Gambo's Questionable Park Factors

[terpsfan101]

Gambo obviously put a lot of hard work into his park.dat file. There are 500+ ballparks to choose from. But there are some serious issues with his park factors. Sometimes they don't match reality at all. Take the Polo Grounds for instance. Using Retrosheet, I calculated Park Factors (correcting for the offset of other parks in the league) for the Polo Grounds using data from 1920-1939, 1945-1948, and 1952-1957, with no Yankees data (1920-1922) included. Perhaps too big of a sample size, but the dimensions of the Polo Grounds didn't change at all from the early 1920's until it the Giants left after 1957. Here are the factors:

Batting Average: .96
Home Runs/PA: 1.55
2B/PA: .69
3B/PA: .86

Gambo's park factors are the exact opposite in almost every respect:

Batting Average LHB: 1.167
Batting Average RHB: 1.116
Home Runs LHB: .844
Home Runs RHB: .768
2B: 1.143
3B: 1.096

So in real life, the Polo Grounds reduced batting averages 4%, decreased doubles and triples by 25%, and increased home runs by over 50%. According to Gambo, the Polo Grounds increased batting averages at 13-14%, increased doubles and triples around 12-13%, and decreased homeruns roughly 20%.

His park factors for Fenway Park II are also terrible. Everyone knows Fenway Park inflates batting averages, especially for left handed batters. But Gambo has Fenway Park II reducing batting averages, showing a rediculous .954 park factor for left handed batting average.

Gambo, you might want to look into fixing these. I know that you don't have empirical data for most seasons, but you at least need to tweak something so your park factors at least agree with common sense.

[Athletics17]

Maybe you should pm him and help him out a bit without posting that his work is terrible. I understand that maybe some of these are off, but be cool about it. The man has put countless hour into the game for our enjoyment and he isn't perfect. I don't think this was the best way to bring it to his attention.

[terpsfan101]

And how was my last post offensive? I was just pointed out that his park factors are way off in some instances. I may be guilty of cherry-picking, showing only evidence for the Polo Grounds in my example. I'm sure he can handle the criticism. It's not like I didn't take the time to carefully research the issue. present any evidence. I've released a lot of work for free on the web (park factors, linear weights, a pro basketball databse with complete stats from all the major pro leagues, going back to the 1938 NBL). I love it when people criticize my work. It shows they actually took an interest in what I did.

[Athletics17]

I am just saying it from how I read it. Constructive criticism is great no doubt about it. You presented evidence and that's great, I am not saying you are wrong with what you are suggesting at all.When you say Fenway park II is "terrible" or or the work at least needs to agree with common sense. I felt it was kind of rude how you came across and could have been said better. That's it.

[terpsfan101]

Well, the spreadsheet Gambo uses to calculate his park factors wasn't created by him. I tried to tweak some of the formulas, but I really don't know what the heck I'm doing. To get the homerun PF higher for the Polo Grounds, I weighted the dimensions down the lines more heavily. To lower the doubles PF, I substituted the triples PF. For the record, I am all for combining doubles and triples into one park factor, especially when working with real park data. Anyway, I like the idea of calculating a theoretical park factor from dimensions, altitude, etc... I am just not a fan of the spreadsheet Gambo uses to accomplish this.

In fact, the ideal way to calculate park factors is to calculate a theoretical park factor first using dimensions, fence heights, altiude, weather, and then regress the actual park factor to the theoretical park factor. Obviously this is a lot of work that I don't feel like doing, so I just regress all my park factors to 1. For instance, I should probably regress my park factors for Coors Field to something like 1.15, Petco to approximately .85. In it's brief existence so far, Petco has depressed run-scoring more than any other park in baseball history. Even more than Braves Field of the late 1930's and Dodger Stadium in the mid to late 1960's.

[Dr. P.R. Park III]

So you think you know a thing or two about calculating and adjusting park factors, so what? You can punch numbers into a spreadsheet, big deal. Hey Big Man, use your magik to come up with an updated and superior Historical Ballparks database, otherwise your "constructive criticism" is worthless. Occasionally socially inept people wander into the forums, I just hate when I stumble into their ignorant posts.

