08012018, 09:30 PM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


Quote:
Originally Posted by BobNJ1960
Not sure I agree Phils were lucky. IMO some teams perform well under pressure, some wilt, and the gap vs Phythag reflects that regularly.
.

Their luck was not dependent upon your agreement, it was their good luck. It has already happened. If it was something else, it could be found in the numbers, if it was a skill, it would be repeatable. And the gap between the Pythag wins and actual wins is luck, that is precisely what the formula was designed to measure...to what degree does a team owe its successes or failures to factors out of their control?
Consider: Suppose that a baseball manager knew in advance that his team was going to score 700 runs and give up 600 runs in the coming season. Further, the manager has control over how many runs score in each game, being able to add or stop at will. Each game then would become a bidding war, with the manager upping his runs scored total just enough to beat whatever the opposing manager offered. All games would be one run games because it would be foolish to exhaust your run supply with a redundant investment of unneeded runs.
But of course that isn't how baseball works. managers do not know in advance how many runs the team will score and give up, nor are they able to control how many score in a single game. Thus there will be games where you win by a superfluous amount and games where you lose by a superfluous amount. Putting your runs scored and prevented together in the most productive possible pattern is beyond any team or manager's ability. Consequently we know that luck in the distribution of your runs does exist. And we know that Bill James figured out a formula to measure that luck, and we know that the formula works because year after year, most teams either hit or stray only one or two games from the projections. Each year there are anomalies and it is in the anomalies that we find the good or rotten luck. A team which wins far more games than projected will nearly always have an impressive record in one run games, a team which underachieves badly will nearly always have poor records in those games.
It isn't a theory, it is something which does exactly what it sets out to do which is to provide us with a greater understanding of a team's true talent level. It doesn't matter whether or not you believe in it any more than it matters whether or not you believe that the application of sufficient heat will cause water to boil. It works whether or not you believe in it.

08012018, 09:45 PM



3,576 posts, read 1,234,228 times
Reputation: 3732


Even better than pythag is base runs.
Runs themselves distributions of hits, walks, errors, 3rd outs etc
Base Runs have the Phillies at 5750 and the Nats at 5947

08012018, 10:33 PM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


Quote:
Originally Posted by CBeisbol
Even better than pythag is base runs.
Runs themselves distributions of hits, walks, errors, 3rd outs etc
Base Runs have the Phillies at 5750 and the Nats at 5947

Not a huge difference at the moment for those two teams. Base runs is 7 wins off the actual records while the Pythag is ten wins off.
I don't have time now, but later I'll try and take a look to see how the two approaches are doing for all teams.

08022018, 06:47 AM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


Okay, I ran a comparison for the American League teams. As of this morning:
Actual wins: 800
Pythag wins: 807
Base run wins: 806
Well, that does not establish which method is more accurate, but it sure does establish the accuracy of the theories behind the calculations. And of course both methods benefit here from expanding the data base from a single team to a league.
I've no problem with base runs being more accurate, but has that been established? In one of his early Abstracts Bill James stated that he had discovered that the Pythag formula worked more accurately than did the offensive components version, but then he was employing runs created and that has since then been shown to be less accurate than the base runs formula.

08022018, 10:42 AM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


I've run the figures for the NL as well. As of this morning:
Actual wins: 819
Pythag wins : 814
base runs wins 816
Which gives a ML total of:
Actual wins: 1619
Pythag wins: 1621
base run wins: 1622
A difference of one win between the two systems. We cannot say one system is more accurate than the other based on this year's results. Has this been studied on a year to year basis?
For those who are wondering what we are talking about, the difference between the base runs formula and the pythag formula is that the Pythag uses actual runs scored and prevented while the base runs uses theoretical runs scored and prevented based on all of the teams accomplishments in scoring and preventing runs. It is two roads leading to the same end.

08022018, 10:57 AM



3,576 posts, read 1,234,228 times
Reputation: 3732


I don't think looking at total wins is the way to go. They're always going to be close because over any given number of games there's going to be a certain number of wins.
Better to look at wins for each team
And the correlations and standard deviation of those wins to see which is more accurate
Also, the discussion started because we were talking about current record can future performance.
So, looking at actual record, pythag and base runs wins on a certain date and comparing that winning percentage to rest of season winning percentage would be what we're really looking for

08022018, 11:15 AM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


Quote:
Originally Posted by CBeisbol
I don't think looking at total wins is the way to go. They're always going to be close because over any given number of games there's going to be a certain number of wins.
Better to look at wins for each team
And the correlations and standard deviation of those wins to see which is more accurate

Has this been done? You stated that the base runs were more accurate, but how would you know unless someone has conducted a study and established this?

08022018, 11:17 AM



3,576 posts, read 1,234,228 times
Reputation: 3732


I can find this information on base runs
Quote:
Historically, BaseRuns has worked really well. For the years we have historical BaseRuns data (2002 to 2014), one standard deviation was right around four wins, and the data appears to be normally distributed; 73% of teamseasons have fallen within one standard deviation, 97% of teamseasons have fallen within two standard deviations, and no team had ever exceeded three standard deviations.

https://www.fangraphs.com/blogs/the...serunsfailed/
Full disclosure: this article is entitled The Year BaseRuns Failed because the data wasn't fitting very well. I looked for a follow up article, and did not find it.

08022018, 11:26 AM



3,576 posts, read 1,234,228 times
Reputation: 3732


Quote:
Originally Posted by Grandstander
Has this been done? You stated that the base runs were more accurate, but how would you know unless someone has conducted a study and established this?

I would assume yes. But, I, now, can't find any actual studies. Maybe I saw them in the past, I don't remember. I just know I've had it in my mind that BaseRuns was better than Pythag.
Quote:
if you want to get a better sense of how well the teams have actually played, Base Runs is probably the best bet.
No one would every argue it’s a perfect estimator, only that it’s usually a better tool than Pythagorean Record and almost certainly better than the W/L record if you’re interested in underlying performance.

https://www.fangraphs.com/library/te...andbaseruns/
Logically, it should be, as it, as I said, strips out the sequencing of the actual runs.
Pythag might have a stronger correlation to past wins, since it measures actual runs, and it's actual runs that create wins and losses, but BaseRuns should have a higher correlation to future wins because there's not proof that teams can control their sequencing.

08022018, 12:32 PM



Location: Parts Unknown, Northern California
40,823 posts, read 18,553,245 times
Reputation: 18653


Quote:
Originally Posted by CBeisbol
I would assume yes. But, I, now, can't find any actual studies. Maybe I saw them in the past, I don't remember. I just know I've had it in my mind that BaseRuns was better than Pythag.
https://www.fangraphs.com/library/te...andbaseruns/
Logically, it should be, as it, as I said, strips out the sequencing of the actual runs.
Pythag might have a stronger correlation to past wins, since it measures actual runs, and it's actual runs that create wins and losses, but BaseRuns should have a higher correlation to future wins because there's not proof that teams can control their sequencing.

Thank you for finding the information. I am going to continue to use the Pythag when talking about true team talent levels, not because I insist that it is more accurate, rather because it is a lot simpler to explain to our audience here. We know that both methods work even if one is a bit more precise than the other.

Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: Citydata.com.

