Examining Baseball’s Most Extreme Environment
“The Coors Effect.”
These three words evoke a strong reaction from most people and are impossible to ignore when discussing the offensive production of a Rockies player. Ask anyone who was around for the Rockies of the ‘90s and they will tell horror stories of games with final scores of 16-14. Ask anyone at FanGraphs and they will laugh and point at the Rockies’ 2015 Park Factor of 118. Heck, ask Dan Haren and see what he has to say:
I would count out the days about a month in advance to see if I was gonna pitch in Coors field.
— dan haren (@ithrow88) January 4, 2016
Suffice it to say that Coors is a hitter’s park. Nobody will argue that. But there have been murmurs recently about another effect of playing 81 games at altitude, an effect that actually decreases offensive production. These murmurs have evolved into a full-blown theory, which has been labeled the “Coors Hangover.”
This theory supposes that a hitter gets used to seeing pitches move (or, more accurately, not move) a certain way while in Denver. When they go on the road, the pitches suddenly have drastically different movement, making it difficult to adjust and find success at lower elevations. In other words, Coors not only boosts offensive numbers at home, it actively suppresses offensive numbers on the road, which can take relatively large home/road splits for Rockies players and make them absolutely obscene.
The concept seems believable, but thus far we have no conclusive evidence of its merit. FanGraphs’ Jeff Sullivan recently tested this theory, as did Matt Gross from Purple Row. Although neither article revealed anything promising, Jeff is still a believer, as he recently shared his personal opinion that the Coors Hangover might simply last longer than any 10-day road trip. With this is mind, I decided to approach the problem by examining the park factors themselves.
If you haven’t read the article about how FanGraphs calculates its park factors, I highly recommend you do so before continuing. The basic approach detailed in that article is the same approach that I use here. As a quick example, the park factor for the Rockies is calculated by taking the number of runs scored in Rockies games at Coors (both by the Rockies and the opposing team) and comparing that to the number of runs scored in Rockies games away from Coors. Add in some regression and a few other tricks, and we have our final park factors.
This method makes a number of assumptions, most of which are perfectly reasonable, but I was interested in taking a closer look at one critical assumption. By combining the runs scored by the Rockies with the runs scored by their opponents, we are assuming that any park effect is having an equal (or at least, an indistinguishable) impact on both teams. This seems like an obvious assumption, but it becomes invalid when the Rockies play on the road. According to the Coors Hangover, Rockies hitters experience a lingering negative park effect after leaving Coors which the opposing team is not experiencing.
In other words, we expect a gap to exist between a hitter’s performance at Coors and his performance at an average park. If the Coors Hangover is true, this gap would be larger for Rockies hitters than anyone else.
Let’s start by taking a look at the park factors we have now. The following tables only contain data from NL teams for simplicity sake.
| Park Factors, 5-year Regressed (2011-2015) | |||
| Team | Total Runs (team + opponent) | Park Factor | |
| Home | Away | ||
| Rockies | 4572 | 3205 | 1.18 |
| D-backs | 3657 | 3328 | 1.04 |
| Brewers | 3588 | 3306 | 1.04 |
| Reds | 3385 | 3215 | 1.02 |
| Phillies | 3365 | 3341 | 1.00 |
| Nationals | 3240 | 3213 | 1.00 |
| Cubs | 3346 | 3345 | 1.00 |
| Marlins | 3200 | 3229 | 1.00 |
| Braves | 3086 | 3199 | 0.99 |
| Cardinals | 3243 | 3397 | 0.98 |
| Pirates | 3070 | 3394 | 0.96 |
| Dodgers | 2995 | 3323 | 0.96 |
| Mets | 3109 | 3556 | 0.95 |
| Padres | 2936 | 3440 | 0.94 |
| Giants | 2900 | 3537 | 0.92 |
No surprises. Teams score a ton of runs at Coors and hardly ever score at AT&T Park in San Francisco. Now let’s split up those middle columns to get a closer look at who is scoring these runs.
