Rare things that NEVER happen

muz · 08-14-2013, 02:29 AM

Perhaps somebody can give me some insight here.

1. I've played now over 100 seasons on OOTP 14, spread over two games. One was a real life game, one a fictional game.

In both games combined, never did a player hit 4 HR in a game. Most seasons, there were multiple players in both leagues who hit 40 HR, a few 50 HR seasons, even a 62 HR season. Yet in 100 seasons, nobody could do what Mark Whiten and Bob Horner both did. That doesn't make sense.

Any ideas?

2. Hitting streaks. These are way too uncommon. Even 25 game hitting streaks are very rare in OOTP; while they are rare in MLB as well, they aren't unheard of. Dan Uggla and Jimmy Rollins, of all people, had 30+ game streaks within a 5 year span (and I think there was one more as well).

I think part of the problem is that the AI will 1) pinch hit for star players in blowouts and 2) use star players to pinch hit in off days. Both factors obviously challenge hitting streaks.

But I set a bunch of star players to never pinch hit or be pinch hit for, and still the hitting streaks weren't happening.

So, to experiment, I created a super-batter. Maxed out his contact ratings and made him rarely walk. Batted him at the top of a stacked lineup to get lots of ABs (more than 700 each season). For four years in a row, he hit about .420 and had between 280-300 hits each season.

His hitting streaks during that span? 51 games, 37 games, 32 games and nothing else above 30. Only a handful of 20+ streaks. These are way too low and infrequent for a guy who *averages* almost 2 hits a game. If he never gets to 56 -- that I can understand. But over a full season, it's hard to believe that he wouldn't at least once put together 40 games with at least one hit.

Admittedly, the artificiality of this player makes it hard to estimate what his streaks would be in real life. But given that streaks are so rare overall, I think this is further evidence that something is amiss. Plus, this ties in with the observation in #1: he actually never got more than 6 hits in a game (6-7), and only once batted 1.000 with at least 5 hits (one 5-5 game). Again, we're talking about a guy with 300 hits for a season. He seems to get too many 4 hit games, and not enough 1 hit games and not enough 6 or 7 hit games.

Any ideas would be appreciated.

VanillaGorilla · 08-14-2013, 02:52 AM

A .400 hitter getting 5 AB in a game is a 92% favorite to get a hit.

A .400 hitter getting 5 AB each game is an 11/9 underdog to hit in 10 straight games.

A .400 hitter getting 5 AB each game is a 4/1 underdog to hit in each of 20 games.

A .400 hitter getting 5 AB each game is an 11/1 underdog to hit in each of 30 games.

A .400 hitter getting 5 AB each game is a 24/1 underdog to hit in each of 40 games.

A .400 hitter getting 5 AB each game has a 2.8% chance of hitting in each of 44 games and tying Pete Rose's NL mark.

A .400 hitter getting 5 AB each game has a 1.7% chance of hitting in each of 50 games.

A .400 hitter getting 5 AB each game has a 1.1% chance of hitting in 56 of 56 games, tying Joltin Joe.

A .400 hitter getting 5 AB each game has a 1% chance of hitting in 57 of 57 games.

You ran 4 seasons. Your guy probably failed to get a hit about 10 times a year. So you have run about 40 trials. Try six more seasons.

muz · 08-14-2013, 03:13 AM

Thanks for the reply. I'm not sure your analysis is correct, though. Maybe I'm missing something.

It looks to me that you've calculated the odds of getting a hit in every one of X games (i.e. 44 out of 44, 56 out of 56, etc). But in any given season, there are about 100 or so 56 game stretches, and around 130 30 game stretches. Thus, to find the probability of getting a streak of X length, you'd have to calculate the chances of getting hits in all games of each X game stretch in the season.

Of course, because the stretches and the hitting streaks overlap, they aren't independent events. The probabilities can't just be added together. The statistics, I think, are complex.

What I find suspect isn't that the guy didn't match DiMags, but rather that he didn't routinely get more 20 or 30 game streaks. DiMags streak is ridiculous and we all know that. But the difference between a 30 game streak and a 56 game streak is huge.

To put it differently, if he had 10 games in a season without hits, the odds are not good that they would be so evenly distributed as to not allow a 30 game streak.

VanillaGorilla · 08-14-2013, 04:00 AM

Quote:

Originally Posted by muz

Thanks for the reply. I'm not sure your analysis is correct, though. Maybe I'm missing something.

You are welcome.

Quote:

Originally Posted by muz

It looks to me that you've calculated the odds of getting a hit in every one of X games (i.e. 44 out of 44, 56 out of 56, etc). But in any given season, there are about 100 or so 56 game stretches, and around 130 30 game stretches. Thus, to find the probability of getting a streak of X length, you'd have to calculate the chances of getting hits in all games of each X game stretch in the season.

Not so.

Quote:

Originally Posted by muz

Of course, because the stretches and the hitting streaks overlap, they aren't independent events. The probabilities can't just be added together. The statistics, I think, are complex.

