Offensive offense

mookie1 · 11-19-2020, 04:04 PM

Over the last several months, I've been running a fun historical project where I take the top 8 teams from each league, starting in 1969, and run a tournament to see who comes out on top. There have been numerous surprises (the 1976 Indians being the most recent), and I hope to write all about it some day. I have been happy with run production falling in line with historical averages, even accounting for the fact that I am allowing the DH in the National League.

I had one hiccup in 1971 when I believe that I messed up the stats modifiers so it was running under 2019 settings. As expected, this led to an offensive explosion. But I have encountered another one in the first round of my 1977 tournament. In real life, scoring jumped from 1976 to 1977 by more than half a run, largely due to the addition of 2 expansion teams. So I certainly expected a bump, but not to the degree that I'm seeing.

19 out of 28 games so far have had a combined score matching or exceeding 11 runs. One of these games was 22-17, which would have been the second highest scoring game ever at that time. 1977 scoring in real life on average per team was 4.47 runs per game, so I'm doubling that number and adding a full run to that combined average for the NL DH (that's generous, in my opinion) to set my bar for what should be an average of 11 total runs per game.

I know: small sample size. But for fun, I took a random sample of 15 real life dates during the 1977 season, and on only one of those days did more than half the games have 11 runs or more (and that was on a light schedule of only 7 games). Hardly definitive, maybe even a stupid comparison, but that's all I've got.

Given the numerous times this sort of complaint has popped up anecdotally, given the handful of times that I have seen this in earlier projects, I can't shake the feeling that something is wrong here. I know it's probably small sample size, but I'm having trouble fully believing that.

Any other possibilities? And one question -- one variable in the equation is that I updated my ballparks with a bunch of SilvaM downloads. Is it possible that ballpark factors are out of whack on those? Is there any adjustment that needs to be made when swapping out ballparks like that?

Charlie Hough · 11-19-2020, 06:30 PM

As you said, it may be a small sample size. But I wonder whether some of this inflation is happening due to the fact that you're playing with the top 8 teams from each league during a season when there was a fairly significant increase in offense.

That could be creating a double effect where the overall batting talent among the top 8 teams is already higher than the usual distribution you see among teams, and then the expansion year effect may be amplifying their output. Then again, the quality of pitching among the top 8 teams should usually help balance things out.

This may resolve itself as you simulate more games. But it would be better to take a random sample of games in 1977 involving the teams that are in your tournament. Compare your results to the combined scores of those games involving only those teams, and that is a more direct and valid comparison.

mookie1 · 11-20-2020, 11:19 AM

Thank you for your response.

A few things:

#1) Clearly, I'm an idiot because 4.47 doubled would give an average of 9 per game. So the bar should be lowered for my comparison.

#2) I loved your idea so much about looking at that subset of teams that I went through the 1977 schedule and picked out games between the teams playing each other in my tournament. I am again adding an extra run as an average for my league because of the NL DH (and, again, I think that's overly generous). Here's the result:

Reality (>= 9 runs per game, combined):

KCR vs CAL (7 out of 15)
NYY vs MIN (4 out of 10)
BOS vs CHW (6 out of 10)
BAL vs TEX (0 out of 10)

PHI vs SFG (7 out of 12)
LAD vs CHC (3 out of 12)
PIT vs HOU (4 out of 12)
CIN vs STL (6 out of 12)

Total: 37 out of 93 games (approx 40%) had a combined total of 9 runs or more.

My tournament (>= 10 runs per game, combined):

KCR vs CAL (2 out of 4)
NYY vs MIN (3 out of 4)
BOS vs CHW (4 out of 4)
BAL vs TEX (2 out of 4)

PHI vs SFG (2 out of 4)
LAD vs CHC (2 out of 4)
PIT vs HOU (3 out of 4)
CIN vs STL (4 out of 4)

Total: 22 out of 28 games (approx 79%) had a combined total of 10 runs or more.

Conclusion? Probably best to complete the tournament (which should end up in the vicinity of 93 total games) before reaching one. But my claim that the scores are skewed too high is looking good at the moment. Of course, last night after compiling this, I played 4 more games, and only 1 hit the "too high" threshold. So maybe it will even out. But not looking likely.

Charlie Hough · 11-20-2020, 02:02 PM

Nice update, and I agree that you should finish out the tournament and then analyze things. One other thing you might want to do is calculate the normal statistical variation you could expect to see based on probabilities.

Sometimes things can seem skewed in your results, and they might be statistically improbable, but they are within the statistical variance you can reasonably expect to see based on probability theory.

In other words, even in those real life games, theoretically there was a statistical probability of the combined score being considerably higher. If enough of those instances happen to occur over a stretch of games, then you might see results similar to what you're getting.

You might be getting an improbable result, but it's likely within the normal realm of statistical variance.