| Runs Scored, 2011-2015 | ||||
| Team | Home Stats | Away Stats | ||
| Team | Opponent | Team | Opponent | |
| Rockies | 2308 | 2264 | 1383 | 1822 |
| D-backs | 1844 | 1813 | 1641 | 1687 |
| Brewers | 1823 | 1765 | 1619 | 1687 |
| Reds | 1731 | 1654 | 1606 | 1609 |
| Phillies | 1676 | 1689 | 1576 | 1765 |
| Nationals | 1749 | 1491 | 1651 | 1562 |
| Cubs | 1625 | 1721 | 1547 | 1798 |
| Marlins | 1541 | 1659 | 1464 | 1765 |
| Braves | 1606 | 1480 | 1569 | 1630 |
| Cardinals | 1779 | 1464 | 1797 | 1600 |
| Pirates | 1586 | 1484 | 1688 | 1706 |
| Padres | 1443 | 1493 | 1604 | 1836 |
| Dodgers | 1557 | 1438 | 1758 | 1565 |
| Giants | 1481 | 1419 | 1797 | 1740 |
| Mets | 1482 | 1627 | 1817 | 1739 |
These are the two pieces of run differential — runs scored and runs allowed — and we generally see agreement between the home and away stats. If a team out-scores their opponents at home, they can be expected to do the same on the road. Good teams are better than bad teams, regardless of where they play. Although, if you subtract a team’s run differential on the road from their run differential at home, the difference will actually be around 100 runs due to home-field advantage. Doing this for all 30 teams yields a mean difference of 83 runs with a standard deviation of 122.
Where do the Rockies fall in this data set? Not only have they scored over 400 more runs at home than the next-best NL team — they have also scored almost 200 runs less on the road than the next-worst NL team. Comparing their home and road run differentials, we see a difference of 483 runs (+44 at home, -439 on the road), or 3.3 standard deviations above the mean. To put it plainly: that’s massive. This is a discrepancy in run differentials that cannot be explained by simple home-field advantage.
Furthermore, I followed the same process of calculating park factors for each team explained above, but I split up the data to calculate a park factor once using the runs scored by each team (tPF), and again using the runs scored by each team’s opponents (oPF). Generally, these new park factors are closely aligned with the park factors from before…except for, of course, the Rockies.
| Alternate Park Factors, 5-year Regressed (2011-2015) | ||
| Team | tPF (Team Park Factor) | oPF (Opponent Park Factor) |
| Rockies | 1.27 | 1.10 |
| D-backs | 1.05 | 1.03 |
| Brewers | 1.05 | 1.02 |
| Reds | 1.03 | 1.01 |
| Phillies | 1.03 | 0.98 |
| Nationals | 1.02 | 0.98 |
| Cubs | 1.02 | 0.98 |
| Marlins | 1.02 | 0.97 |
| Braves | 1.01 | 0.96 |
| Cardinals | 1.00 | 0.96 |
| Pirates | 0.97 | 0.94 |
| Padres | 0.96 | 0.92 |
| Dodgers | 0.95 | 0.97 |
| Giants | 0.93 | 0.92 |
| Mets | 0.92 | 0.97 |
On average, a team’s tPF is about two points higher than its oPF — again, this can be attributed to home-field advantage. The Rockies, however, are in an entirely different zip code with a discrepancy of 17 points. We aren’t talking about home-field advantage anymore. We are talking about something deeper, something that should make us stop and think before averaging the two values to get a park factor that we apply to the most important offensive statistics.
We have no reason to believe that any team should have a 17-point difference between their tPF and oPF; the fact that the Rockies are in this situation either means that they are enjoying hidden advantages at home, or they are suffering hidden disadvantages on the road. To date, we don’t have a theory supporting the former, but we do have one supporting the latter. This is the Coors Hangover.
Does this mean that the Rockies’ Park Factor should actually be their oPF of 110? Should it be some weighted average of different values? I don’t know. But I do know these numbers can’t be ignored. Something is going on here, and we need to talk about it.
Jacob is a mechanical engineer who spends an unhealthy amount of his free time researching baseball.
Looks like their pitchers have it too, except they go from awful at home to bad away.
That’s completely true, but I think that really shows the regular Coors Effect. Their pitchers give up 400 more runs at home than on the road (from awful to bad, like you said). That, on its own, is worthy of a 110 park factor, which is still extreme by all means.
Based on this, we expect their hitters to score somewhere around 500 more runs at home than on the road. They are actually scoring 1000 more at home though, suggesting that something else is going on.
“Something else” is simply the unique home field advantage at Coors — well-known to baseball analysts for more than 20 years.
“This is a discrepancy in run differentials that cannot be explained by simple home-field advantage.”
Yes, it can. There is no law that says HFA is constant among teams. Denver teams always have the highest home advantages in their respective leagues — and the Rockies are no exception. Their HFA is generally about 3.5 STDevs above average.
BTW, to do park factors correctly, you need to look at how many half-innings were actually played, which account for winning home teams not batting in the 9th and extra-inning games.
