June Release

[kq76]

Originally Posted by Syd Thrift

Less talk, more coding! Get back in your hole!

I can't argue with that, but I gotta say I kinda like this talkative Markus guy. It's like Superman's Bizarro or something.

[pstrickert]

Originally Posted by Markus Heinsohn

If you have 3-year recalc enabled, which is the default, then the results are no wonder. Try 1 year recalc, that will get you the results you prefer.

I just imported 1986 into OOTP 9, using 3 year-recalc, and the Mets closer is Randy Myers, which makes sense since he has the best ratings... you need to check in the player editor, you see the expected ERA there which is computed using the ratings. Myers has an expected ERA of 2.92 (in neutral enviroments), and Orosco 3.47 ... so, the AI in OOTP 9 did exactly the right thing, and if you check the K/BF and K/BB numbers, you see that Myers was the better pitcher stats-wise in that period as well.

Edit:
Checked the 86 import with 1-year recalc, and this time the closer is Anderson, who indeed has the best expected ERA. Orosco was pretty lucky in real life in 86, with a .254 BABIP, a BB/9 of almost 4 and just a 6.89 K/9. Again, the AI does the right thing.

The real-life manager may have chosen to place Anderson in less pressure-packed situations, since he was only a rookie. Had he been placed in the closer role in real-life, he may have melted down. Orosco, on the other hand, was a proven commodity. He knew how to handle the pressure of being a closer. It seems to me that the OOTP AI assumes "all things being equal" whereas real-life is never that. For that reason, the OOTP AI assumes too much when comparing the expected ERA of two different players.

[1998 Yankees]

Originally Posted by kq76

I can't argue with that, but I gotta say I kinda like this talkative Markus guy. It's like Superman's Bizarro or something.

LOL! It's nice having him here, even if he is a forum noob.

[OldFatGuy]

Originally Posted by pstrickert

The real-life manager may have chosen to place Anderson in less pressure-packed situations, since he was only a rookie. Had he been placed in the closer role in real-life, he may have melted down. Orosco, on the other hand, was a proven commodity. He knew how to handle the pressure of being a closer. It seems to me that the OOTP AI assumes "all things being equal" whereas real-life is never that. For that reason, the OOTP AI assumes too much when comparing the expected ERA of two different players.

I have to agree with Markus on this one. I mean just imagine the complaining about the AI if it put seemingly lesser (going by the ratings) people in the lineup based on some sort of ability to "handle pressure" or name recognition.

No, I WANT the AI to put the best players in the best positions, period; and I think most of us want that. We don't want the AI to start Andruw Jones in CF just because his name is Andruw Jones anymore than current Dodger fans want him starting there. The ability to adjust just HOW the AI values those decisions (using the ratings vs stats percentages) is one of the most remarkable features of OOTP that I truly love, even though I've now read that that feature may not work exactly properly. But, I think that we all want the AI to field the best team possible, following the guidelines we're able to suggest. Names and reputations should mean nothing to the AI, and I could make a good argument they should mean a little less to RL managers, using Andruw Jones as one example.

[Markus Heinsohn]

Originally Posted by pstrickert

The real-life manager may have chosen to place Anderson in less pressure-packed situations, since he was only a rookie. Had he been placed in the closer role in real-life, he may have melted down. Orosco, on the other hand, was a proven commodity. He knew how to handle the pressure of being a closer. It seems to me that the OOTP AI assumes "all things being equal" whereas real-life is never that. For that reason, the OOTP AI assumes too much when comparing the expected ERA of two different players.

There are several real life examples of rookies being used as the closer... The OOTP AI cannot use more than it has available, which is stats & ratings, and it does a good job IMO.

I mean, it would be very easy to code a mode which would assign players certain roles based on real life stats, and the AI would stick to them. But that is pretty cheap IMO.

[Markus Heinsohn]

Originally Posted by Markus Heinsohn

I mean, it would be very easy to code a mode which would assign players certain roles based on real life stats, and the AI would stick to them. But that is pretty cheap IMO.

Thinking about it, maybe it is not that cheap I'll add an option for that to the league setup...

[pstrickert]

Originally Posted by Markus Heinsohn

There are several real life examples of rookies being used as the closer... The OOTP AI cannot use more than it has available, which is stats & ratings, and it does a good job IMO.

I mean, it would be very easy to code a mode which would assign players certain roles based on real life stats, and the AI would stick to them. But that is pretty cheap IMO.

Sure, rookies have been used as closers in real life. But that's not the point. The point is that a real life manager makes decisions based on more than a formula. Had, for example, Anderson been used as a closer in real life, would he have put up the same stats? The OOTP AI assumes, based on an artificial formula, that he would. I think that assumption is ridiculous (or "pretty cheap," if you will). The real life manager has to take more into consideration than that. He knows that setup men and closers are not interchangeable. The closer role, for example, brings a lot more pressure. He knows, for example, that Orosco has handled that pressure quite well. He also may know that Anderson is not ready to handle that role. (Did history prove him right? Anderson never saved another game, other than the 1 he saved in 1986.)