[Athletics17]

http://www.ootpdevelopments.com/boar...-creators.html

Check out this thread. Ry1220 is creating some of the old stadiums. Polo grounds, Baker field, ect. I don't think anyone has done park factors for them. Might be a good project.

[ootpFox07]

Have you guys seen this page on ESPN? Park Factors for the last decade. Trying to find a listing of them with l/r/splits.

MLB Park Factors - Major League Baseball - ESPN

[Gambo]

I'm not mad about the criticism as it may be justified but there are a few things that need to be straightened out.

1) First, I didn't create the park factors, I just implemented them into my database with some minor tweaks. I originally was skeptical of them and then when I started crunching the numbers was pleased with what I saw.

2) Park Factors - a statistical analysis of stats produced in a ballpark compared to the league average of that season.
Gambo Park Ratings - a statistical analysis of what stats should be produced by a parks dimensions and location based purely on the average of all historical major league parks. (1.0 generated as the average of all current parks)

"Park Factors" are not the end all and be all of rating a ballpark. The problem with park factors is two fold. 1st, it is based on the players who played there's stats and not the field at all. People assume that over a long period this corrects itself but 1/2 of the stats used are still based on the same set of players. 2nd, park factors are rated on each season. My park ratings are based on modern ballparks being 1.0. Older parks were generally larger with odd dimensions so compared to modern parks you would expect higher numbers.
Take your Polo Grounds example. During it's existence ballparks like Braves Field, Old Yankee Stadium, Navin Stadium existed with HUGE playing areas. Overall, compared to those parks (it's short corners and deep center combined with the playing style ) resulted in less hitting compared to the average park. Park Factors reflect this. Now put the Polo Grounds in a modern environment. Those deep gaps and short porches would result in modern hitters and speedsters having a field day. If you want to see the truth load up any "arcade" game with the Polo Grounds in it and play a few games with modern players.

3) I'll reiterate. The ratings are based purely on dimensions. I would not be naive enough to say the ratings are the best they could be BUT, you need to understand the difference between official "park ratings" and ballpark ratings in my DB. What would be better to do to analyze my numbers is just take the parks in existence those years, adjust my ratings to 1.0 for their average and see where the Polo Grounds lay in that comparison. It will be much closer than you think.

4) As for the Fenway example. In reality Fenway allowed for certain left handed hitters to have crazy averages. Wade Boggs made a career of tailoring his swing to bounce the ball of the monster, it was his game plan and he admitted to it. Many other lefty hitters who could did this did also. Fenway is a freak.

5) If your going to do a historical replay with players on their actual teams then I would recommend using park factors and neutralized statistics if your going to a fantasy league and use a ballpark (or a bunch from different eras) use my ratings. You'll get what your looking for.

[Gambo]

Just another little expansion on what I mean is the flaw to park factors.
This is just a little story about a fake ballpark and it's run in history. Comparisons to any real ballparks are purely coincidental (or are they?)

In 1914 Gambo Park is built and is fairly standard for it's time. During that season the games in that ball park produce fairly normal stats compared to others. The wind in this park is not a problem as hitters dont hit many homeruns. (In 1914 this parks factor averages 1.0 for hitters)

During the 1960's, domes are built that are have deeper and higher fences, larger gaps and artificial turf. As players hit more homeruns the wind in Gambo's parks cut down many of them. It's smaller size means less extra base hits. People start considering it a pitchers park (In 1965 this parks factor averages .8 for hitters)

Later all those speed domes start going away in favor of even larger parks. The tiny Gambo field is seen as a small place that larger more powerful steroid injected freaks start tearing apart. Wind can't even stop these blasts. Hitters start to love hitting there. (In 2000 this parks factor is a 1.2)

That's 3 different ratings though the park never changed. Compared over it's history the park would be a 1.0 I suppose. In my database that park will probably be rated a 1.0 or close. That's what I hope to achieve with my DB. It's for now,I feel, the best way to take away the people and time factor away from a park and give it the most appropriate ratings for use in OOTP.

[terpsfan101]

Thanks for sending the private message Gambo. I realized after I posted that the formulas your spreadsheet uses to calculate Park Factors, weren't yours. You just tweaked them to fit your averages. But I think the formulas could be improved upon. How? I don't know.