Each hitting streak is an individual event. Once the player gets 0 hits the streak is completed and the next one begins with the next game in which he gets a hit.

Quote:

Originally Posted by muz

What I find suspect isn't that the guy didn't match DiMags, but rather that he didn't routinely get more 20 or 30 game streaks. DiMags streak is ridiculous and we all know that. But the difference between a 30 game streak and a 56 game streak is huge.

For a player that has 10 streaks during a 162 game season (a streak can be only one game, here) who hits .400 and gets 5 AB each and every game, the probability of him having one streak of 30 games is about 0.6. That would include carrying over whatever streak he might be on from game 162 into the following season.

Quote:

Originally Posted by muz

To put it differently, if he had 10 games in a season without hits, the odds are not good that they would be so evenly distributed as to not allow a 30 game streak.

Actually, the odds of a .400 hitter not having at least one* 30 game hit streak in a season where he plays each game and has 5 AB in every game are about the same as his odds of getting a hit in any individual AB. Whether or not the chances of either are "not good" is a qualitative determination.

ADD: With a 60% chance for at least one 30 game hit streak in a season, we can here, through addition, say that we expect 2.4 seasons out of 4 with a 30 game streak. Though you do not indicate if any of his three 30 game streaks were in the same season, the fact that you observed 3 in four seasons is about what would be expected.

Sticking with 10 streaks a season over 4 seasons for a trial set of 40, the probability of a .400 hitter who gets 5 AB every game and plays in each of 162 games over 4 seasons to have fewer than 3 30 game hit streaks is 0.30. The probability of him getting exactly 3 is 0.23. The probability of him getting more than 3 is 0.47. Based on these parameters, the odds are against your guy having more than 3 30 game streaks over the course of 4 seasons.

Your posted results are expected.

muz · 08-14-2013, 05:01 AM

I confess that I have 1) not thought deeply about this and 2) that I am not an expert on stats and probability. However, I have thought it through somewhat and I have some knowledge of statistics, so that's the authority level with which I claim to speak.

I don't think the issue is whether a hitting streak is an individual event. The question is how you are calculating its odds. In my opinion, the analysis requires a complex series of calculations.

For simplicity, let's assume a 50 game season. What are the odds of a player hitting 30 games in a row? Well, the prob of him hitting in the first 30 games is easy to calculate: it's X^30, where X is the chance of getting at least one hit in a game.

The problem, though, is that this 30 game streak could fail at the outset. So the odds of him getting a 30 game streak in the first 31 games is: odds of hitting in first 30 games, plus the odds of him hitting in the last 30 games conditional on not hitting in the first game.

When you ask about the whole season, it's the probability of hitting in the first 30 games, plus the odds of hitting in games 2-31 conditional on not hitting in game 1, plus the odds of hitting in games 3-32 conditional on not hitting in one of the first games, plus the odds of hitting in games 4-33 conditional on not hitting in one of the first 3 games, etc.

So the formula looks something like (I'm sure I don't have it exactly correct, but I think I'm in the ballpark) -- (chance of a hit in 30 out of 30 games) * (sum of chances of not hitting in one game, two games, three games, etc). For a .400 hitter, the second parenthesis would be about 10, according to your original calculations. That is just for a 50 game season. For a 162 game season, it would be larger, but I don't know by how much (it's not a linear function).

Maybe if I get some time, I will do a random simulation with a sample size in the thousands.

VanillaGorilla · 08-14-2013, 05:44 AM

Quote:

Originally Posted by muz

I confess that I have 1) not thought deeply about this and 2) that I am not an expert on stats and probability. However, I have thought it through somewhat and I knew a few things about stats, so that's the authority level with which I claim to speak.

I don't think the issue is whether a hitting streak is an individual event. The question is how you are calculating its odds. In my opinion, the analysis requires a complex series of calculations.

It's very straight forward. The complexity is minimized thanks to software.

Quote:

Originally Posted by muz

For simplicity, let's assume a 50 game season. What are the odds of a player hitting 30 games in a row? Well, the prob of him hitting in the first 30 games is easy to calculate: it's X^30, where X is the chance of getting at least one hit in a game.

So far so good.

Quote:

Originally Posted by muz

The problem, though, is that this 30 game streak could fail at the outset. So the odds of him getting a 30 game streak in the first 31 games is: odds of hitting in first 30 games, plus the odds of him hitting in the last 30 games conditional on not hitting in the first game.

When you ask about the whole season, it's the probability of hitting in the first 30 games, plus the odds of hitting in games 2-31 conditional on not hitting in game 1, plus the odds of hitting in games 3-32 conditional on not hitting in one of the first games, plus the odds of hitting in games 4-33 conditional on not hitting in one of the first 3 games, etc.