Brad K · 11-20-2020, 02:02 PM

Your input - the team selection - is vastly different from historical. Why do you expect historical output? And if it is different, with so many things changed, how can you identify the cause?

Charlie Hough · 11-20-2020, 02:07 PM

Quote:

Originally Posted by Brad K

Your input - the team selection - is vastly different from historical. Why do you expect historical output? And if it is different, with so many things changed, how can you identify the cause?

I don't think the OP is expecting historical output. The issue is that he has already run a number of tournaments using the same method of team selection, but this particular instance is producing run scoring much higher than all previous tournaments. I suspect that it's all within the realm of statistical variance, but the OP is curious as to what might be causing this, especially since the teams happen to be from an MLB expansion year when run scoring increased.

Brad K · 11-20-2020, 02:37 PM

He's expecting historical output plus an adjustment factor for DH. Results are higher than that.

Charlie Hough · 11-20-2020, 03:03 PM

Quote:

Originally Posted by Brad K

He's expecting historical output plus an adjustment factor for DH. Results are higher than that.

Sure, it seems the OP was happy with the previous tournaments because the run scoring was similar to historical output plus a "fudge" factor for the DH. But if those tournaments were in line with historical numbers, then I suppose it's not unreasonable that he now expects future sims to be in line with historical output as well.

However, I think the OP might be overlooking that the real life historical output could have been significantly different based on statistical variance. That's the one thing that people don't usually consider when they're using real life stats for comparison. Those stats could have been markedly different.

For example, in 1971 and 1973, Willie Stargell had a huge increase in his home run totals and slugging percentage compared to the seasons before and after those years. There are numerous factors that probably contributed to this, but the slugging percentages show that it wasn't just from getting more plate appearances.

Since the totals are separated by a year in between, did he suddenly develop more power, then lose it, then regain it, and then lose it again? I doubt it. A big part of it is probably statistical variance, where he happened to make the right contact or get the right pitches, or the right conditions converged more often.

However, if this happened in an OOTP league, especially if a player's power rating didn't change, a lot of users would think something must be "wrong." But it's probably within the range of statistical variance.

mookie1 · 12-06-2020, 10:54 PM

Thank you both for your responses. I completed the tournament, and scores remained higher than expected -- 51 out of 85 (60%) were 10 runs or higher. I concur that the likelihood of statistical variance accounting for the problem is high. Still, it's unsettling to see a 7 game series between the two best AL teams (Yankees and Royals) with these scores:

17-3, 8-2, 13-2, 15-6, 12-4, 7-0, 13-7

And the corresponding NL series (Reds and Astros) had its last 2 games with scores of 24-4 and 16-2.

Statistical variance or not, I can't think of a single "real" series with consistent scores along those lines -- the 1960 World Series springs to mind, but even that managed to squeeze in a handful of lower scoring contests.

It is what it is. Probably random, doesn't feel that way. I am going to start the 1978 tournament soon, and we'll see if the trend continues. I do wonder if OOTP makes some calculation based on prior years, and when one of those years has a jump in offense, it skews the numbers somehow. I should know more in 1979, as that season has a nearly half run jump as well.

Thanks again.

11-19-2020, 04:04 PM	#1
mookie1 Minors (Rookie Ball) Join Date: Feb 2011 Posts: 46	Offensive offense Over the last several months, I've been running a fun historical project where I take the top 8 teams from each league, starting in 1969, and run a tournament to see who comes out on top. There have been numerous surprises (the 1976 Indians being the most recent), and I hope to write all about it some day. I have been happy with run production falling in line with historical averages, even accounting for the fact that I am allowing the DH in the National League. I had one hiccup in 1971 when I believe that I messed up the stats modifiers so it was running under 2019 settings. As expected, this led to an offensive explosion. But I have encountered another one in the first round of my 1977 tournament. In real life, scoring jumped from 1976 to 1977 by more than half a run, largely due to the addition of 2 expansion teams. So I certainly expected a bump, but not to the degree that I'm seeing. 19 out of 28 games so far have had a combined score matching or exceeding 11 runs. One of these games was 22-17, which would have been the second highest scoring game ever at that time. 1977 scoring in real life on average per team was 4.47 runs per game, so I'm doubling that number and adding a full run to that combined average for the NL DH (that's generous, in my opinion) to set my bar for what should be an average of 11 total runs per game. I know: small sample size. But for fun, I took a random sample of 15 real life dates during the 1977 season, and on only one of those days did more than half the games have 11 runs or more (and that was on a light schedule of only 7 games). Hardly definitive, maybe even a stupid comparison, but that's all I've got. Given the numerous times this sort of complaint has popped up anecdotally, given the handful of times that I have seen this in earlier projects, I can't shake the feeling that something is wrong here. I know it's probably small sample size, but I'm having trouble fully believing that. Any other possibilities? And one question -- one variable in the equation is that I updated my ballparks with a bunch of SilvaM downloads. Is it possible that ballpark factors are out of whack on those? Is there any adjustment that needs to be made when swapping out ballparks like that?

