Historical replay accuracy

[bailey]

Originally Posted by Jah

When individual player stats are out of whack, team performance can be affected as well. In OOTP - or any other purported replay or replay-like sim - it isn't plausible for a real-life .269 lifetime hitter to hit .350...

Ummm --- Norm Cash, lifetime .271, hit .361 in 1961.

[Cryomaniac]

Originally Posted by bailey

Ummm --- Norm Cash, lifetime .271, hit .361 in 1961.

I think he meant that he shouldn't hit 3.50 lifetime in the sim.

[StyxNCa]

Originally Posted by bailey

Ummm --- Norm Cash, lifetime .271, hit .361 in 1961.

I think he is talking about a particular season. Player A hit .250 in 1970. He shouldn't hit .350 in a 1970 replay.

[StyxNCa]

Originally Posted by Le Grande Orange

That's exactly the point—what does historical accuracy mean in this context? If Player A in 1974 typically batted third in the line up, played in 100 games, and hit .270, does that mean in OOTP he must typically bat third, play in 100 games, and hit .270? How far from those figures can the game diverge and still be considered historically accurate?

What I'm getting at is that any major league season is the result of an enormously interwoven tapestry of events. Once you start pulling on some of the threads, the results should not expected to be identical since you are changing the nature of the tapestry. OOTP already does this in one small way—the schedules used are the original schedules, not as they were played. This means variations in off days from how the real season unfolded, and consequently alterations to pitching rotations, which necessarily should be expected to impact pitcher usage. The effect may be small, but it's there, and the more of it that goes on the more the results may likely be different at the end. The same sort of thing would be true with a myriad of other small influences.

Basically, it comes down to this: what items in a historical league in OOTP are allowed to vary from reality, and to what degree? I don't think there'll be much consensus on this, especially once you get down to the specific details, and because much of it is subjective judgement.

I think this whole issue is being misunderstood. I have said, as others have said, we don't expect exact results. That would be as boring as results that bear no resemblance to reality.

League leaders in categories should not finish in the lower 25% of the league in whatever stats they were a leader in. For example, I don't care if Williams didn't hit .400 in the game for that season, but I would expect to see him close to the top of the leader board in that stat, even if the number was .340 or something. This is what people are trying to say. We can not give certain numbers when we say we want accuracy because that is not what we're talking about. We are talking about the realm of believability. I can't believe that the 1969 Orioles constantly finish with a higher team ERA than all but 2 other teams, constantly. The actual team ERA doesn't matter, but being the #1 team in ERA, I would expect them to finish in the top 5 anyway. (Hm, that sounds like a contradiction but maybe you can understand what I'm trying to say) I see most of the problem among pitchers as far as things being way out of whack, but that's not saying a fair number of batters can never seem to perform at a level I would expect.

The overall league stats always seem to look good, so that part is pretty accurate.

Now I have to agree that "believability" in itself can be a problem since my idea of believability is probably different than Joe Blow's and so forth and so on.

Anyhoo, like others, I am using another game for historical so I reckon I shouldn't be saying anything about that aspect of OOTP.

[Jah]

Originally Posted by bailey

Ummm --- Norm Cash, lifetime .271, hit .361 in 1961.

I meant for the particular season. In the example I was speaking of (see the specific info in the thread link I posted), the guy was a lifetime .250 hitter and never hit hire than .269 in any season in his career...so, no matter how you look at it, his big season didn't make any sense.

[pstrickert]

Part of the fun of single-season historical replays is being able to "drop in" and immediately recognize it. If I "drop in" to the 1982 season, I'll expect to see (when I face Milwaukee) Molitor, Yount, and Cooper at the top of the batting order. I'll expect to face Vuckovich or Haas or Caldwell in a key, late-season series. But if the AI (not beholden to as-played lineups, for example) establishes the rotation, it may drop one of these pitchers out of the rotation. Why? Because the expected ERA for Caldwell, for example, may be 0.05 higher than Jerry Augustine. So, the AI will swap Augustine for Caldwell in the rotation. It may "make sense" to the AI, but -- multiplied enough times -- this kind of move begins to ruin the enjoyment of replaying a historical season.