To your second point: That is correct. Also, to do them “correctly” you can’t assume a team has an away park factor of 100. We ignore these relatively minor differences because they are essentially insignificant in the large scale, especially when considering the time it takes to implement them.
To your first point: I agree there is no such law. I took this as more of a red flag that led me to explore the concepts of tPF and oPF in the rest of the article, which is where I really made my conclusion. The 17-point difference for the Rockies is almost 5 standard deviations above the mean.
You bring up an interesting concept though-seeing how these results compare across different sports. My initial thought was to compare Runs Scored to Runs Allowed because pitchers wouldn’t be affected by the Coors Hangover, but using another sport entirely should do the same trick. Worth looking into at least.
I just ran a quick look in the nfl and was having a hard time finding easily accessible splits. I did find an article that looked at home field advantage using win% between 2002-2011, and the Broncos came in with the 20th ranked home field advantage.
I really would like to look at point differential, but I have a hard time believing a team with a below average home field advantage using win% would jump up to 3.5 standard deviations above the mean using a different method.
Home/away schedules are way too lop-sized in the NFL to use a raw home vs. away win pcts. — not to mention the incredibly small sample. Where did I argue that the broncos’ HFA is exactly 3.5 stdevs above the mean, btw? I said that the Rockies’ HFA historically has averaged about 3.5 stdevs above average (this is not difficult to calculate) — which of course is not the same as saying that the observed HFA over every 4-5 month period will be 3.5 stdevs above average, esp. if you are not taking the extra 2 minutes to adjust for actual innings played. See: variance.
I’ve been making a living betting sports for 10+ years, so I prob. know a thing or two about HFA. But by all means, if you think there is something magical going on at Coors beyond the obvious (human beings having difficulty adjusting to extreme altitude), run with it..
Did you read the series about 2016 park factors calculated from Statcast data? Tony calculated these using only batted ball data, meaning any effect that the Rockies experience on the road would not show up anywhere in the data.
His park factor for Coors is 124. If you cut it in half and regress it like I discuss in the article, this gives the Rockies a final park factor of 107. A more conservative regression constant gives a park factor of 109. Keep in mind this is a completely independent author taking a look at totally different data from a different angle.
Ohhhhhh. Shoot. You’ve spent a decade betting on sports. I didn’t know that. I guess I can ignore all this evidence then.
Evidence of what exactly? You’ve produced nothing that cannot be explained by an extreme home field advantage plus natural variance. You seem unable to grasp the concept that different environments will have different HFAs — in some cases, very different.
And if you (and/or Blengino) don’t understand why single-season park factors are not predictive, I really cannot help you.
Here’s an independent source that tells you the Coors PF is the same as it has been over the past few seasons:
http://www.baseball-reference.com/teams/COL/2016.shtml
Let’s back up – I think there is some confusion. Looking at winning% or run differential will include both home field advantage and road field disadvantage. Pointing to either of these does nothing to prove the existence or strength of either.
There is also no argument that the Coors PF is different for this season. My argument is that it has stayed consistently lower than what we have been calculating. My 5-year evidence and the Statcast 1-year evidence both support that. (Note I’m not using the word prove. This is merely support, pointing in the same direction)
Brief evidence summary-
1) All other teams experience a PF of 110 when coming to Coors. The Rockies experience a PF of 127. This discrepancy is 5 standard deviations above average.
2) The Rockies have sustained high K rates, middling base running, and poor defense, which would all presumably be ways to boost home field advantage using offense at Coors.
3) The Rockies have been completely lost in terms of a pitching philosophy, suggesting they have yet to find any ways to boost their home field advantage using pitching.
4) Statcast data from 2016 suggests the Rockies PF should be under 110.
Separating home field advantage from road field disadvantage is extremely difficult. These 4 pieces of evidence combined provide support that the Rockies have significant road field disadvantage in addition to an already extreme home field advantage.
The phenomenon of altitude creating an abnormally large HFA/HCA has already been well-documented:
http://www.espn.com/nba/insider/story/_/id/9014283/nba-analyzing-real-home-court-advantage-utah-jazz-denver-nuggets
http://predictionmachine.com/college-basketball-homecourt-advantage
I.e., it would be quite interesting if the Rockies for some reason did *not* have an extreme HFA; but as your data confirms, they do.
I think one issue with comparing HFA or HCA for NBA and NFL to baseball is they are much more constantly active sports. In Football and Basketball altitude comes into effect in a players stamina while in baseball it’s primary effect is on ball movement and carry. Realistically the HFA created by altitude is not comparable sport to sport in that respect. Field Goals in football is about the only comparable aspect, maybe deep pass distance as well.