[pstrickert]

Originally Posted by Markus Heinsohn

Thinking about it, maybe it is not that cheap I'll add an option for that to the league setup...

I'm glad you reconsidered. I'm impressed, Markus.

It seems to me the GF (Games Finished) stat identifies the real-life closer quite accurately.

[Markus Heinsohn]

Originally Posted by pstrickert

I'm glad you reconsidered. I'm impressed, Markus.

It seems to me the GF (Games Finished) stat identifies the real-life closer quite accurately.

I don't think the Lahman DB has GF, if not I'll just use saves...

[Markus Heinsohn]

Originally Posted by Markus Heinsohn

I don't think the Lahman DB has GF, if not I'll just use saves...

It does have GF, so that works

[Markus Heinsohn]

Originally Posted by pstrickert

The OOTP AI assumes, based on an artificial formula, that he would. I think that assumption is ridiculous (or "pretty cheap," if you will).

Well, the OOTP version of Anderson would, as there is no such thing as clutch/pressure in OOTP. Closers already get a slight ratings boost since in real life they most likely faced more pinchhitters than middle relievers, thus the competition is a little bit better.

[Garlon]

Markus, be careful with this Games Finished stat being used to assign bullpen roles. This should probably only be taken into account in the modern era or post-1975. In early time periods managers would often use their best SP in relief to finish games when needed.

Christy Mathewson finished 10 games in 1908, and depending on what team he ends up on in the simulation that could be the most on the team. There's no way we want Mathewson in the Closer role.

So perhaps limit the realistic Closer selection to pitchers with an Endurance of less than 49 (maybe even lower) who have the most Games Finished on the team based on their historical stats. This way a Mathewson would never get chosen as the Closer.

I would also argue that IRL teams misuse their relief aces, their best relief pitcher, by making them Closers. Your best pitcher in your bullpen will have a much bigger impact on winning games if used in the 7th and 8th inings in close ballgames than in the 9th inning. OOTP's AI would probably be wiser using Mariano Rivera as a Middle Reliever or Setup Man than as a Closer, although it may seem more realistic to see him racking up Saves as a Closer.

[Markus Heinsohn]

Originally Posted by Garlon

Markus, be careful with this Games Finished stat being used to assign bullpen roles. This should probably only be taken into account in the modern era or post-1975. In early time periods managers would often use their best SP in relief to finish games when needed.

Christy Mathewson finished 10 games in 1908, and depending on what team he ends up on in the simulation that could be the most on the team. There's no way we want Mathewson in the Closer role.

So perhaps limit the realistic Closer selection to pitchers with an Endurance of less than 49 (maybe even lower) who have the most Games Finished on the team based on their historical stats. This way a Mathewson would never get chosen as the Closer.

I would also argue that IRL teams misuse their relief aces, their best relief pitcher, by making them Closers. Your best pitcher in your bullpen will have a much bigger impact on winning games if used in the 7th and 8th inings in close ballgames than in the 9th inning. OOTP's AI would probably be wiser using Mariano Rivera as a Middle Reliever or Setup Man than as a Closer, although it may seem more realistic to see him racking up Saves as a Closer.

The AI will not assign players to relief roles if they are already in the rotation, and the rotation gets chosen first... so this will be no issue

There are many studies who support your point regarding proper bullpen usage, but real life managers often shy away from strategy that defies common wisdom. I mean, I would always bat my best hitter first, so he gets more AB in the season... studies have shown that this is worth about 10 runs a season compared to hitting your best guy 3rd or 4th.

[pstrickert]

Would it be possible, Markus, to include a 3-year recalc option (as per Garlon's suggestion in his VERY detailed proposal) that DOUBLES the weight of the current season? This could produce better historical results than 1-year (and better lineups, rotations, etc.).

[Markus Heinsohn]

Originally Posted by pstrickert

Would it be possible, Markus, to include a 3-year recalc option (as per Garlon's suggestion in his VERY detailed proposal) that DOUBLES the weight of the current season? This could produce better historical results than 1-year (and better lineups, rotations, etc.).

Yes, that would work...

[Long_Long_Name]

Wjile you're around, Markus... will non-English characters (é, ç, etc.) be supported in the Mac version, unlike it is right now?

[Markus Heinsohn]

Originally Posted by Long_Long_Name

Wjile you're around, Markus... will non-English characters (é, ç, etc.) be supported in the Mac version, unlike it is right now?

I don't know... can you PM Andreas about it? His username is Andreas Raht.

[Long_Long_Name]

Originally Posted by Markus Heinsohn

I don't know... can you PM Andreas about it? His username is Andreas Raht.

All right, thank you!

[PSUColonel]

Man, is OOTP turning into a replay sim? I really hope not. The focus should be and always continue to be on an unknown an undetermined future where you(the player) are taking on the role of GM.

[Kelric]

Originally Posted by PSUColonel

Man, is OOTP turning into a replay sim? I really hope not. The focus should be and always continue to be on an unknown an undetermined future where you(the player) are taking on the role of GM.

The focus should be on baseball. Past, present and fictional.