[StyxNCa]

Originally Posted by pstrickert

Part of the fun of single-season historical replays is being able to "drop in" and immediately recognize it. If I "drop in" to the 1982 season, I'll expect to see (when I face Milwaukee) Molitor, Yount, and Cooper at the top of the batting order. I'll expect to face Vuckovich or Haas or Caldwell in a key, late-season series. But if the AI (not beholden to as-played lineups, for example) establishes the rotation, it may drop one of these pitchers out of the rotation. Why? Because the expected ERA for Caldwell, for example, may be 0.05 higher than Jerry Augustine. So, the AI will swap Augustine for Caldwell in the rotation. It may "make sense" to the AI, but -- multiplied enough times -- this kind of move begins to ruin the enjoyment of replaying a historical season.

I imported the 1969 season a few days ago to sim it and some of the roster/reserve list decisions it made were crazy. Mike Cuellar, the Cy Young winner that year, on the reserve squad? Dick Bosman on the reserve for Washington while Camillo Pascual was in the rotation? These are a few examples. I had to go and expand all rosters to 50 players so the AI would not keep putting these players back in the reserve. As it was it gave Cuellar 9 starts for the entire season even though I manually put him in the rotation.

[Bluenoser]

Originally Posted by StyxNCa

Well, the claim is that it is the best game out there, even for historical. That last part is not true, and the one thing I pointed out is one of the reasons why.

If you're going to claim historical accuracy then it better be just that.

Someone suggested long ago that OOTP should be divided into 2 games, fictitious and historical, and that would be great, but it won't happen. As people have pointed out time and time again, the engine isn't "really designed" to be an historical simulation, even though the newsletter mentions how this version is going to be the "most accurate" version for SINGLE SEASON historical replay. So what everyone is claiming is opposite of what Markus himself states in the newsletter.

But in reality, because the game is primarily for fictional play, it will never be the best historical game out there, and that's fine. I'm happy with the fictional part and that's what I use OOTP for, and I don't think anyone will ever be able to top OOTP for that.

Claiming you're the best and claiming you're accurate are two different things. To me, OOTP is the best historical, but that's because I'm not concerned with accuracy as in mimicing real baseball.

[StyxNCa]

Originally Posted by BruceM

To me, OOTP is the best historical, but that's because I'm not concerned with accuracy as in mimicing real baseball.

And that's fine if that's what you consider historical. It suits your purpose.

But as I have pointed out, Markus put out a newsletter stating:

"Improved Historical Simulation


The system for calculating the fielding ratings have been reworked, a recalculation period for the pitcher endurance was added, new total modifiers were added and the accuracy of the initial season was improved so that single-season replays are now more realistic. Also, imported pitchers now throw the pitches they did in real life, we have enhanced the historical database with real pitch data"

Just seems his idea of historical is what some of us have been talking about, and his phrase "now more realistic" says to me that even he thought it wasn't as realistic as it should be previously.

[Le Grande Orange]

Originally Posted by pstrickert

Part of the fun of single-season historical replays is being able to "drop in" and immediately recognize it. If I "drop in" to the 1982 season, I'll expect to see (when I face Milwaukee) Molitor, Yount, and Cooper at the top of the batting order. I'll expect to face Vuckovich or Haas or Caldwell in a key, late-season series. But if the AI (not beholden to as-played lineups, for example) establishes the rotation, it may drop one of these pitchers out of the rotation. Why? Because the expected ERA for Caldwell, for example, may be 0.05 higher than Jerry Augustine. So, the AI will swap Augustine for Caldwell in the rotation. It may "make sense" to the AI, but -- multiplied enough times -- this kind of move begins to ruin the enjoyment of replaying a historical season.

The last sentence is part of why I find the philosophical question of what historical leagues should be an interesting one to consider. I think I would tend to agree with your view.

Another interesting thing is, with Retrosheet putting up ever more seasons' worth of boxscores and play-by-play data, the ability to have historically correct starting line ups and player usage for nearly every game in a season is only growing.

[Pdubya64]

I don't see the "ideal" that many of us here long for happening in our lifetime. Why? Well, if I have had any grand concept of life demonstrated to me time and again it's this one:

In order to get something, one must give up something. aka "There is no Free Lunch". There is a cost to everything and a consequence to any decision that is made.

So, how is this related to OOTP you ask?

The instant you decide that OOTP is no longer a strictly historical (replay) simulation of baseball where players always follow their RL results, I maintain that you have just given up the capability (at least in our lifetimes) to "have the best of both worlds" where a team will follow their basic history with only a small deviation in performance of its players.

Essentially, there is no turning back. You may start with historical stats and capabilities for year X, but as gametime moves forward, with each and every game simmed in the leagues the likelihood of adhering to a standard result with minimal deviation becomes less and less of a possibility, and it only grows worse with time. The analogy I think of is our own lives. Consider how your life might be different if a handful of people you know never entered your life. I think that a line from Star Wars is appropriate- "Always in motion is the future." The sheer number of variables combining boggle the mind.