11-19-2020, 06:30 PM	#2
Charlie Hough Hall Of Famer Join Date: May 2006 Posts: 3,642	As you said, it may be a small sample size. But I wonder whether some of this inflation is happening due to the fact that you're playing with the top 8 teams from each league during a season when there was a fairly significant increase in offense. That could be creating a double effect where the overall batting talent among the top 8 teams is already higher than the usual distribution you see among teams, and then the expansion year effect may be amplifying their output. Then again, the quality of pitching among the top 8 teams should usually help balance things out. This may resolve itself as you simulate more games. But it would be better to take a random sample of games in 1977 involving the teams that are in your tournament. Compare your results to the combined scores of those games involving only those teams, and that is a more direct and valid comparison. Last edited by Charlie Hough; 11-19-2020 at 06:32 PM.

11-20-2020, 11:19 AM	#3
mookie1 Minors (Rookie Ball) Join Date: Feb 2011 Posts: 46	Thank you for your response. A few things: #1) Clearly, I'm an idiot because 4.47 doubled would give an average of 9 per game. So the bar should be lowered for my comparison. #2) I loved your idea so much about looking at that subset of teams that I went through the 1977 schedule and picked out games between the teams playing each other in my tournament. I am again adding an extra run as an average for my league because of the NL DH (and, again, I think that's overly generous). Here's the result: Reality (>= 9 runs per game, combined): KCR vs CAL (7 out of 15) NYY vs MIN (4 out of 10) BOS vs CHW (6 out of 10) BAL vs TEX (0 out of 10) PHI vs SFG (7 out of 12) LAD vs CHC (3 out of 12) PIT vs HOU (4 out of 12) CIN vs STL (6 out of 12) Total: 37 out of 93 games (approx 40%) had a combined total of 9 runs or more. My tournament (>= 10 runs per game, combined): KCR vs CAL (2 out of 4) NYY vs MIN (3 out of 4) BOS vs CHW (4 out of 4) BAL vs TEX (2 out of 4) PHI vs SFG (2 out of 4) LAD vs CHC (2 out of 4) PIT vs HOU (3 out of 4) CIN vs STL (4 out of 4) Total: 22 out of 28 games (approx 79%) had a combined total of 10 runs or more. Conclusion? Probably best to complete the tournament (which should end up in the vicinity of 93 total games) before reaching one. But my claim that the scores are skewed too high is looking good at the moment. Of course, last night after compiling this, I played 4 more games, and only 1 hit the "too high" threshold. So maybe it will even out. But not looking likely.

11-20-2020, 02:02 PM	#4
Charlie Hough Hall Of Famer Join Date: May 2006 Posts: 3,642	Nice update, and I agree that you should finish out the tournament and then analyze things. One other thing you might want to do is calculate the normal statistical variation you could expect to see based on probabilities. Sometimes things can seem skewed in your results, and they might be statistically improbable, but they are within the statistical variance you can reasonably expect to see based on probability theory. In other words, even in those real life games, theoretically there was a statistical probability of the combined score being considerably higher. If enough of those instances happen to occur over a stretch of games, then you might see results similar to what you're getting. You might be getting an improbable result, but it's likely within the normal realm of statistical variance.

11-20-2020, 02:02 PM	#5
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	Your input - the team selection - is vastly different from historical. Why do you expect historical output? And if it is different, with so many things changed, how can you identify the cause?

11-20-2020, 02:37 PM	#7
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	He's expecting historical output plus an adjustment factor for DH. Results are higher than that.

12-06-2020, 10:54 PM	#9
mookie1 Minors (Rookie Ball) Join Date: Feb 2011 Posts: 46	Thank you both for your responses. I completed the tournament, and scores remained higher than expected -- 51 out of 85 (60%) were 10 runs or higher. I concur that the likelihood of statistical variance accounting for the problem is high. Still, it's unsettling to see a 7 game series between the two best AL teams (Yankees and Royals) with these scores: 17-3, 8-2, 13-2, 15-6, 12-4, 7-0, 13-7 And the corresponding NL series (Reds and Astros) had its last 2 games with scores of 24-4 and 16-2. Statistical variance or not, I can't think of a single "real" series with consistent scores along those lines -- the 1960 World Series springs to mind, but even that managed to squeeze in a handful of lower scoring contests. It is what it is. Probably random, doesn't feel that way. I am going to start the 1978 tournament soon, and we'll see if the trend continues. I do wonder if OOTP makes some calculation based on prior years, and when one of those years has a jump in offense, it skews the numbers somehow. I should know more in 1979, as that season has a nearly half run jump as well. Thanks again.