Home field advantage: for OOTP to match MLB...

[gmo]

I know there is talk of the HFA in other places, but I want to take a statistical tack here. This is a looong post, but hopefully it will be meaningful and shed some light not directly on how the OOTP HFA black box works, which I'm not concerned about provided it is reasonable, but on the results it should (or at least may) produce.

There has been plenty of speculation as to why HFA exists in baseball - getting to bat last, less fatigued players, players more comfortable in & knowing better their home stadiums, teams built to fit their stadiums, etc. I don't care here directly about why HFA exists. I only care here about why it exists as much as it applies to getting OOTP to simulate the observed data from MLB, though I may comment a bit.

What does MLB data show? Just a reminder to many of you, as it is oft-cited by Le Grande Orange in these forums, these results come from 1901-2003 (or something very close to that range of years):
*Home teams have an overall winning percentage of ~.540, meaning 54% of games are won by home teams
*~83% of the time individual teams at the end of a full season have a greater winning percentage in their home games than they have in their away games

I did a substantial amount of testing with OOTP5. Initially I saw what looked like a statistically significant, though very tiny, advantage for away teams. Further analysis did not support this idea, so I had to conclude that initial result was simply because I got by luck a very unusual sample. The full collection of data I accumulated for that test league seemed to point to there being in the long run no statistical advantage for either home or away team - neglecting outside factors, in an infinite number of games home and away teams would each win half the games. Though they were highly variable on all timescales, extra inning games were split virtually exactly 50-50 also. That points to no structural advantage to batting last. Whether there is a psychological advantage that OOTP can not simulate could be debated. But remember, batting last also makes a team more likely to be behind when they come up in the bottom of an inning, and I think LGO posted something once saying at one time home teams batted first because there was a perceived advantage in jumping ahead not last at-bat.

In MLB, home teams win more than away teams, presumably because they play better, which should show up in the stats. Others here, most notably I believe Skipaway, mentioned some stats illustrating this, so I went to espn.com and grabbed some of their team Home/Away split data from 2001-2003 and averaged it up. Here is what that data shows:

First off, the wins and losses. Home teams had a 3-year average winning percentage of .538 (from seasons of .524, .542, .550), much in line with the historical average. That translates to an average of 87.25-74.75 record per 162 games for home teams. Using runs scored and runs allowed, home teams get an average pythagorean record of 83-79 (82.98-79.02) per 162 games, meaning a pyth diff of about +4. It is logical home teams would score less in this way (less than what their win pct would lead you to believe) because they bat less overall; from the 9th inning on home teams bat only as much as they need to. Not though, the +.500 pyth rec means that despite coming to bat less, home teams score a greater number of absolute runs than away teams.

Now the non-W/L numbers. Batting average, OBP, slugging, and OPS are all about 3.5% higher for home teams (3.35%, 3.72%, 3.68%, 3.70%, respectively to be exact).

Per plate appearance, home teams
*walk more (by 7.5%),
*strike out less (by 6.0%),
*score more runs (by 6.5%).

Also per plate appearance, home teams hit more home runs (by 2.4%), more doubles (by 2.5%), and more triples (by 17.4%). But I think those HR, 2B, and 3B rates are misleading if you look at extra base hits divided by total number of hits. Here are the percentages of total hits-
*going for HR: 11.98% for home teams, 11.99% for away teams;
*going for 2B: 20.07% for home teams, 20.08% for away teams;
*going for 3B: 2.27% for home teams, 1.98% for away teams;
*going for 1B: 65.69% for home teams, 65.95% for away teams.
I was shocked at how close to exactly the same those numbers were in general, and specificly in this sample 2B & HR proportions being within 0.01%. The small difference with home teams getting more 3B may be from the small number of those events, but could also reflect something real like away team defenses playing more balls into triples, for example.

A few other stats show home teams sacrifice more (5.1% more sac hits, 8.7% more sac flies) and are not surprisingly intentionally walked more (11% more). Stolen base, GIDP, and HBP rates per plate apperance home/away differences are all within about ~1% or less of each other.