So, as much as many of us here would like it, we can't have it both ways. OOTP will provide us with a lot of entertainment I am certain... but it can't be true to history without losing the capability to surprise us- and vice-versa.

At least, not until computers become a lot closer to human capabilities, and that's something I doubt I will see in my lifetime.

[StyxNCa]

Originally Posted by Pdubya64

I don't see the "ideal" that many of us here long for happening in our lifetime. Why? Well, if I have had any grand concept of life demonstrated to me time and again it's this one:

In order to get something, one must give up something. aka "There is no Free Lunch". There is a cost to everything and a consequence to any decision that is made.

So, how is this related to OOTP you ask?

The instant you decide that OOTP is no longer a strictly historical (replay) simulation of baseball where players always follow their RL results, I maintain that you have just given up the capability (at least in our lifetimes) to "have the best of both worlds" where a team will follow their basic history with only a small deviation in performance of its players.

Essentially, there is no turning back. You may start with historical stats and capabilities for year X, but as gametime moves forward, with each and every game simmed in the leagues the likelihood of adhering to a standard result with minimal deviation becomes less and less of a possibility, and it only grows worse with time. The analogy I think of is our own lives. Consider how your life might be different if a handful of people you know never entered your life. I think that a line from Star Wars is appropriate- "Always in motion is the future." The sheer number of variables combining boggle the mind.

So, as much as many of us here would like it, we can't have it both ways. OOTP will provide us with a lot of entertainment I am certain... but it can't be true to history without losing the capability to surprise us- and vice-versa.

At least, not until computers become a lot closer to human capabilities, and that's something I doubt I will see in my lifetime.

I've said any number of times I am willing to give up the historical aspect of the game, since there are other options for people who want that type of game, so Markus can spend all his time on the fictional part since there isn't really any other game out there for that.

Funny how the fictional only people haven't jumped for that idea.

What I have noticed is that, in the past, fictional players would raise all kind of heck if the stats for the era they were playing in were screwed up for their league, so Markus had to get things as historically accurate for the fictional stats to be as realistic as possible, yet for people who use real players for historical replays, players who stats are real and were created in any real season or era, well, we are supposed to accept, according to those same people, ridiculously screwed up results. The same kind of results they would not accept.

[Siv]

I am mostly a semi historical sim fan who likes to manage a team and see if i can make it better or worse with trades and such. I have played a few single season replays and the results were "believable" but not necessarilly accurate based on the individual player performance's.

My wish is that in the single season replays the players that were overused (approaching too many at bats and innings pitched) would be tagged and their ratings would diminish, this I believe would help to bring in line the accuracy of the replay. I know other sims do this and I feel this helps to straighten out the believability of the sim.

[StyxNCa]

Originally Posted by Siv

I am mostly a semi historical sim fan who likes to manage a team and see if i can make it better or worse with trades and such. I have played a few single season replays and the results were "believable" but not necessarilly accurate based on the individual player performance's.

My wish is that in the single season replays the players that were overused (approaching too many at bats and innings pitched) would be tagged and their ratings would diminish, this I believe would help to bring in line the accuracy of the replay. I know other sims do this and I feel this helps to straighten out the believability of the sim.

Well as I stated earlier, that word "believable" doesn't mean the same thing to each person, so that in itself is a problem when discussing that and I can't even pretend to have a solution for it.

I'm just anxious to get back to my fictional leagues. My players have gotten extremely rusty and I'm paying them to do nothing.

[RchW]

Originally Posted by StyxNCa

What I have noticed is that, in the past, fictional players would raise all kind of heck if the stats for the era they were playing in were screwed up for their league, so Markus had to get things as historically accurate for the fictional stats to be as realistic as possible, yet for people who use real players for historical replays, players who stats are real and were created in any real season or era, well, we are supposed to accept, according to those same people, ridiculously screwed up results. The same kind of results they would not accept.

Apples and oranges. Historical players get upset when a name they know does much worse than they did IRL. Or a name they don't know does well. There are other reasons obviously but for the most part historical hackles get raised when individual (team too?) results don't meet the expectations of that year or set of years.

You already stipulated in another post that the era specific league totals are generally close to the mark**. The issue then becomes one of a dynamic sim (OOTP) vs a static replay of historical results. I thought recalc was supposed to push historicals towards the static model but I'm not a pure historical simmer so I don't use recalc.