So the formula looks something like (I'm sure I don't have it exactly correct, but I think I'm in the ballpark) -- (chance of a hit in 30 out of 30 games) * (sum of chances of not hitting in one game, two games, three games, etc). For a .400 hitter, the second parenthesis would be about 10, according to your original calculations.

Maybe if I get some time, I will do a random simulation with a sample size in the thousands.

You are complicating this problem. I also erred in something I wrote, previously. I will not edit it, but revise it here: each trial begins after a game without a hit, it doesn't start when he gets a hit. It was poor language use for me to have intermingled "streak" and "trial".

I also threw in a discrepancy by using a .400 hitter when you have a .420 hitter. I based getting a hit in 152 games on that number, not .400 (a .400 hitter would expect to go hitless 13 times, rounding from 12.6).

All of this is sticking with 5 AB and 5 AB only each and every game played.

Your .420 hitter has a 0.934 probability of getting a hit in a given game (am calculating with longer numbers, but rounding here).

When he enters a game after having 0 hits in the previous game, that is when we start counting. We do not need to sum the chances of not getting in a hit in game 1,2,3...etc. We are just increasing the exponent. The chance of getting a hit implies the chance of not getting a hit. No need to introduce the chance of not getting a hit, it is accounted for already.

The probability of a .420 hitter hitting in 30 consecutive starting at the beginning of any game is 0.13. The idea of having 10 trials for a season is sound, because we don't count a 33 game hit streak as 4 30 game hit streaks.

Now, using .420 for the basis, we have a 0.753 chance of getting a 30 game hit streak in a season that is defined by 10 attempts. This would indicate a 30 game streak in three of 4 seasons. You reported 3 30 game streaks.

The probability of the .420 hitter having fewer than 3 30 game hit streaks in 4 seasons is 0.124. The probability of having exactly 3 is 0.092. The probability of more than 3 is 0.784.

Using .420 as the basis we would expect to see more than 3 instances in 40 trials. However, we would not expect to see more than 4 instances in 40 trials as the player is an underdog to have 5 30 game hit streaks in 40 trials.

Even now using .420 instead of .400 as the basis, your reported results do not look to be unexpected.

The Game · 08-14-2013, 06:50 AM

Quote:

Originally Posted by muz

Perhaps somebody can give me some insight here.

1. I've played now over 100 seasons on OOTP 14, spread over two games. One was a real life game, one a fictional game.

In both games combined, never did a player hit 4 HR in a game. Most seasons, there were multiple players in both leagues who hit 40 HR, a few 50 HR seasons, even a 62 HR season. Yet in 100 seasons, nobody could do what Mark Whiten and Bob Horner both did. That doesn't make sense.

Any ideas?

2. Hitting streaks. These are way too uncommon. Even 25 game hitting streaks are very rare in OOTP; while they are rare in MLB as well, they aren't unheard of. Dan Uggla and Jimmy Rollins, of all people, had 30+ game streaks within a 5 year span (and I think there was one more as well).

I think part of the problem is that the AI will 1) pinch hit for star players in blowouts and 2) use star players to pinch hit in off days. Both factors obviously challenge hitting streaks.

But I set a bunch of star players to never pinch hit or be pinch hit for, and still the hitting streaks weren't happening.

So, to experiment, I created a super-batter. Maxed out his contact ratings and made him rarely walk. Batted him at the top of a stacked lineup to get lots of ABs (more than 700 each season). For four years in a row, he hit about .420 and had between 280-300 hits each season.

His hitting streaks during that span? 51 games, 37 games, 32 games and nothing else above 30. Only a handful of 20+ streaks. These are way too low and infrequent for a guy who *averages* almost 2 hits a game. If he never gets to 56 -- that I can understand. But over a full season, it's hard to believe that he wouldn't at least once put together 40 games with at least one hit.

Admittedly, the artificiality of this player makes it hard to estimate what his streaks would be in real life. But given that streaks are so rare overall, I think this is further evidence that something is amiss. Plus, this ties in with the observation in #1: he actually never got more than 6 hits in a game (6-7), and only once batted 1.000 with at least 5 hits (one 5-5 game). Again, we're talking about a guy with 300 hits for a season. He seems to get too many 4 hit games, and not enough 1 hit games and not enough 6 or 7 hit games.

Any ideas would be appreciated.

14 may be different then past versions. i have been around since 11 and I just had my 1st 4 HR game. Everything was perfect for that to happen. Clone of Mickey Mantle, wind at 15 mph blowing out to RF. RF wall of 281, a team struggling when they came in. IRL i think most teams would have IBB him after the 3rd one. It was a 9-9 tie game in the 7th at the time 2 outs I believe and the next guy was 0-4 in the game.
So far on 14 my longest hit streak is 37 games. I have seen 5 30+ game hit streaks between 4 leagues. Longest overall was 47 games on 11 for a career .291 hitter.
Hitting streaks of 40+ games are extremely rare IRL and OOTP reflects that. Personally I think 14 makes a Perfect Game, Triple Play, unassisted Triply Play and a 4 HR game too easy. I had none of these on 13 and only the Perfect Game on 12 & 11.
What your best modern hitting streak vs your best pre 1970 hitting streak?
While i have yet to get very far in my historical league on 14, in past versions 40+ game hitting streak would happen every 10 yrs or so pre 1970 and once every 20 yrs post 1970. This was on 11 - 13.
Wee Willie Keller holds the record at 51. Joe D's longest streak was 43 games. Pete Rose at 41. OOTP will never be exact RL and i like that.