Pitching stats of course cover the same ground as the hitting stats. It can of course be debated how much of the statistical W/L differences between home and away teams are due to hitters and how much pitchers. But pitching stats may say a little something about fielding, for which espn.com does not have home/away splits. The percentage of total runs allowed that were unearned does not appear different for home and away teams. For the 3 seasons here are the percentages of runs allowed being earned -
for home teams: 91.43%, 91.80%, 92.14%,
for away teams: 91.46%, 91.44%, 91.98%.

In the average a very slightly greater percentage of away teams runs allowed are unearned, but I would not make much of that. My first thought was that since the same percentage of runs are unearned for home and away teams, the error rates would be the same, and though I think this makes sense I'm not completely sure on it. Plus, maybe there is something to official scorers giving hits more often to home batters instead of errors to the away team. So I don't know what to make of that overall. My gut inclination would be to say away teams field slightly worsely than home teams, but my head says if so it is only ever so slightly and stats may not be able to show it. [Edit: well, look down a few posts]

In summary, it is not from the whole history of baseball, but 2001-2003 MLB data shows clear home/away differences in batting average, BB, SO, and thus home teams score at a better rate per PA. Home teams get more extra base hits, but only because they get more hits period as the proportions of 1B/2B/3B/HR remain the same. If teams are slightly better fielders in their home parks than when away, this is not obvious from the limited stats I have seen. [Edit: again, look down a few posts] So if OOTP can by some combination of "effects" on hitters and on pitchers make home/away stats (Avg, BB, SO are I think the keys) work out with the same differences as observed, I believe the same home team win pct will come out of it.

[gcsii99]

^^^^ wow...great post...good info!

[pollaro70]

Very impressive. So, is this what the home field advantage upgrade takes into consideration?

[Eugene Church]

Good post, gmo.

Much appreciated.

[Dargone]

Very good info...

[gmo]

Turns out maybe there is a meaningful difference in defense shown in the pitching data.
Skipaway pointed it out before, but I just failed to look at it above.

Below is the MLB BABIP data from 2001-2003 for home and away pitchers.

BABIP is approximated here as:
(Hits - HR) / (Innings*3 - GDP - K + Hits - HR).

BABIP   Home   Away   Home less by
2001   .2880  .2934    1.81%
2002   .2854  .2904    1.75%
2003   .2856  .2937    2.76%
Total  .2863  .2925    2.16%

Interesting how in each column a number from a different year stands out from the rest. Anyway, as best as I can estimate the home/away BABIP difference adds up to 1 extra hit on a ball in play for home teams every 6 games. Maybe it involves familiarity with outfield fences, the speed of the infield grass, artificial turf, dome ceilings, foul territory, or just being more comfortable or less tired at home. Sound reasonable that on average once sometime every six games the home defense makes one play for an out rather than allowing it to result in a hit because they can operate better in the home park for some combination of those reasons? I think it needs to since in this age of DIPS were not going to be attributing that BABIP difference to the pitchers. Well, we could put it on the hitters I suppose, but I like my defense hypothesis.

[gmo]

So how well is OOTP6.01 doing with its home field advantage option?

Below are the differences between home and away stats in 2001-2003 MLB data and a simple test league I made in OOTP6 that was played through a number of games equal to 10 MLB seasons, though the configuration was nothing like MLB. The stats that interested me most from the MLB data were Batting Average, BABIP, OBP, SLG, and walks, strikeouts, and runs per plate appearance. A positive number denotes the home team rate or value was higher than the away team's, a negative number vice versa.

Stat    MLB    OOTP6
Avg     3.4%   1.8%
BABIP   2.2%   1.7%
OBP     3.7%   1.3%
SLG     3.7%   2.0%
BB/PA   7.5%  -0.1%
SO/PA  -6.0%  -0.2%
R/PA    6.5%   2.2%

Then the most important number of all in looking at home field advantage - wins. The winning percentage for home teams in MLB year 2001-2003 is .538, and in my test league it was .514. In other cases I have looked at in measuring the HFA in 6.01 I have gotten percentages of .512, .510, .507. So, my experience shows home teams winning 51% in the game, not quite up to the historical level of 54%. But I think the game is closer to matching that value than those test percentages may imply.

I think all those stats above are intertwined. Hopefully the following logic is correct. Let's compare OOTP to MLB, which I will consider "correct" for the sake of this discussion...