Markus' use of "Improved Historical Simulation" is completely valid as OOTPD has never claimed to be a static replay sim. Even if you dispute that, the use of "puffery" in marketing and sales claims is legitimate.

If Markus has in fact made programming and structural changes that "show" better results by his and/or beta testing the claim becomes more legit and less "puffery".

StyxNCa ** The overall league stats always seem to look good, so that part is pretty accurate.

fictional players would raise all kind of heck if the stats for the era they were playing in were screwed up for their league, so Markus had to get things as historically accurate for the fictional stats to be as realistic as possible,

Markus gets league totals close. It's not possible to claim any accuracy for individual fictional stats where league structure, set up and strategy is determined by the gamer. Most gripes about outrageous stats can be traced to out of whack league settings. My ongoing gripe is that the default settings in the game do not come close to modern MLB totals. Over the last 3.5 versions I've spent considerable time adjusting league total modifiers, including all minor leagues, player creation, player development and aging, league strategy and so on. To give only one example, double plays often need to be increased by 40% or more via LTM to produce correct results for a modern MLB league.

Doing this I came to realize that the real strength of this game is its volatility. As an example, I wanted modern MLB stats but starting pitcher CG from the 70's. The result is a complete change in MR use compared to the modern game. That's strictly my problem, not OOTP's. If the game were to be "tightened up" to reduce this volatility it may lose some flexibility. I decided to live with it, even though I'd like things to be easier to change.

Originally Posted by StyxNCa

yet for people who use real players for historical replays, players who stats are real and were created in any real season or era, well, we are supposed to accept, according to those same people, ridiculously screwed up results.

To me there is a structural difference in player creation between fictional leagues and historical leagues that may explain this. I imagine that fictional players are created in one large pool that is "fitted" (bell curve) to the league set up. This may allow for a better (smoother) distribution of ratings at league creation.

Historical players are constructed one at a time (I think) from the database or from a roster set. I wonder if the game gets out of whack when it calculates ratings that don't match its internal league structure "fit". Is it possible that some players get modified to satisfy this structure. This is pure speculation on my part.

[dashby]

I'm assuming skewed results in historical leagues will also translate to skewed results in this year's "hyped" best MLB roster ever?

I've played a whole lot of MLB seasons over the years in OOTP and played around with a whole lot of current MLB rosters and never.ever.ever played out a season that was even what could be conceived as similar to the real-life season. Standings are always alarmingly different. Team stats (not just talking about individual) are quite different.

Never understood how a game with as deep and diverse statistical engine could blow it in such a major way.

I do love the Fictional League though. However, I've always purchased the OOTP games hoping upon hope that I can recreate and play out some real-life seasons and get close proximities. Never happened so, the Ficitional League option is just something I have to indulge in order to justify my yearly investment. Not that I'm bitching...Fictional play is fun and a decent 2nd choice.

Keep hoping, though, one of these years I'll be pleasantly surprised with MLB rosters and can simulate accurately with these.

[jabrown6940]

Without an actual schedule, real daily lineups, real injuries,etc. I wouldn't expect to get good replay results. Every little thing affects both individual player, team and league results. Now if someone does take the time to setup these things on a daily basis for every team and the results come out terrible, there may be an issue. Also a computer program follows logic, no matter how random variables are put in, it follows logic. Humans however are totally illogical and anything can happen. Julio Lugo is attempting to prove this all by himself.

[StyxNCa]

Originally Posted by RchW

Apples and oranges. Historical players get upset when a name they know does much worse than they did IRL. Or a name they don't know does well. There are other reasons obviously but for the most part historical hackles get raised when individual (team too?) results don't meet the expectations of that year or set of years.

You already stipulated in another post that the era specific league totals are generally close to the mark**. The issue then becomes one of a dynamic sim (OOTP) vs a static replay of historical results. I thought recalc was supposed to push historicals towards the static model but I'm not a pure historical simmer so I don't use recalc.

Markus' use of "Improved Historical Simulation" is completely valid as OOTPD has never claimed to be a static replay sim. Even if you dispute that, the use of "puffery" in marketing and sales claims is legitimate.

If Markus has in fact made programming and structural changes that "show" better results by his and/or beta testing the claim becomes more legit and less "puffery".