D-BacksJosh · 08-14-2013, 10:26 AM

I love the smell of hardcore mathematics in the morning.

muz · 08-14-2013, 10:43 AM

I ran some random simulations of a .420 hitter getting 5 AB per game for 156 games per season (that's about how many my guy played).

In a 1000 season sample, this hypothetical hitter got about 1245 30 game hitting streaks. This doesn't count streaks that overlap from season to season. That is, starting with a streak of zero on opening day every season, I'm getting, on average, around 1.25 30 game hitting streaks per season.

Of those hitting streaks, 862 were 35 games or longer, or almost one per season. 594 were 40 games or longer, or about 60% of seasons. 37% of the seasons featured a 45 game hitting streak; 29% a 50 game streak; and 17% a 56 game streak.

So the results in OOTP underperformed -- I had 3 30 game streaks in 4 seasons, when 5 would be expected. Granted, the 4 season sample size is pretty small, but this is consistent with my general observation that hitting streaks are too short and rare.

It could be, though, that variation in # of ABs per game makes a big difference. We've been assuming here 5 AB per game, which is about right on average. But some games he would have only 4 (or, I suppose, possibly 3 if he walked in a game where he only got 4 plate appearances). The effect of sometimes getting only 4 ABs is not cancelled out by sometimes getting 6, so that would tend to shorten hitting streaks.

As for 4 HR games, I think we're all in agreement that there are more in real life than in OOTP. Any ideas why?

Matt Arnold · 08-14-2013, 10:51 AM

Takeaway is that hitting streaks are rare even for great players. I've definitely not done insane amounts of testing, but every season I do see a handful of players with 20 game hitting streaks, and once every couple seasons I think someone hits 30. As far as I'm concerned, it's not significantly off from real life.

I don't think I've seen a 4HR game either, but I don't always pay attention to the whole league. And if you actually only look at 1 team, even 100 seasons might not be enough. I mean, there have been something like 16 or 17 of them in the last 100 years in the MLB as a whole, which is how many team-seasons? For one team, they probably only average 1 4hr-game per team per 100 years, so the fact that you haven't seen one is probably not all that uncommon. If you haven't seen one league-wide in that time, then maybe there's an issue. I imagine OOTP could add some more adjusting, maybe changing players' hot vs cold in-game, so that after the first 2-3 HRs in the game, he's hot and thus boosts his power slightly. Although a tweak like that is tough to balance properly.

Cinnamon J. Scudworth · 08-14-2013, 11:20 AM

Edit: never mind, different issue.

RchW · 08-14-2013, 11:24 AM

Quote:

Originally Posted by UWHabs

Takeaway is that hitting streaks are rare even for great players. I've definitely not done insane amounts of testing, but every season I do see a handful of players with 20 game hitting streaks, and once every couple seasons I think someone hits 30. As far as I'm concerned, it's not significantly off from real life.

I don't think I've seen a 4HR game either, but I don't always pay attention to the whole league. And if you actually only look at 1 team, even 100 seasons might not be enough. I mean, there have been something like 16 or 17 of them in the last 100 years in the MLB as a whole, which is how many team-seasons? For one team, they probably only average 1 4hr-game per team per 100 years, so the fact that you haven't seen one is probably not all that uncommon. If you haven't seen one league-wide in that time, then maybe there's an issue. I imagine OOTP could add some more adjusting, maybe changing players' hot vs cold in-game, so that after the first 2-3 HRs in the game, he's hot and thus boosts his power slightly. Although a tweak like that is tough to balance properly.

I have 7 4 HR games in 77 years and 5 in 62 years in two of my current leagues. Slightly low but certainly not in the least bit concerning. There are many other statistical variations far more important than 4HR games. I'll check after 137 seasons.

VanillaGorilla · 08-14-2013, 11:24 AM

Quote:

Originally Posted by muz

In a 1000 season sample, this hypothetical hitter got about 1245 30 game hitting streaks. This doesn't count streaks that overlap from season to season. That is, starting with a streak of zero on opening day every season, I'm getting, on average, around 1.25 30 game hitting streaks per season.

I posted that a .420 hitter has a 13% (rounded prob) chance of a 30 game streak at the start of the trial. If he gets 10 trials a season and you run 1000 seasons, you would expect to get 1305 (using actual prob) 30 game streaks.

If you reduce your players GP to 156 out of 162, then we need to reduce the expected total by the same 3.7%. When we do that we get an expected number of 30 game hit streaks 1256, which compares to your 1245.

