doesn't make a difference. I'm with you here, unless the hitter is Barry Bonds (or someone of equal value) or some other extraordinary circumstance, putting the tying run on base here is just something that wouldn't happen.

I agree completely. Personally, I don't care for the "put the winning run on base" strategy, and I think it's very VERY rarely used in real life, but the AI should do it once in a while. If the AI did it one time in ten, I'd be okay with it.

This case is far more egregrious because that run on second matters not at all. There is literally NO reason to put that man on first. This is exactly the same as intentionally walking the leadoff man when you're protecting a one-run lead in the ninth. No team would ever do that, except for the aforementioned Babe Ruth/Barry Bonds-type exceptions.

[Eckstein 4 Prez]

August 26, 1873 - Brooklyn Atlantics at Baltimore Canaries

This was another one where, as luck would have it, the intentional walk did not affect the outcome of the game. But it really demonstrates the AI's shortcomings.

Baltimore leads 2-1 going into the top of the ninth. First batter, Jack Burdock, reaches on an error. He promptly steals second and goes to third on a throwing error. Seeing that there's a man on third with no out, the AI goes into action and intentionally walks the next hitter, Herm Doscher.

The next hitter lined out, and then the following guy hit a grounder, on which the AI tried for a game-ending double play but the batter was safe at first and the runner scored. The next guy was out, leaving the game tied. Baltimore scored in the bottom of the ninth and that was that.

But here's another thing that this pointed out to me. The AI compounds its mistake by ALWAYS going for the double play rather than the out at home. Let me see if I can explain. If the guy hadn't lined out, there would have been runners on first and third with no out. If the hitter hit a ground ball, the AI would INVARIABLY go for the double play to make it a two-out, nobody on situation.

In a given situation, that might actually be the best strategy. If I'm a good-hitting team, I might say "okay, I'll concede the tying run at third to get the winning run off the bases, because I can score in the bottom of the ninth or win in extra innings." The problem is, the AI just one batter ago was so heedless of putting the runner on base that it just conceded an intentional walk as a matter of course. Then, when the next hitter comes up, it's suddenly so concerned about the runner at first that it decides "I'd better just concede the run so I can get that other runner off the basepaths" - the VERY SAME GUY the AI just intentionally put on the basepaths.

If a manager in real-life tried that little sequence of events, he'd not just be fired - he'd probably be institutionalized.

[Dan Theman]

LOL - you know, some of you guys have a talent for bringing up one issue and combining it with another from the past. Why this reminds me of every woman I've ever dated is completely beyond me.
(I realize this was not exactly a "constructive" comment, but for some reason I felt it was needed )

[JDOldSchool]

Originally Posted by Dan Theman

LOL - you know, some of you guys have a talent for bringing up one issue and combining it with another from the past. Why this reminds me of every woman I've ever dated is completely beyond me.
(I realize this was not exactly a "constructive" comment, but for some reason I felt it was needed )

I blame all the women I've known for negatively influencing me in this manner.

[Eckstein 4 Prez]

August 30, 1873 (if this seems like it occurs almost every day in my league, it does - and remember, these are only the games that I'm watching)

Philadelphia Athletics at Boston Red Stockings

Again, this one didn't change the outcome of the game, but it almost did. Top of the ninth, Boston on top 3-0. Leadoff hitter doubles, second batter grounds to third. The next guy up, George Heubel, doubles to make it 3-1. The following three hitters are Philly's 3-4-5 hitters, top-notch players all. So naturally the AI walks the next guy onto first to put the tying run 90 feet closer to home. The cleanup hitter promptly hit an RBI single, putting men on first and second with one out and the score 3-2. The next batter grounded into a fielder's choice at second. So the tying run - the intentional walk guy - was on third with two out. The next guy hit a liner that got the "WOW! He makes the catch!" from the pbp to end the game.

Yet another case of the AI doing it all wrong. I think we've found the source of the "comeback code" right here - and it's all in the AI's stupid handling of in-game decisions.

[darkhorse]

Originally Posted by Eckstein 4 Prez

Yet another case of the AI doing it all wrong. I think we've found the source of the "comeback code" right here - and it's all in the AI's stupid handling of in-game decisions.

I have this bad habit of using the word "Markus" where you might use "AI".

[dougaiton]

Playing my solo league, the AI decide to implement this ploy as early as the 6th in a tie ballgame. Needless to say, a bevy of runs scored.

[draven085]

The issue discussed in this thread is but one example of many that make it very hard for me to take OOTP seriously as a baseball simulator.

[Anachronism]

Originally Posted by Vezna31

I'm still not so sure that is a bug. Just say that next hitter hit a grounder and it went for a DP. Now you have 2 out and a man on third and a good chance of getting out of the inning. If you don't do anything and just leave the man on second, then you leave no options for the defense except to hope that the pitcher either strikes out the next 3 hitters, flies out w/o advancing, or a grounder to the left to hold the runner. With a man on first, you get the DP option, even a grounder to 3rd for a force at third which happens. It just gives your defense more options to get outs. Just my

That's okay logic if you are up three or more runs... but not when you are up by one or two. You [i]rarely[/b] want to put the tying/winning run on base.

The only recent exception to that was a recent Toronto/Seattle game when the Blue Jays walked Ichiro Suzuki.

Tied in the seventh inning with 2 out and a runner on first and third, they walked Ichiro (.372) to load the bases (and put another man in scoring position) in order to face Randy Winn(.285). It worked as Winn popped up weakly to the infield.

BUT having said that this was a SPECIAL case. Ichiro is a freaking hitting machine and walking him in that situation makes perfect sense. Walking Joe Schmoe (.312) to face John Doe (.299) 100% of the time in late innings as it outlined in this thread doesn't make sense at all....

[Le Grande Orange]

Just to be fair, AI shortcomings are not just limited to OOTP. If you check the PureSim boards you'll find a number of posts there about some less-than-ideal AI moves as well.

It would seem there is something inherently difficult in programming a good AI manager for a text-based baseball sim...

[draven085]

Originally Posted by Le Grande Orange

Just to be fair, AI shortcomings are not just limited to OOTP. If you check the PureSim boards you'll find a number of posts there about some less-than-ideal AI moves as well.

It would seem there is something inherently difficult in programming a good AI manager for a text-based baseball sim...

I played a fair amount of Diamond Mind before I found OOTP and in my experience that DMB AI is about five times smarter than OOTP's. It isn't perfect but it's a hell of a lot better than what we have now in OOTP.