OOTP Decision Engine

[Flatbush]

Greetings OOTP Community. This is my first post - I purchased a copy of the game a couple weeks ago, and have been reading the manual and "lurking" around the forums. Thanks to the OOTP team for giving us a great game - and to the veterans here who graciously give their time and experience to help the newcomers. Someday I hope to know something that might be helpful to someone.


With full respect to "industry secrets," what is known about how the OOTP engine makes decisions? For example, remember the dice games where the roll determined if the result came from the batters' card or the pitchers' card? (they were fun...not as much as OOTP though) Or does OOTP factor in the pitcher and batter ratings into a composite of possibilities?


Do managerial decisions (guard the lines, corners in, outfield deep) change the table of possibilities or just only randomly impact the result? If I knew how the engine made decisions, it might impact how I managed - and at 2 and 5 with a staff ERA over 6, I could use an edge.


In your experience, are games won and lost because of good or bad managerial decisions?


How possible is it to see a statistical anomaly over a season? A career .240 hitter have a .300 season... or an all-star pitcher have a 10-10 year.... or a homer hitter fall from 40 to 25? In my 1957 replay, is it possible to improve on the Dodgers actual 84-70 season?


I read a post recently (I think from Charlie Hough, but I'm not sure...) that indicated that the game has a built in capability to adjust statistical anomalies on a team basis as the season progresses. (if I'm not interpreting this correctly...please let me know!)


Thanks for whatever consideration you can give these questions. Mainly, I wanted to say "hello" and introduce myself - I'd like to stop by and read the posts and once in awhile chime in, after I learn more about how things work.

[Déjà Bru]

Welcome, Flatbush. I will take a shot at some of these.

Originally Posted by Flatbush

With full respect to "industry secrets," what is known about how the OOTP engine makes decisions? For example, remember the dice games where the roll determined if the result came from the batters' card or the pitchers' card? (they were fun...not as much as OOTP though) Or does OOTP factor in the pitcher and batter ratings into a composite of possibilities?
Rest assured that it is not pure luck or randomness, else the game would be chaos. It is those factors of chance built into a boatload of algorithms that take all of those ratings into the realm of possibility. (How's that for a bunch of bull? Heh. But it's the best way this layman can describe the process.)

Do managerial decisions (guard the lines, corners in, outfield deep) change the table of possibilities or just only randomly impact the result? If I knew how the engine made decisions, it might impact how I managed - and at 2 and 5 with a staff ERA over 6, I could use an edge.
As I described above, the fact that you guard the lines in a certain situation does affect the algorithm at work in determining the outcome of the pitch/at-bat. It alters the odds of one outcome over the other, let's say. Then, of course, luck enters into it as in real life.

In your experience, are games won and lost because of good or bad managerial decisions?
Certainly. My owner wants my head on a platter but I have protections (I cannot be fired)! Seriously, you can sim results without taking over any team and not have a hand in it. However, if you do manage an organization, the better you manage everything, from coaching changes to roster moves to lineups to on the field strategy, the better your results. Up to a point. As in real life, you cannot make a silk purse out of a sow's ear.

How possible is it to see a statistical anomaly over a season? A career .240 hitter have a .300 season... or an all-star pitcher have a 10-10 year.... or a homer hitter fall from 40 to 25? In my 1957 replay, is it possible to improve on the Dodgers actual 84-70 season?
Yes. This is often stressful to strict reconstructionists who want to see outcomes exactly as they occurred in reality. There are some settings that will attempt to do so in a historical replay. However, without those settings, the game is free to, and likely to, alter outcomes according to chance and how things are managed. That's what makes the game interesting, at least in my opinion. HOWEVER, if you were to take the same 1957 Dodgers and play that season over and over, the results should gravitate to the norm of them ending up 84-70. That is an important concept in terms of game integrity and simulation quality.

I read a post recently (I think from Charlie Hough, but I'm not sure...) that indicated that the game has a built in capability to adjust statistical anomalies on a team basis as the season progresses. (if I'm not interpreting this correctly...please let me know!)
On a team basis, no. There are settings for you to change to alter statistical outcomes on the league level.

Thanks for whatever consideration you can give these questions. Mainly, I wanted to say "hello" and introduce myself - I'd like to stop by and read the posts and once in awhile chime in, after I learn more about how things work.
Again, welcome to the community and enjoy the game.

[mgoetze]

I'm pretty sure the authors have read a lot of sabermetric research and included as much of it as possible. That includes, of course, research into how large the variance of results is. For instance, given a player whose skill should lead to a long-term on-base percentage of .330, a season with 500 plate appearances should show an on-base percentage between .309 and .351 ... 68% of the time (Table 12 in "The Book" by Rango, Lichtman, Dolphin). I have no idea how lucky the 1957 Dodgers were but if you play that season often enough you will eventually be luckier.

Obviously management will have some effect. OOTP will work the way people writing about suboptimal use of relievers by actual baseball managers think things should work.