Quote:

Originally Posted by muz

So the results in OOTP underperformed -- I had 3 30 game streaks in 4 seasons, when 5 would be expected. Granted, the 4 season sample size is pretty small, but this is consistent with my general observation that hitting streaks are too short and rare.

No. Your sample size of 4 seasons is wayyyyyyyyyyyyyyyy too small to reach that conclusion. Yes, 5 is the expected number (I brain cramped in earlier post saying it was 4....am not editing that, either) but having only 3 is still a better than a 20% chance in outcome from that sample size.

Quote:

Originally Posted by muz

It could be, though, that variation in # of ABs per game makes a big difference. We've been assuming here 5 AB per game, which is about right on average. But some games he would have only 4 (or, I suppose, possibly 3 if he walked in a game where he only got 4 plate appearances). The effect of sometimes getting only 4 ABs is not cancelled out by sometimes getting 6, so that would tend to shorten hitting streaks.

This is correct. I used 5 since you had him at the top of the line-up of an offensive team, and figured that was a fair number to work with. The thing about a real streak is that, as you mention, he may go 0-for-1 with 4 BB and a sac fly, or something, like that. If we used base hit per PA we would get a better read on what we should actually expect for a rate from the sim.

The 5 AB gives a ball park estimate. Your Original Post indicated there were way too few 30 game streaks. The numbers I posted are good, for a baseline rate. Your 1000 sims (however you did that) confirmed what I posted.

The Wolf · 08-14-2013, 11:35 AM

The Streak of Streaks by Stephen Jay Gould | The New York Review of Books

"But 'treasure your exceptions,' as the old motto goes. There is one major exception, and absolutely only one—one sequence so many standard deviations above the expected distribution that it should not have occurred at all. Joe DiMaggio’s fifty-six–game hitting streak in 1941. The intuition of baseball aficionados has been vindicated. Purcell calculated that to make it likely (probability greater than 50 percent) that a run of even fifty games will occur once in the history of baseball up to now (and fifty-six is a lot more than fifty in this kind of league), baseball’s rosters would have to include either four lifetime .400 batters or fifty-two lifetime .350 batters over careers of one thousand games. In actuality, only three men have lifetime batting averages in excess of .350, and no one is anywhere near .400 (Ty Cobb at .367, Rogers Hornsby at .358, and Shoeless Joe Jackson at .356). DiMaggio’s streak is the most extraordinary thing that ever happened in American sports. He sits on the shoulders of two bearers—mythology and science. For Joe DiMaggio accomplished what no other ballplayer has done. He beat the hardest taskmaster of all, a woman who makes Nolan Ryan’s fastball look like a cantaloupe in slow motion—Lady Luck."

Read the whole thing. It's by one of the greatest of American scientists. It starts slow but it delivers.

Le Grande Orange · 08-14-2013, 11:46 AM

According to The SABR Baseball List & Record Book published in 2007 and which covers the 1876 through 2006 seasons:

The number of times in which a player hit four home runs in a game was 15 (of which 2 were in the 19th century).

The number of times in which a player had a hitting streak of 30 or more games within a single season was 42 (of which 7 were in the 19th century). Breaking it down by length, there were 29 with 30-34 games; 7 with 35-39 games; 5 with 40-44 games; none with 45-49 games; none with 50-54 games; and 1 with 55-59 games.

The number of times in which a player had a hitting streak of 30 or more games which spanned two seasons was 10 (of which 2 were in the 19th century). By length two were 30 games, two were 31 games, two were 32 games, one was 35 games, one was 36 games, one was 38 games, and one was 45 games.

Le Grande Orange · 08-14-2013, 12:01 PM

In regards to Joe DiMaggio, while everyone remembers his 56-game hitting streak in the majors, what is less well remembered is that he had a 61-game hitting streak in the minors. It occurred in 1933 while he was playing for San Francisco of the PCL. He was just 18 years old.

The all-time longest hitting streak was set by Joe Wilhoit in 1919. The length? 69 games.

muz · 08-14-2013, 02:07 PM

Quote:

Originally Posted by VanillaGorilla

I posted that a .420 hitter has a 13% (rounded prob) chance of a 30 game streak at the start of the trial. If he gets 10 trials a season and you run 1000 seasons, you would expect to get 1305 (using actual prob) 30 game streaks.

If you reduce your players GP to 156 out of 162, then we need to reduce the expected total by the same 3.7%. When we do that we get an expected number of 30 game hit streaks 1256, which compares to your 1245.

If you say so. I don't mean that to be snide; I just don't really understand your methodology. I still think you've oversimplified what is actually a complex mathematical model, but 1) maybe I'm wrong or 2) maybe the simplification doesn't make much difference in the end.

As far as too few 30 game streaks, well there were way too few -- about 40% too few. The problem we're having now is that I didn't run enough seasons to distinguish random chance or some factor in the game.