Gambo, I share many of your same concerns with park factors. I didn't realize you were comparing each park, to all parks throughout history, not each park to it's contemporaries. Thanks for clearing this up.

Dr P.R. Park III, why are you so quick to verbally attack me? How am I being igonorant and socially inept? And calculating good park factors requires more than just punching numbers in a spreadsheet. I have spent way too many hours refining these:

http://spreadsheets.google.com/ccc?k...HS09wdFE&hl=en

[terpsfan101]

Does the game use park factors or park adjustments? Where park factors are the figures for the home park only, and park adjustments are figrues for home and road games (PF + 1) / 2.

[Sausage]

Gambo, I just stumbled upon your ballparks spreadsheet, because I am interested in forming a fantasy league with friends using all-time greats from the Majors, Negro Leagues, Japan, Cuba, etc (especially after reviewing this spreadsheet). One thing I noticed is that there is potential to have a "negative" factor, depending on the dimensions. For instance, 59th Street Sandlot in Manhattan has a triples factor of -.474. To be mathematically correct, I believe you should use the reciprocal for factors below 1.0. After all, since we are talking about modifying probabilities, factors greater than 1 should approach infinity and factors less than 1 should approach zero. To accomplish this:

For any factor less than 1, let x = 2 - (factor). Then, take 1 and divide it by x. In this instance:

2 - (-.474) = 2.474. Then 1 / 2.474 = 0.404

I believe I like this better, since this puts Yankee Stadium at .664 for LHR instead of .495, which I think is a bit more realistic. Ruth and Gehrig weren't at THAT much of a disadvantage.

Terps Fan, one idea I have thought about is that if you wanted, you can combine the actual park factors with the theoretical park factors. I estimate the theoretical should be weighted twice. The reason I say this is that Gambo's spreadsheet takes into account the fact that players affect ballparks as much as vice versa. Since it also takes into account actual physical dimensions without being skewed by comparing stadiums from the same era (see Gambo's post above), this counts as two for Gambo's park factors. However, since expected values don't always match actual values, the actual park factors can be added to the formula. For Polo Grounds, this would yield:

LBA = 1.098
RBA = 1.064
LHR = 1.093
RHR = 1.058
2B = 0.992
3B = 1.017

This makes more sense to me since power hitters would be aiming for those short porches down the lines. Let me know if any of you disagree.

Speaking of the league I'm attempting to start, I'm most intrigued by OOTP 11, since it includes all seasons since 1871. I'm curious as to how neutralized/normalized stats work. I used to play Old Time Baseball back in the day, but the "Time Machine" would yield some crazy results. Babe Ruth with 100 HR's and Satchel Paige striking out two per innings amongst other all time greats seemed a bit of a stretch for me. Is OOTP 11 the way to go as far as this goes? What about other sims such as DMB, SBS, Dombrov, etc?

In any event, Gambo, you and the original creator are to be commended for such great work on this spreadsheet. How many hours of sleep did you lose concocting this? This spreadsheet singlehandedly stirred up conversations amongst friends and has prompted us to rekindle our old fantasy league (which was done on Mirco League WAY back in the day, our Old Time Baseball league didn't pan out due to differences of opinion with the Time Machine).

[Gambo]

It took a long time to compile the data but research is my thing. It took long enough that I havent even thought of touching it in two years.

Maybe I''ll get back to it and do some updates but I doubt it will be anytime soon. I still use it though!

[Charlie Hough]

Maybe this would help. The park factors do not equate to actual percentage adjustments in OOTP. That is not how they are used. I don't profess to know exactly how the formulae work, but I know that a HR factor of 1.200 does NOT mean that there will be 20% more home runs hit in that ballpark. And likewise a batting average factor of .900 does not mean that batters will hit for a 10% lower average.

This may be the source of the confusion and protest. So I want to caution anyone NOT to try and create new formulae or new park factors based on percentage adjustments. You will probably not get the desired results because OOTP does not use these factors that way. So whoever developed the original spreadsheet that Gambo used had probably developed the park factors according to how OOTP implements them and not the way people are thinking in this thread.

I would recommend some discussion with Markus to clear this up.

[Sausage]

If that's true Charlie Hough, then OOTP's park factors are different than the ones Pete Palmer/John Thorn/ESPN use. We'll have to reach out to Markus and get his input. Thanks for the replies everyone.