This is absolutely true. When the 4th quarter rolls around in Broncos or nuggets games, the other team is completely spent. Baseball is not the same.
In fact, this might actually be bad for the Rockies in the long run. Baseball is a marathon. Playing 81 games at altitude has to be rough on the body. They might run out of gas during the “4th quarter” of the season, while their opponents are doing much better.
I’ve looked at this on occasion on my own time, and while giving the Rockies hitters a little more creidt doesn’t change single-season numbers a whole bunch, I’m thinking Todd Helton might be a little undersold on FG with his 55 WAR.
Not entirely sure how much, but considering he might get borderline HOF support with that number it might end up being crazy important for future voters. Tulowitzki, Gonzalez and Holliday probably see a little change too, while Rockies pitchers are technically worse than you think (Boo! Halloween comes early)!
I think part of the discrepancy can be explained by the fact that besides Arizona, all other Ballparks in the NL West are graveyards.
I took a quick look at their schedule this year and they had to play 44 games in ballpark that surpress run scoring by your defintion compared to just 12 which increase run scoring on the road. The rest was neutral.
To make it even more extreme, 34 of the 44 road games in ballparks that surpress run scoring came against the 4 most extreme run surpressing ballparks (10@SF, 10@SD, 4@NYM, 10@ LAD)
The way I understand it, that shouldn’t really change the discrepancy. The fact that the Rockies play in pitcher-friendly away stadiums means the gap between their runs scored at home vs on the road will be larger than it should, which means their “tPF” should actually be a bit smaller. But the same should theoretically happen to their pitchers, meaning the “oPF” should actually be a bit smaller too.
If both go down a few points, it doesn’t make the discrepancy any smaller, it just means that we’re overcompensating for Coors in a completely different way as well.
Both really good points!
Accounting for the road Park Factors in 2015 drops the Rockies oPF to about 109 and their tPF to about 124. So GG_huson is correct that both numbers should drop because pitching and offense are both affected, but AC_Butcha_AC is correct that this still causes the discrepancy to get slightly smaller, albeit indirectly.
I know you referred readers to the FG explanation of park factors, but it would have been helpful if you had explained in slightly more detail how you came up with yours. In the first table, if you divide the total runs scored by the Rockies and their opponents at Coors by the total runs scored in road games, you get 1.42. How does one get a park factor of 1.18 from this?
First, you have to cut the portion > 1 by 50%, because the 1.42 would apply if all games were played at Coors, whereas of course players play half their games on the road, where the park factor is assumed to average out to close to 1.0. But second, there is apparently a regression factor of about .84, at least that is what one needs to multiply the ratios in the first table by to get the park factors you actually come up with. And the regression factor for the second table is different, though apparently not by very much.
Now on to the actual analysis. You note the large discrepancy between the factors of 1.27 and 1.10 in the second table, where one considers just the runs scored by the Rockies at home or on the road. More precisely, 1.27 is the ratio of the runs scored by the Rockies at home vs. on the road, while 1.10 is the ratio of runs scored by their opponents at Coors vs. on the road. You argue that either the Rockies have some benefit from playing at Coors not available to their opponents, or that there is a hangover effect on the road, so that they play worse than other teams on the road (thus inflating their apparent home benefit).
Couldn’t both factors be in play? I take it that the Coors factor is partly because the low density air results in less movement on pitches, and partly because batted balls carry further. The second factor obviously will be the same for both home and visitors, but the first does not have to be. In the first place, Rockies pitchers may adjust to the different conditions by throwing fewer breaking balls, or different kinds of breaking balls. Second, whether they do or don’t, visiting players, not used to the different movement, may not be able to take as much advantage of it as the home players. It could be a matter of expecting more break, or even if they go in to Coors expecting less break, simply not able to adjust as well as the home players.
You are correct that you have to cut the portion >1 in half (this is where the assumption that all road games average to a Park Factor of 100, which is discussed in above comments).
The actual formula for the raw Park Factor that I pulled from the linked article is: H*T/((T-1)*R+H) where “H” is the Runs per Game @ Home, “R” is the Runs per Game on the road, and “T” is the # of teams in the league. The regression factor suggested in the article when using 4+ years of data is 0.9 (this regression factor does stay constant throughout my analysis). For the Rockies overall Park Factor, the first step yields a raw PF of 1.407, which gets cut in half to 1.203, then regressed to 1.183.