As for the sims, it's very easy: I just generate a random number to determine whether the guy gets a hit in a game. Do it 156 times and I have results for 156 games. Count the hitting streaks. Repeat X times for X simulations.

muz · 08-14-2013, 02:10 PM

Quote:

Originally Posted by Le Grande Orange

According to The SABR Baseball List & Record Book published in 2007 and which covers the 1876 through 2006 seasons:

The number of times in which a player hit four home runs in a game was 15 (of which 2 were in the 19th century).

The number of times in which a player had a hitting streak of 30 or more games within a single season was 42 (of which 7 were in the 19th century). Breaking it down by length, there were 29 with 30-34 games; 7 with 35-39 games; 5 with 40-44 games; none with 45-49 games; none with 50-54 games; and 1 with 55-59 games.

The number of times in which a player had a hitting streak of 30 or more games which spanned two seasons was 10 (of which 2 were in the 19th century). By length two were 30 games, two were 31 games, two were 32 games, one was 35 games, one was 36 games, one was 38 games, and one was 45 games.

This is interesting. I would have thought the hitting streak number would be larger, but 2006 happens to be an unfortunate cutoff -- there were four more streaks in the next 5 years. Perhaps the recent frequency of streaks gave me the impression that they are generally more common than they are.

Le Grande Orange · 08-14-2013, 02:40 PM

Quote:

Originally Posted by muz

This is interesting. I would have thought the hitting streak number would be larger, but 2006 happens to be an unfortunate cutoff -- there were four more streaks in the next 5 years. Perhaps the recent frequency of streaks gave me the impression that they are generally more common than they are.

Yes, streaks (and other things) can be 'streaky'. In terms of the four home runs in a game, for example, there was a long stretch when there were none at all, then two happened in the same season (2002) within three weeks of each other.

injury log · 08-14-2013, 04:40 PM

You guys are ignoring about a dozen crucially important factors that should make long streaks far less likely than your calculations are telling you:

- players don't average 5 AB per game. They walk, they sacrifice, they get hit by pitches, the rest of their team doesn't hit.

- players leave games due to injury, ejections, etc

- .400 hitters are not .400 hitters in every game. One day they play against Nate Cornejo and they become .500 hitters, and one day they play against Pedro Martinez and they become .300 hitters. And not only are they less likely to hit against Martinez, they're also less likely to get chances to hit, because their whole team will get outs more often.

I've never had the impression that long hitting streaks are too rare in OOTP - if anything, they seem to happen more often than in real life, but I'v enever studied it in detail. As for 4 HR games, OOTP is just using a batter's probability of hitting an HR in each AB. If 4 HR games are less common in OOTP than in real life (and I don't know that they are), then it must be that on rare occasions a player is more likely than in OOTP to hit a HR in each at bat, for whatever reason (extreme weather, because he is 'hot', because on some days pitchers are worse in real life than they ever are in OOTP, something like that).

08-14-2013, 02:52 AM	#2
VanillaGorilla All Star Starter Join Date: Dec 2011 Posts: 1,371	Hit streak A .400 hitter getting 5 AB in a game is a 92% favorite to get a hit. A .400 hitter getting 5 AB each game is an 11/9 underdog to hit in 10 straight games. A .400 hitter getting 5 AB each game is a 4/1 underdog to hit in each of 20 games. A .400 hitter getting 5 AB each game is an 11/1 underdog to hit in each of 30 games. A .400 hitter getting 5 AB each game is a 24/1 underdog to hit in each of 40 games. A .400 hitter getting 5 AB each game has a 2.8% chance of hitting in each of 44 games and tying Pete Rose's NL mark. A .400 hitter getting 5 AB each game has a 1.7% chance of hitting in each of 50 games. A .400 hitter getting 5 AB each game has a 1.1% chance of hitting in 56 of 56 games, tying Joltin Joe. A .400 hitter getting 5 AB each game has a 1% chance of hitting in 57 of 57 games. You ran 4 seasons. Your guy probably failed to get a hit about 10 times a year. So you have run about 40 trials. Try six more seasons.

08-14-2013, 05:01 AM	#5
muz Minors (Rookie Ball) Join Date: Jun 2013 Posts: 39	I confess that I have 1) not thought deeply about this and 2) that I am not an expert on stats and probability. However, I have thought it through somewhat and I have some knowledge of statistics, so that's the authority level with which I claim to speak. I don't think the issue is whether a hitting streak is an individual event. The question is how you are calculating its odds. In my opinion, the analysis requires a complex series of calculations. For simplicity, let's assume a 50 game season. What are the odds of a player hitting 30 games in a row? Well, the prob of him hitting in the first 30 games is easy to calculate: it's X^30, where X is the chance of getting at least one hit in a game. The problem, though, is that this 30 game streak could fail at the outset. So the odds of him getting a 30 game streak in the first 31 games is: odds of hitting in first 30 games, plus the odds of him hitting in the last 30 games conditional on not hitting in the first game. When you ask about the whole season, it's the probability of hitting in the first 30 games, plus the odds of hitting in games 2-31 conditional on not hitting in game 1, plus the odds of hitting in games 3-32 conditional on not hitting in one of the first games, plus the odds of hitting in games 4-33 conditional on not hitting in one of the first 3 games, etc. So the formula looks something like (I'm sure I don't have it exactly correct, but I think I'm in the ballpark) -- (chance of a hit in 30 out of 30 games) * (sum of chances of not hitting in one game, two games, three games, etc). For a .400 hitter, the second parenthesis would be about 10, according to your original calculations. That is just for a 50 game season. For a 162 game season, it would be larger, but I don't know by how much (it's not a linear function). Maybe if I get some time, I will do a random simulation with a sample size in the thousands. Last edited by muz; 08-14-2013 at 05:12 AM.