[Charlie Hough]

OOTP does not use a traditional approach to resolving batter/pitcher matchups using "player cards" and "dice rolls" like most games have done over the years. I don't profess to know how OOTP resolves the action for each at bat, but users have to accept that the game engine is not going to be transparent. I don't think the developers have ever revealed much of the detail.

I can tell you a few things, though. OOTP's focus is on generating highly realistic league totals across all statistical categories. Generally speaking, as results are determined, these totals are shared or divided across the players based on their ratings. And, for historical or current year MLB games, those ratings are based on their real life statistics. So, in theory, as games are simmed and a season progresses, player performances should be within a standard statistical variation of their real life performances, and the overall league totals should be very accurate.

Depending on the settings that you use in the game, this can mean that some players will significantly outperform their real-life results, while others will underperform. Many gamers don't have a grasp of standard statistical variations, so they don't realize how wide the range of potential results can be with statistical probabilities. So, yes, you will occasionally see a .250 hitter bat .280 for a season or a 30 home run player hit 20 or 40.

Unlike most other baseball sims, OOTP does not cheat in order to force results to resemble real life statistics. It doesn't 'see' that a certain player is exceeding or lagging behind his real life stats, and it doesn't apply penalties or bonuses to modify future sim results and artificially force the issue. Instead, it uses algorithms to calculate player ratings, distribute statistics realistically, and simulate games so results are fairly accurate and believable.

Overall, OOTP does a very solid job of simulating an accurate and believable baseball league, and you will find that league-wide statistics vs. real life are very close. And individual player performances are pretty solid as well.

If you prefer single season accuracy and a replay style baseball sim, then you can use the one-year ratings recalculation and other settings to make all player ratings based on the season in question. You can turn off player development and all other settings that may allow your game to deviate more from real life. So the end result should be player and team performances that are very close to real life.

Or, if you prefer a bit more variability and want to take a career sim approach to historical leagues, you can use other settings to allow more smoothing of player ratings and performances, deviation in results, transactions, player ratings development, etc.

So, for example, if you choose to use a 3-year basis for calculating player ratings rather than one year for the particular season you're simulating, player performances may be smoothed out a bit, so they're not going to follow each individual season as closely. They will be more representative of a three-year average of their real life performances with standard deviations still unfolding.

Or, if Babe Ruth is never traded to the New York Yankees in your baseball world, that can have a huge impact on pennant races, the performance of the Boston Red Sox, and even Ruth's performances depending on who bats around him in the lineup.

As a manager, you can definitely affect results and the outcomes of games, but you also have to consider the abilities of your players, the competition, and probabilities. You're not going to take a 55-win team to the pennant if you're using 1-year recalculation and taking a replay sim approach. You might win 60 or 65 games, or you might win only 50. But you're not going to win 90. And sometimes your players will have slumps or streaks, and sometimes your team will overperform or underperform. There is a limit to what you can influence as a manager, just as there is in real life.

Managerial strategies definitely matter, and I would read the manual to get a sense for what each option does. For example, when it comes to holding the runner, the game does this automatically and the option to 'hold runner' means that you're asking the pitcher to alter his delivery, pacing, and throws to first to keep the runner off guard and make a further effort to hold him. This is probably best used only when there is a top flight base stealer on base and you're in a situation where you really want to keep him from advancing. Because there is a potential drawback, and I believe it's that your pitcher can't focus as much on the batter and may be more prone to issue a walk or miss a pitch.

So take some time to learn the strategies, experiment, and see if they help or hurt you. For example, I've found that playing with the corners in to defend against the bunt doesn't seem to yield plays at second place. This is presumably because the second baseman has to back up first base. So I've determined that I'm better off leaving the corners in a normal position on expected sacrifice bunts, and this helps get runners out at second on occasion.

But keep in mind that the game seems generally designed to use the default resolution of each at bat. So the managerial strategies are more for addressing special situations, batters with particular tendencies, etc. I wouldn't use any of them too heavily. It's not as if you need to be executing a special strategy with every pitch.

[The Wolf]

The game engine is DIPS based.

[Flatbush]

Thanks for the input. That was stimulating. Just like the other Charlie Hough, (1970 to 1994, I think?) he's a good man to have on the roster.

I understand why so many of you appreciate OOTP. Let's face it - it is possible to have a computer program that would within 200 at bats be within percentage points of the ball players actual BA (in fact, it would be an easy program to write - I just retired from that line of work) But the problem with that level of reality...is that it is no fun!

However, the other extreme is worse - "chaos" as Bru mentioned - where your lead off man leads the team in homers....with 13.

I am more and more impressed with every game...because I see OOTP has walked a very fine line in giving us a weighted game where a hitter like Mike Trout can be expected to do better than an average player, but there are still some surprises.

Thanks to all, especially Charlie, for a very thorough and insightful post.

[PSUColonel]

[QUOTE=As a manager, you can definitely affect results and the outcomes of games, but you also have to consider the abilities of your players, the competition, and probabilities.[/QUOTE]


Does the AI do the same?