First BABIP, which the game appears to at least almost be getting "right". Note the MLB total of ~2.2% is based on seasons of 1.81%, 1.76%, 2.76%, suggesting a precise MLB value may be difficult to pin down. Now onto BB and SO per plate appearance, and clearly OOTP is not simulating the stark home/away differences shown in MLB where home hitters walk more and strike out less. What if the game got those? More home team walks will obviously give the home team higher OBP. Taking the BABIP formula (what I used is shown in the above post) and adding in strikeouts nearly gives the formula for Avg. So adding more K to the away side than the home side will give an even larger percentage difference between home and away in Avg than in BABIP. The same argument should apply for SLG. Also, I noted above the percentages of hits of each type (1B/2B/3B/HR) were the same for home and away teams in MLB. Though not with the same precision as noted in the MLB data, I'd say I also found this in my OOTP test league. Those statistical changes I would think would get the home teams scoring more relative to the away, and that would result in their winning percentages coming more in line with MLB.

Must say kudos to Markus for this added feature to increase realism. It was a very nice surprise for me to see. But how long before the calls to turn the option button into a slider bar? I can hear it now - "The Death Leagues need the ability to have home field disadvantage because the players are more rattled at home by fans cheering for their demise"

[Malleus Dei]

The professional baseball players that I used to know explained the home field advanatge by saying that many of the home players were more rested than the away team players were because most of the home teamers had been sleeping in their own beds the night before and had not been out all night with wild women like all of the players on the away team had.

[Gastric ReFlux]

Thanks for studying this GMO, and I endorse your call for a slider bar so we can proper Death League implementation.

[Le Grande Orange]

GMO, that's some very nice detailed stuff you've done there.

However, let me ask you about the other stat: over the course of your sims, what is the number of teams finishing better at home as compared to those finishing better on the road? Historically, this figure is near 84% of teams finishing the season with a better home winning percentage than away winning percentage (that's 5 out of 6 teams). What are your leagues generating?

[potatoman]

Very good research done GMO...

[danwaechter]

Excellent research GMO, its much appreciated.

[gmo]

Originally posted by Le Grande Orange
However, let me ask you about the other stat: over the course of your sims, what is the number of teams finishing better at home as compared to those finishing better on the road? Historically, this figure is near 84% of teams finishing the season with a better home winning percentage than away winning percentage (that's 5 out of 6 teams). What are your leagues generating?

The league where teams only play either home or away is obviously not good for looking at this directly. But I can show the following...

       Winning  Losing   .500     Winning Season
Teams  Seasons  Seasons  Seasons  Percentage
------------------------------------------------
Home     164      114      22       .583
Away     105      176      19       .382

I have tested a 10 team league with a balanced schedule (9 games each home and away against every other team), simming 30 seasons. In those 300 team seasons, 178 times teams had more wins at home than away, 104 times teams had more wins away than at home, and 18 times teams had the same number of wins home and away. Compared to your 84%, this is about 62%. This league had a home team winning percentage in all the games of .5115 (12432-11871).

From a "theory" standpoint...

Perhaps I finally found some math to back up my assumption that if you get the overall home game winning pct up to .540, you'll get that 84% of teams winning more at home than away. For that 10 team league I just mentioned, for each team season I took home wins minus away wins - that is how I got the 178-104-18 numbers. That set of numbers ranges from +19 to -19, has a mean of +1.87, and a standard deviation of 6.14. Adjust the average team with no home field advantage setup of 40.5 wins home and 40.5 wins away, and you get 41.435 home wins and 39.565 away wins (that overall .5115 home win %). It seems reasonable to assume that home wins minus away wins is pretty close to normally (bell-curve) distributed about its mean value. In this case, .500 home winning pct is about 0.3 standard deviations (1.87/6.14) from the mean of .5115, which is the peak of the bell curve. Half the bell curve area plus 0.3 standard deviations, the amount between the peak and .500 here, is 50% plus about 12% equalling 62%, which is what I found for the value of home teams winning more at home that away.

To cut to the chase, assuming a bell curve centered at 6 to 6.5 more home wins than away wins (what you get when home teams win ~54% of the games) and that same standard deviation of about 6 to 6.5 in the data noted above, you get the following. A zero value season, one with equal home and away wins, is then right about 1 standard deviation from the mean, so negative value seasons, more away than home wins, make up the area under the bell curve on one side of the peak outside one standard deviation. That area is ~16% of the total area, meaning 16% of teams will be better away than home, thus 84% win be better home than away (neglecting the equal cases).

