Preseason Predictions

[Jeff Reed]

I found this a little odd.

My team, the St. Louis Cardinals are predicted to score 731 runs while giving up 595. 88-74 is the surmised record.

The Cincinnati Reds are predicted to score 748 runs while giving up 662, resulting in a record of 92-70, a 4 game margin of victory in the division over my Cardinals.

My run differential is predicted to be 136 runs.

Cincinnati's is projected to be 86 runs. Yet they are going to have a better record by 4 games?

What is going on here?

[Déjà Bru]

That is odd. Can't be the schedules; they're in the same division.

I wonder if it's got something to do with the Pythagorean expectation (scroll down and read the section on Baseball) wherein teams do better or worse than expected due to luck and other intangible factors.

For example, the Reds look like they are going to be able to manufacture more runs than the Cards; look at their projected SB's. Power is another game-winning factor, as shown in the HR's.

The game may be looking at other factors like coaching quality and minor league prospects. In short, I don't know. But, I am willing to guess that it's quite complicated.

[Jeff Reed]

Yeah, I'm not saying anything is wrong, but I figured the predictions would follow a straight mathematical formula of Runs Scored - Runs Allowed = Difference = Won/Loss record. It's interesting that this is not what is going on.

Also, if you look at the Phillies and Braves in the East they are expected to have the same record. Now look at the RS and RA for each. The Braves are to score 39 more and allow 25 less. 64 runs better than the Phillies.

[Déjà Bru]

"There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy."
- Hamlet, Scene V

[Bluenoser]

I don't understand - are you saying that the team with the best run differential should always be ranked ahead in predictions?

I would think there's more to what determines predicted finish than just run differential.

[Jeff Reed]

Originally Posted by Bluenoser

I don't understand - are you saying that the team with the best run differential should always be ranked ahead in predictions?

I would think there's more to what determines predicted finish than just run differential.

No, I'm not saying that at all. I am saying I expected a computer simulation to spit out a mathematically based prediction based solely on RS vs. RA.

Since there is obviously something else (or probably more accurately, many other things) being factored in, I am curious as to what that thing or things may be.

[Jeff Reed]

Originally Posted by Déjà Bru

"There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy."
- Hamlet, Scene V

O, God, I could be bounded in a nutshell
and count myself a General Manager of infinite space
were it not that I have
bad bullpens.

- First folio edition, rehearsal copy

[SteveP]

Here's my guess about the preseason predictions:

The game runs a fast simulation of the upcoming season, in order to fine tune the league total modifiers, before turning the game back over to you to play out that season. Since the simulation produces results, the games presents those results to you as so-called preseason predictions.

Now Markus gets a chance, once again, to post a note telling me how wrong I am.

[Sweed]

Originally Posted by SteveP

Here's my guess about the preseason predictions:

The game runs a fast simulation of the upcoming season, in order to fine tune the league total modifiers, before turning the game back over to you to play out that season. Since the simulation produces results, the games presents those results to you as so-called preseason predictions.

Now Markus gets a chance, once again, to post a note telling me how wrong I am.

I've always figured it worked this way except I don't believe it adjusts any modifiers. Auto-calc lets you decide if and when to do this. I don't think Markus would make the game automatically change any modifiers. Can you imagine the screaming and hollering that would follow if users found out the game was making these changes without us having any part in the decision?

[mpejkrm]

Originally Posted by SteveP

Here's my guess about the preseason predictions:

The game runs a fast simulation of the upcoming season, in order to fine tune the league total modifiers, before turning the game back over to you to play out that season. Since the simulation produces results, the games presents those results to you as so-called preseason predictions.

Now Markus gets a chance, once again, to post a note telling me how wrong I am.

This is exactly how it works! I talked to someone who asked Markus. The game runs a simulation of the season three times and averages the results.

[Bluenoser]

Originally Posted by mpejkrm

This is exactly how it works! I talked to someone who asked Markus. The game runs a simulation of the season three times and averages the results.

Quote please

[Jeff Reed]

I didn't realize there is a second set of predictions after Spring Training as well.

Now Cincinnati does have a better run differential than St. Louis, 171-140. I certainly know who the competition is.