[Charlie Hough]

Here is the relevant passage from the OOTP manual. In principle the park factors are designed to increase or decrease output compared to the "neutral park" factors of 1.000, but it's not clear exactly what the neutral settings are based on. The critical thing I want to point out is that the manual explains that the adjustments are not based on straight percentages. Hence my previous post.

Even so, there may still be a problem with the ballpark factors that have been around the OOTP community for a while. It is hard to square the factors calculated based on OOTP and ballpark dimensions with those that people are finding elsewhere. That's because the others are based on the sabermetric calculation of home offensive statistics vs. road offensive statistics rather than the actual dimensions of the park or what OOTP uses.

It's important to note that OOTP's default historical park factors don't match the ones from sabermetric sources either. So either way you're not going to find a good match. I think it might be the case that the default historical park factors in OOTP are based on a park calculation vs. league totals for that given season. I see some evidence of this when looking at specific seasons and how certain teams fared in offensive categories against other teams in the same league. Always remember that OOTP is based fundamentally on league totals. But, unfortunately, since the modifiers are not based on straight percentages, I can't figure out a formula that OOTP uses.

Ballpark factors are based on a "norm" of 1.000. That is, a ballpark with all 1.000 factors is essentially a "neutral" park where hitters will all perform similarly. As the numbers increase, that factor becomes more common. So, for example, if your AVG Overall factor is 1.100, you can expect that if you had identical players in this park and a neutral park, the player in the park with the 1.100 AVG Overall factor would have a slightly higher average. The modifiers are not straight percentages. So, a 2.000 doesn't mean you will do "twice as well."

Also, for those interested, I believe this is the original source material that Gambo later built on for ballpark factors.

[Sausage]

What's the best way to get a hold of Markus for his input? I am very intrigued how these work if it's not percentage-based. Maybe it's percentage-based on certain events (ie home runs for long fly balls)

[ortforshort]

I'm in the process of researching a book on what went into the success and failure of baseball franchises over the course of history. One of the aspects I'm looking at is whether owners, GM's and managers put teams on the field that worked with or against the advantages of what their ballparks offered.

It's a bit of a different perspective on ballparks than what you guys are looking to get out of the information.

There are obvious ones such as lefthanded hitters and pitchers in Yankee Stadium and, conversely, righthanded hitters and pitchers in Fenway. Speed guys in Busch Stadium and Comiskey Park, etc.

I haven't started looking at the park.dat file, but are there some tips you guys can give me when looking at the data or some insights that you've come up with for the types of players that management should be trying to procure for each park. Hitting, pitching, baserunning and fielding.

For example, what do you do in Coors Field when looking for a pitcher? If you're a fly ball pitcher, you're dead. However, the field is baked so dry and hard that if you're a ground ball pitcher, you're dead. What should you try to be getting and how well has management succeeded in getting what they need? And, if they do tailor a team for their ballpark, will that kill them on the road?

Also, I've never understood this obsession with trying to factor out ballpark effects when evaluating player's performance. For example, as a Yankee fan, I commend Cashman for obtaining Curtis Granderson (one of the only things I commend Cashman for). Granderson hit 41 home runs last year. He's made for Yankee Stadium and he produced. Why would you want to normalize his hitting by saying that if he was in a "normal" park, he would have only hit, say, 30 home runs? That's who he is and that's what he produced.

[Gambo]

With alot of secondary statistics the popularity kind of ebbs and flows. Park factors i think is one that previously peaked and has now diminished in importance. I think after all the number crunching the results didn't justify the effort put into it.

Now dont get me wrong, in certain cases it does come into play. Colorado is an example of this. But to what extent? In it's early history playing in Colorado didn't rejuvenate the careers of many of the litany of hitters they brought in there. (Dale Murphy?)

Increasing HRs by 10% is 4-5 a year for the best power hitters, but then again thats only over 81 games. What difference did it really make?

But so many other things come into play when looking at park factors on a season by season basis. What was the coaching style? Where they ahead or behind in a lot of games? Where was that batter in the lineup?

I agree that park factors are overrated when looking at specific players in specific seasons but when looking at the ballpark overall it does make a difference.