I really like the points you bring up in your last paragraph. My personal belief is that if the Rockies had zero home field advantage, their oPF of 110 should be taken as their true overall Park Factor. Obviously, this is NOT the case and changing their PF to 110 would be a dangerous overreaction. Everyone should have some home field advantage, and I think it’s fair to believe that the Rockies enjoy a slightly larger one than most teams.
But I don’t think that Rockies pitchers have it all figured out. At all. The Rockies have changed their pitching strategy a number of times in recent years. They went with sinkerball GB pitchers for a while (think Aaron Cook). They tried finesse guys with command (think Jeff Francis). They tried power pitching fireballers (think Ubaldo Jimenez), big curveballs (Jhoulys Chacin) and sliders (Houston Street). Each of these guys enjoyed some success, but the underlying theory never worked long-term. They even tried a piggy-backing strategy where they had 4 only starters and numerous long-relievers, and no starter went more than 4 innings of 75 pitches! If the Rockies had a significant pitching advantage at Coors, they would have found a solution and stuck with it.
The bottom line for me: I agree with you, I think both factors are in play. They have a large home field advantage and a not-insignificant road field disadvantage. But as long as that road field disadvantage exists to some degree, it means the Rockies current PF is too high.
I wonder how much of this is the Rockies’ front office prioritizing players that will succeed in Coors’ extreme environment. It makes sense that contact-oriented guys who can put the ball in play and run well will do much better at Coors than a standard hitter will, given the enormous dimensions and high BABIP of the park. Players who can hit fastballs well but struggle vs breaking pitches would also presumably have big home/away splits at Coors.
Similar to the above discussion about pitching, I completely agree with this idea in theory, but it just doesn’t seem like the Rockies have successfully done this. I looked at the team leaderboards for the years in this data set (2011-2015)…
Regarding a contact-oriented approach, the Rockies had the 18th highest K%. The underlying numbers agreed, as they also posted the 2nd highest swing% but only the 19th highest contact%. Not terrible, but below average and not at all what you want at Coors. Base running was better, as they posted the 11th highest BsR score. But even though this is above average, it should be much higher if they were actually using speedy contact guys.
I also thought of defense – Coors inflates BABIP, so running out an elite defense to suppress BABIP seems smart. Unfortunately, the Rockies are pretty much the bottom of the barrel here, with agreement between DRS and UZR leading to an overall DEF scores that ranks 27th.
The only think left I can think of is the “fastball hunters” that you mentioned. I haven’t taken the time to look at this (I would love if someone would), but if the rest of the data is any indication, I think we can safely assume the Rockies FO has not been able to maximize their home field advantage.
I really just think metrics don’t really properly handle Coors Field because it is so much different than other places. This is a personal opinion so take it as that and not some argument but I think it skews things majorly like WAR and Defensive metrics. The highest WAR ever by a Rockies player was Larry Walker in 1997 at 9.1. That’s a pretty good WAR but keep in mind his stats line was .366/.452/.720 with 49 HR 143 RBI and 33 SB. This was pre-humidor and all. The best post humidor season was just 7 WAR, yes just considering some of the seasons guys have had. Helton in 2004 hit .347/.469/.620 32 HR 115 RBI. Am I saying he should have had 12 WAR no but park factors etc skew so much because of the oddity of Coors it makes WAR and defensive metrics just a mild input to me.
With WAR starting to factor heavily into things the Rockies will never have another MVP. I wrote an article a while back about what it would take and the only one they’ve ever had (Walker, 97) will be impossible to get close to as everything he did in all the big categories has barely been accomplished at all in the league the past 10 years.
I remember reading that article! I couldn’t agree more. I think we have surpassed the realm of “he’s a rockie, so take it with a grain of salt” and gone straight to “he’s a rockie, so LOLCOORZ it’s not real.” If we don’t adjust, we will directly influence MVP races and HoF voting.
But of course, as you said, this is a purely opinionated tangent.
Glad someone read it 😉 . I love sabermetrics and the calculations but as someone who has watched Rockies baseball for 23 years I understand how different Coors is. At the same time every time I look at league WAR vs Rockies WAR it just never lines up. 90% of analysis I read/hear from scouts/former players on the radio etc say Nolan Arenado is a top 5 position player in MLB. You’d think that has some weight but if you just focus on WAR.. Arenado is ranked 23rd among position players in WAR from 2013-present. If Coors has such a high park factor shouldn’t Arenado’s 3 straight gold gloves/defensive prowess have a higher influence on WAR because it doesn’t seem to.