The point is now I am quite certain getting the 54% of games won by home teams will force the other hand and give you the 84% of teams winning more at home than away. I tried to be clear in the explanation, but it probably just ended up confusing people regardless of their statistics background. If I lost you there, it is much more straightforward than how I described it.

[Goodsox]

This post deserves to be back at the top. Great research and analysis!!

[KaiFuller]

here's my added thoughts, what do you guys think?

FACTS: biggest differance comes with walks and strikeouts. then it's hits, then its defense

Is it possible that:
A: The umpire favors the home team, causeing a BB\K ratio favorable to the home team? This would also end up making the away team have to swing at more difficult pitches, causing their hits numbers to fall a bit?

B: If the home team is on base more, the away team is forced to hold the runner at first and play double play depth, causing that extra ball to get through every six games

thoughts?

[tward13]

Some of those walks will be explained by bottom of the ninth strategy that almost only happens to the home team.

Earlier in a game and in the top of the ninth teams don't intentionally (or semi-intentionally) walk as many batters, not named Bonds) to set up a force as they do in the bottom of the ninth. Teams will sometimes walk a couple of batters to set-up a force. It's not a huge number, but it's a situation that is more unique to the home team.

I think any results that show something odd for the home team needs to be reviewed with the bottom of the 9th in mind. It won't answer all of them, but might help explain some results.

[ctorg]

This is some great research here, gmo. Thanks for the insight. I've wondered about this myself but have never had the time to do any of the leg work.

[gom]

I, for one, STRONGLY disagree with inputting a home-field advantage in OOTP.

I believe, and I do not have substantiated data to back up my theory, that home-field advantage in OOTP is compensated with the ball park factors. Example: Using Ankit's current ballpark factors, Yankee Stadium has a homerun rating of 117 for lefties and 98 for righties.

Any team manager/owner would try to stock up on lefty power hitters and lefty pitchers. Therein lies the advantage for home teams, as they tend to cater to their ball parks. This should allow for the disparity between home and away winning percentages.

I am not one of those people who believes my way of playing is better than other people's, but I am worried that the over-analyzation and customization can ruin the product per se, much akin to too many cooks spoiling a broth.

When I was in college, I tutored other students in some subjects, and I always told my subjects we could prepare for exams in two ways. We could concentrate on the 95% of the material and get that down pat, or we could concentrate on 5% of the material and make sure we get all of the "trick" questions. I feel like many people are concentrating on the 5% here, rather than the overall game play constituting 95% of the game. I would rather the programmers concentrate on the bigger picture rather than the minor details that would, IMHO, actually


Markus, I hope you take my previous request for CSVs in an upcoming patch: http://www.ootpdevelopments.com/boar...threadid=60729

[Gastric ReFlux]

Originally posted by gom
I, for one, STRONGLY disagree with inputting a home-field advantage in OOTP.

I believe, and I do not have substantiated data to back up my theory, that home-field advantage in OOTP is compensated with the ball park factors. Example: Using Ankit's current ballpark factors, Yankee Stadium has a homerun rating of 117 for lefties and 98 for righties.

Any team manager/owner would try to stock up on lefty power hitters and lefty pitchers. Therein lies the advantage for home teams, as they tend to cater to their ball parks. This should allow for the disparity between home and away winning percentages.

I am not one of those people who believes my way of playing is better than other people's, but I am worried that the over-analyzation and customization can ruin the product per se, much akin to too many cooks spoiling a broth.

When I was in college, I tutored other students in some subjects, and I always told my subjects we could prepare for exams in two ways. We could concentrate on the 95% of the material and get that down pat, or we could concentrate on 5% of the material and make sure we get all of the "trick" questions. I feel like many people are concentrating on the 5% here, rather than the overall game play constituting 95% of the game. I would rather the programmers concentrate on the bigger picture rather than the minor details that would, IMHO, actually


Markus, I hope you take my previous request for CSVs in an upcoming patch: http://www.ootpdevelopments.com/boar...threadid=60729

Hello? Are you the Anti-Gmo?

[Henry]

The HFA was added as an option as is usually done when possible.

Henry