Page 1 of 3

Run differential (pythagorean) no longer matters?

PostPosted: Fri Apr 27, 2007 12:13 pm
by geekor
Look, as was mentioned back in 02, there was a formula for how many runs you score versus you give up should give you your record. This NO LONGER APPLIES though. Run differential no longer matters.

Maybe in 05, but with uber accurate pricing in 06, it barely mattered. Now with no extreme ballparks, it is useless.

Lets look at this league: http://fantasygames.sportingnews.com/baseball/stratomatic/2007/league/standings.html?group_id=101

Notice firstly why the playoff seeidng should be changed ONCE AGAIN.

Now look at those top 3 record run differential. None have more that +30. The person with the best RD has the 4th best record. The person (yes me and my b!tching) with the 2nd best RD is barely at .500.

Now before lucky believed that the bullpen and 1 run games play a big part in that. I stnad before you saying no. Look at the team with the best record. He has played in easily the most 1 run games, winning a huge portion of them. With a small RD, you'd think pitchers park with a good closer. Nope and nope. He's in Chase, a good hittes park, with a average at best bullpen.

Goes to show, [b:5ef3716325]LUCK [/b:5ef3716325]is the biggest factor for winning in 07.

PostPosted: Fri Apr 27, 2007 12:55 pm
by fred_1_15301
How come you've only cited one instance where this happened? Do you really think a sample size of 1 is enough to support your conclusion? In the league I'm in, the three teams with the highest RD are all leading their divisions. I don't think your post was serious but we all know this is much more than just luck.

PostPosted: Fri Apr 27, 2007 1:30 pm
by geekor
I tlaked about this in the 06 game as well as ATG3,

I'm just trying to be more detailed about it, but work gets in the way :P

http://fantasygames.sportingnews.com/baseball/stratomatic/2007/league/standings.html?group_id=103

This league the best RD has the best record, but the rest of the league doesn't match up to their pythagorean records.

http://fantasygames.sportingnews.com/baseball/stratomatic/2007/league/standings.html?group_id=105

This league the best RD has the 3rd best record. The Central's leader has a negative RD while the team in 2nd has a +63 RD.

plus I'm only looking at leagues that are close to finishing, as things could change closer to the end of the season.

there is a pythagorean formlua (check out diamonddope.com) for determining your record based off the runs scored vs runs agiasnt. I'm saying this no longer works for the 07 game (really it didn't work for 06 either). With no extreme ballaprks, and the pricing accuracy, the skill level evens out a lot. Haivng your team do well is more determined with luck than previous season. All luck, that's not what I' saying. But a much higher percentage is determined by luck.

There is more to the game than just numbers

PostPosted: Fri Apr 27, 2007 3:12 pm
by KnightErrant
I have similar example, only I'm on the winning side:
[url]http://fantasygames.sportingnews.com/baseball/stratomatic/2007/league/standings.html[/url]

A straight pythagorean calculation should have me in second place. But, the pythagorean formula is just a rule of thumb not a firm calculation. - [i:49c9691c8b]If the Runs Scored and Runs Against are evenly distributed between the games played, I will win [b:49c9691c8b]X[/b:49c9691c8b] number of games.[/i:49c9691c8b]

There is another component. Not luck but Chaos. In my case and yours, the team with the better record in one-run games has better speed and more power. That may translate into an ability to strike quickly in tied games. Or not, who knows?
:roll:
For what is is worth, I'm not getting my share of luck this season. Jason Bay is hitting .[b:49c9691c8b]209[/b:49c9691c8b] while Manny Ramirez is batting a whole [b:49c9691c8b].234[/b:49c9691c8b]. My point is that all the seasons are not played with an [i:49c9691c8b]even[/i:49c9691c8b] distribution of runs but a [i:49c9691c8b]random[/i:49c9691c8b] distribution. This randomness is why Ramirez is hitting .087 below his card. If we played a 1000 game season, Ramirez would probably be hitting over .300, and I would possibly be in second place.

Of course, I prefer the Jeff Francis Theory. Francis is Foxboro's ace and my number 2. The man's got mojo.

PostPosted: Sun Apr 29, 2007 10:56 am
by Terry101
In the long run, there is no doubt the formula will work. It is not some formula to work just in baseball- it works in football, etc. Runs scored and runs scored against are the surest measure of getting the winning pct. It is a linear relationship. Sure, if your one game winning pct. is out of whack it will throw the formula off, if you win 19-1 and lose many one games- yes it will not be totally accurate but I've seen the formula work taking into account over 100 years of major league baseball- tens of thousands of games.
Now why wouldn't it work in Strat 07? That is the question. I proposed a similar question on the other boards, saying "Is the 07 game different for some reason?" Most of the replies were along the lines of - No the game is the same- one just needs to adjust to the nuances.

What I have been reading and have been experiencing is the [b:57aa6fc52a]extremes[/b:57aa6fc52a]- especially on the low end (great hitters batting .180 for over 250 at bats, etc. are way too common. I understand that in a random distrubution this is likely to happen but it seems like it is happening more often than chance would allow.( One of my teams had EIGHT regulars batting .200 or under after 30 games- that is a combined 900 at bats) The odds on a probability of a .280 occurrence(their expected batting average) after 900 events coming out at a .200 is astronomical.

The logical and mathematical answer is that the formula HAS to work and will work if the [b:57aa6fc52a]distribution is a normal random distribution[/b:57aa6fc52a]. If it is [b:57aa6fc52a]not [/b:57aa6fc52a]a normal distribution- in other words, if the game engine is different- for whatever reason- then there will be extremes that can not be accounted for by normal distribution and the law of large numbers.

It may be that we are witnessing a very rare event: an inordinate amount of lopsided games and unusual winning pct. in one run games- resulting in a skewed team winning pct. that does not match the pythagorean theorem. This is the only logical answer. I guess time and more data will give us the answer.

PostPosted: Sun Apr 29, 2007 4:46 pm
by Palmtana
I surveyed 10 '07, 80M, autoleagues (all have played 144 games) and determined the average run differential by standing in the division.

10 leagues x 3 divisions in each league = 30 data points for each average

Average RD for the 30 teams in:

1st = 63.0
2nd = 0.3
3rd = 1.3
4th = -64.6

The numbers show that RD is a good predictor of the top and bottom teams. I'm surprised 2nd and 3rd place were so close.

PostPosted: Fri May 04, 2007 10:44 pm
by durantjerry
I think that it should apply more to 2007 than any previous tear. One thing in SOMO that I believe skews the run to win comparison is the extreme parks. Extreme parks lead to extreme run differentials IMO. With the parks being more neutral it seems the theorem should apply more than in previous years rather than less.

PostPosted: Sat May 05, 2007 10:19 pm
by blue turtle
Geekor,
I hope our two teams make it through the semi-finals...I'd like to see if my low runs scored or given team matches up against your offense!

PostPosted: Tue Jul 17, 2007 2:40 pm
by Valen
I read in one of the Bill James deals that the original formula tends to be less accurate as more runs are scored. His explanation was that with higher scoring teams blowouts which skew the statistics are more common. He finished the article by providing a new formula he said should be applied against modern teams. I do not remember exactly what the new formula was but it had a multiplier that came from some magical source.

PostPosted: Tue Jul 17, 2007 5:31 pm
by Mean Dean
[url=http://en.wikipedia.org/wiki/Pythagorean_expectation]Wiki claims[/url] that, rather than squaring the runs scored and runs allowed, the exponent should be:[code:1:e1439f33b1]((Runs Scored by Team+Runs Allowed by Team)/Games)^.287[/code:1:e1439f33b1]