08-14-2013, 11:20 AM	#11
Cinnamon J. Scudworth All Star Starter Join Date: Jul 2011 Location: Washington, DC Posts: 1,118	Edit: never mind, different issue. __________________ "Sometimes, this is like going to a grocery store. You’ve got a list until you get to the check-out stand. And then you start reading People magazine, and all this other [stuff] ends up in the basket." -Sandy Alderson on the MLB offseason Last edited by Cinnamon J. Scudworth; 08-14-2013 at 11:21 AM.

08-14-2013, 11:46 AM	#15
Le Grande Orange Hall Of Famer Join Date: Feb 2002 Location: Up There Posts: 15,642	According to The SABR Baseball List & Record Book published in 2007 and which covers the 1876 through 2006 seasons: The number of times in which a player hit four home runs in a game was 15 (of which 2 were in the 19th century). The number of times in which a player had a hitting streak of 30 or more games within a single season was 42 (of which 7 were in the 19th century). Breaking it down by length, there were 29 with 30-34 games; 7 with 35-39 games; 5 with 40-44 games; none with 45-49 games; none with 50-54 games; and 1 with 55-59 games. The number of times in which a player had a hitting streak of 30 or more games which spanned two seasons was 10 (of which 2 were in the 19th century). By length two were 30 games, two were 31 games, two were 32 games, one was 35 games, one was 36 games, one was 38 games, and one was 45 games. Last edited by Le Grande Orange; 08-14-2013 at 11:49 AM.

08-14-2013, 02:29 AM	#1
muz Minors (Rookie Ball) Join Date: Jun 2013 Posts: 39	Rare things that NEVER happen Perhaps somebody can give me some insight here. 1. I've played now over 100 seasons on OOTP 14, spread over two games. One was a real life game, one a fictional game. In both games combined, never did a player hit 4 HR in a game. Most seasons, there were multiple players in both leagues who hit 40 HR, a few 50 HR seasons, even a 62 HR season. Yet in 100 seasons, nobody could do what Mark Whiten and Bob Horner both did. That doesn't make sense. Any ideas? 2. Hitting streaks. These are way too uncommon. Even 25 game hitting streaks are very rare in OOTP; while they are rare in MLB as well, they aren't unheard of. Dan Uggla and Jimmy Rollins, of all people, had 30+ game streaks within a 5 year span (and I think there was one more as well). I think part of the problem is that the AI will 1) pinch hit for star players in blowouts and 2) use star players to pinch hit in off days. Both factors obviously challenge hitting streaks. But I set a bunch of star players to never pinch hit or be pinch hit for, and still the hitting streaks weren't happening. So, to experiment, I created a super-batter. Maxed out his contact ratings and made him rarely walk. Batted him at the top of a stacked lineup to get lots of ABs (more than 700 each season). For four years in a row, he hit about .420 and had between 280-300 hits each season. His hitting streaks during that span? 51 games, 37 games, 32 games and nothing else above 30. Only a handful of 20+ streaks. These are way too low and infrequent for a guy who averages almost 2 hits a game. If he never gets to 56 -- that I can understand. But over a full season, it's hard to believe that he wouldn't at least once put together 40 games with at least one hit. Admittedly, the artificiality of this player makes it hard to estimate what his streaks would be in real life. But given that streaks are so rare overall, I think this is further evidence that something is amiss. Plus, this ties in with the observation in #1: he actually never got more than 6 hits in a game (6-7), and only once batted 1.000 with at least 5 hits (one 5-5 game). Again, we're talking about a guy with 300 hits for a season. He seems to get too many 4 hit games, and not enough 1 hit games and not enough 6 or 7 hit games. Any ideas would be appreciated.