[SteveP]

Originally Posted by Sweed

I've always figured it worked this way except I don't believe it adjusts any modifiers. Auto-calc lets you decide if and when to do this. I don't think Markus would make the game automatically change any modifiers. Can you imagine the screaming and hollering that would follow if users found out the game was making these changes without us having any part in the decision?

I think you may have a misconception about how league total modifiers work. They are designed to work automatically so that the game will produce realistic stats on a total season, total league basis. The screaming and hollering occurs when the game does not produce realistic stats.

The game used to adjust LTMs by looking at the results in the previous season and adjusting accordingly. However, that had two problems: (1) the first season always produced bad results; and (2) the adjustments couldn't account for changes in the league that occurred between seasons. Both problems could be fixed by gamers using a little trick which was in fact based on simulating the upcoming season and manually adjusting the LTMs yourself. Markus decided to incorporate this trick in the game to run automatically.

It's entirely my supposition that this simulation feature is what produces the preseason predications.

[RchW]

Why does it then produce completely weird stats where MR appear in 116 games and have records of 21-5?

[slugga27]

Originally Posted by SteveP

I think you may have a misconception about how league total modifiers work. They are designed to work automatically so that the game will produce realistic stats on a total season, total league basis. The screaming and hollering occurs when the game does not produce realistic stats.

The game used to adjust LTMs by looking at the results in the previous season and adjusting accordingly. However, that had two problems: (1) the first season always produced bad results; and (2) the adjustments couldn't account for changes in the league that occurred between seasons. Both problems could be fixed by gamers using a little trick which was in fact based on simulating the upcoming season and manually adjusting the LTMs yourself. Markus decided to incorporate this trick in the game to run automatically.

It's entirely my supposition that this simulation feature is what produces the preseason predications.

I swear I read somewhere last year that how the pre-season predictions are determined is as follows:

The simulation engine does a quick simulation of the season three times, with injuries, transactions, etc. turned off and then averages the results to spit out the prediction.

[Lukas Berger]

Originally Posted by Bluenoser

Quote please

Don't have a quote handy but that's what I've heard from Markus too.

Pretty sure there's something in the beta forum about it, don't have time to dig it up right now though.

[kingcharlesxii]

Originally Posted by RchW

Why does it then produce completely weird stats where MR appear in 116 games and have records of 21-5?

It means the depth chart is not set up correctly for the teams that it happens to. That's one thing I wish the predictions would do is simulate an auto-reorder of the depth charts before running it.

[slugga27]

Originally Posted by lukasberger

Don't have a quote handy but that's what I've heard from Markus too.

Pretty sure there's something in the beta forum about it, don't have time to dig it up right now though.

Didn't realize someone had already beat me to the punch. That's good enough for me. Case closed.

[SteveP]

Originally Posted by RchW

Why does it then produce completely weird stats where MR appear in 116 games and have records of 21-5?

Because the LTM system doesn't care about individual players (although the system as a whole ought to produce reasonable results for individual players).

Another way of saying the same thing is that the game tracks certain stats for the league as a whole and for an entire season, and adjusts only for those stats. Other things that may occur are the result of how the game (or the AI) functions in various ways that aren't controlled by the LTM system.

OTOH, the stats that the game does control for can produce unrealistic results in other stats that the game doesn't control for, for a variety of reasons. For example, controlling IPouts for SPs tended to produce many fewer CGs than occurred in historical seasons. Markus ended up modifying the code for that particular LTM to improve the CG results.

[Mancandy]

Originally Posted by slugga27

I swear I read somewhere last year that how the pre-season predictions are determined is as follows:

The simulation engine does a quick simulation of the season three times, with injuries, transactions, etc. turned off and then averages the results to spit out the prediction.

That's what I've always assumed. When I played Baseball Mogul, I would always make my own preseason predictions of sorts by doing the same thing with their single season simulator tool.

Although, admittedly, I did a lot more three seasons. Usually 10 or 100(for ease of figuring out the averages), depending on how much time I had. Maybe letting us set how many seasons to use(more seasons take more time, but are also more likely to be accurate) would be a good option in the future...