Markus gets league totals close. It's not possible to claim any accuracy for individual fictional stats where league structure, set up and strategy is determined by the gamer. Most gripes about outrageous stats can be traced to out of whack league settings. My ongoing gripe is that the default settings in the game do not come close to modern MLB totals. Over the last 3.5 versions I've spent considerable time adjusting league total modifiers, including all minor leagues, player creation, player development and aging, league strategy and so on. To give only one example, double plays often need to be increased by 40% or more via LTM to produce correct results for a modern MLB league.

Doing this I came to realize that the real strength of this game is its volatility. As an example, I wanted modern MLB stats but starting pitcher CG from the 70's. The result is a complete change in MR use compared to the modern game. That's strictly my problem, not OOTP's. If the game were to be "tightened up" to reduce this volatility it may lose some flexibility. I decided to live with it, even though I'd like things to be easier to change.





To me there is a structural difference in player creation between fictional leagues and historical leagues that may explain this. I imagine that fictional players are created in one large pool that is "fitted" (bell curve) to the league set up. This may allow for a better (smoother) distribution of ratings at league creation.

Historical players are constructed one at a time (I think) from the database or from a roster set. I wonder if the game gets out of whack when it calculates ratings that don't match its internal league structure "fit". Is it possible that some players get modified to satisfy this structure. This is pure speculation on my part.

Well, the AI decisions on who is on active roster and who is on reserve is a bit ridiculous. I pointed out 2 examples earlier. This will also skew things. Even when allowing the maximum # of players on the roster, and manually putting a pitcher in the rotation, the Cy Young winner from that year was allowed to start 9 games? Does anyone really think this is realistic? I still can't figure out how the AI valued Jim Hardin more than Cuellar.

Most of what you said I can't argue with, and your reference to what I said about overall league stats is true. Markus has always claimed he was looking to make the overall stats look correct. Considering his time restrictions, he can't possibly sim season after season and analyze individual stats, so his shortcut was to make the overall stats look right. So, since I have more time on my hands than I care to admit, I am just pointing out things I have seen that he hasn't. What he does with that info is his business.

It is just how randomly the game divides those overall stats among players, particularly pitchers. I can get those same overall stats in other games but those games also make individual/team stats a lot more believable than OOTP.

I believe that if the only accuracy you can get is the league totals while team/individuals stats are whacked, you can not call it historical. You're just using real names to create fictional stats. Just because the names are real doesn't make it historical either.

I don't know, maybe people's ideas of "historical" is different, hence the different points of view. I can't really say Joe Blow's opinion on historical play is wrong and mine is right. It's all a matter of individual preferences. All I am saying is that for me it doesn't work. I have admitted in the past that as far as suspension of disbelief, I don't have that ability.

[skunkle333]

Since the discount ends Wednesday...which is the day my good ol' Social Security is deposited I will be tempted to pre-order v10. I've ended up getting OOTP the past 3-4 years. But each time my play time goes less than a month or two taking up maybe 10-20% of my baseball gaming.

I really did enjoy my time with OOTP9 last year. My interests in the historical realm are limited. I gave up years back thinking that OOTP can get close to what I consider statistically acceptable replays for real life past seasons.

However...the biggest part of my baseball gaming the past two years has been special tournaments and replaying WSeries matches and the season ending playoffs. My big problem with OOTP is my old brains don't do well setting these things up. So last year a fine fella here PStrickert did the work for me setting up just the White Sox and Dodgers of 1959 so I could replay that World Series (I'm a WSox addict...Nellie Fox fan).

And I really enjoyed the play and all the pizzaz and perks in OOTP. Player pics...sounds...ball flight etc. And I thought last years pbp was very well done. The problem that bothered me the most was the managing of the AI for the Dodgers and the fact that my eyes had problems with the small play by play type. I use pitch by pitch about 80-90% of the time. If you wanted larger type then you totally lost the park screen etc with the big box that would open on the screen. That was a killer for me.

I'm considering grabbing this version again just for playoffs and WSeries replays. I like the possibility now to see the ballpark pic in the center of the screen and making other screen adjustments. However, I haven't read anything about being able to increase the script type just a few levels. And assuming you can do that will you again lose the screen look with this giant box opening for the larger script? I only need to increase the type about 2 or 3 levels for comfort.

Also I would hope setting up just playoffs etc wouldn't be such a difficult task so I wouldn't have bother PStrickert again. He probably wouldn't have the time for me again especially since it's a lot of work and I didn't stick with it very long.

My main baseball play is with DMB. But I like OOTP and Action for those times when I want player pics and a little visual sparkle in my play.

Just my thoughts.

[fantom1979]

Thats a bold statement you are making