08-14-2013, 03:13 AM	#3
muz Minors (Rookie Ball) Join Date: Jun 2013 Posts: 39	Thanks for the reply. I'm not sure your analysis is correct, though. Maybe I'm missing something. It looks to me that you've calculated the odds of getting a hit in every one of X games (i.e. 44 out of 44, 56 out of 56, etc). But in any given season, there are about 100 or so 56 game stretches, and around 130 30 game stretches. Thus, to find the probability of getting a streak of X length, you'd have to calculate the chances of getting hits in all games of each X game stretch in the season. Of course, because the stretches and the hitting streaks overlap, they aren't independent events. The probabilities can't just be added together. The statistics, I think, are complex. What I find suspect isn't that the guy didn't match DiMags, but rather that he didn't routinely get more 20 or 30 game streaks. DiMags streak is ridiculous and we all know that. But the difference between a 30 game streak and a 56 game streak is huge. To put it differently, if he had 10 games in a season without hits, the odds are not good that they would be so evenly distributed as to not allow a 30 game streak.

08-14-2013, 10:26 AM	#8
D-BacksJosh Banned Join Date: May 2011 Location: Arizona Posts: 875	I love the smell of hardcore mathematics in the morning.

08-14-2013, 10:43 AM	#9
muz Minors (Rookie Ball) Join Date: Jun 2013 Posts: 39	I ran some random simulations of a .420 hitter getting 5 AB per game for 156 games per season (that's about how many my guy played). In a 1000 season sample, this hypothetical hitter got about 1245 30 game hitting streaks. This doesn't count streaks that overlap from season to season. That is, starting with a streak of zero on opening day every season, I'm getting, on average, around 1.25 30 game hitting streaks per season. Of those hitting streaks, 862 were 35 games or longer, or almost one per season. 594 were 40 games or longer, or about 60% of seasons. 37% of the seasons featured a 45 game hitting streak; 29% a 50 game streak; and 17% a 56 game streak. So the results in OOTP underperformed -- I had 3 30 game streaks in 4 seasons, when 5 would be expected. Granted, the 4 season sample size is pretty small, but this is consistent with my general observation that hitting streaks are too short and rare. It could be, though, that variation in # of ABs per game makes a big difference. We've been assuming here 5 AB per game, which is about right on average. But some games he would have only 4 (or, I suppose, possibly 3 if he walked in a game where he only got 4 plate appearances). The effect of sometimes getting only 4 ABs is not cancelled out by sometimes getting 6, so that would tend to shorten hitting streaks. As for 4 HR games, I think we're all in agreement that there are more in real life than in OOTP. Any ideas why?

08-14-2013, 10:51 AM	#10
Matt Arnold OOTP Developer Join Date: Jun 2009 Location: Here and there Posts: 16,244	Takeaway is that hitting streaks are rare even for great players. I've definitely not done insane amounts of testing, but every season I do see a handful of players with 20 game hitting streaks, and once every couple seasons I think someone hits 30. As far as I'm concerned, it's not significantly off from real life. I don't think I've seen a 4HR game either, but I don't always pay attention to the whole league. And if you actually only look at 1 team, even 100 seasons might not be enough. I mean, there have been something like 16 or 17 of them in the last 100 years in the MLB as a whole, which is how many team-seasons? For one team, they probably only average 1 4hr-game per team per 100 years, so the fact that you haven't seen one is probably not all that uncommon. If you haven't seen one league-wide in that time, then maybe there's an issue. I imagine OOTP could add some more adjusting, maybe changing players' hot vs cold in-game, so that after the first 2-3 HRs in the game, he's hot and thus boosts his power slightly. Although a tweak like that is tough to balance properly.

08-14-2013, 12:01 PM	#16
Le Grande Orange Hall Of Famer Join Date: Feb 2002 Location: Up There Posts: 15,642	In regards to Joe DiMaggio, while everyone remembers his 56-game hitting streak in the majors, what is less well remembered is that he had a 61-game hitting streak in the minors. It occurred in 1933 while he was playing for San Francisco of the PCL. He was just 18 years old. The all-time longest hitting streak was set by Joe Wilhoit in 1919. The length? 69 games.

08-14-2013, 04:40 PM	#20
injury log Hall Of Famer Join Date: Apr 2007 Location: Toronto Posts: 9,162	You guys are ignoring about a dozen crucially important factors that should make long streaks far less likely than your calculations are telling you: - players don't average 5 AB per game. They walk, they sacrifice, they get hit by pitches, the rest of their team doesn't hit. - players leave games due to injury, ejections, etc - .400 hitters are not .400 hitters in every game. One day they play against Nate Cornejo and they become .500 hitters, and one day they play against Pedro Martinez and they become .300 hitters. And not only are they less likely to hit against Martinez, they're also less likely to get chances to hit, because their whole team will get outs more often. I've never had the impression that long hitting streaks are too rare in OOTP - if anything, they seem to happen more often than in real life, but I'v enever studied it in detail. As for 4 HR games, OOTP is just using a batter's probability of hitting an HR in each AB. If 4 HR games are less common in OOTP than in real life (and I don't know that they are), then it must be that on rare occasions a player is more likely than in OOTP to hit a HR in each at bat, for whatever reason (extreme weather, because he is 'hot', because on some days pitchers are worse in real life than they ever are in OOTP, something like that).