Here's a "head-scratcher" for you (after 138 games):
Team A: 592RS/701RA (-109 Run Differential)
http://fantasygames.sportingnews.com/baseball/stratomatic/2007/team/team_other.html?user_id=5420
Team B: 630RS/545RA (+85 Run Differential)
http://fantasygames.sportingnews.com/baseball/stratomatic/2007/team/team_other.html?user_id=3104
TeamB (mine of course) is in 2nd place, 1 game behind TeamA, which has a 75-63 record even though I have +194 Run Differential over my divisional rival????????
WEST W L PCT GB L10 Strk HM RD vs.L vs.R RS RA
Trailer Trash 75 63 .543 - 7-3 L1 43-26 32-37 30-18 45-45 592 701
Colonials 74 64 .536 1.0 7-3 W1 43-29 31-35 21-21 53-43 630 545
Mania! 58 80 .420 17.0 5-5 W2 28-38 30-42 15-19 43-61 608 706
The Equalizers 54 84 .391 21.0 3-7 L2 31-38 23-46 14-28 40-56 585 687
Run differential (pythagorean) no longer matters?
22 posts
• Page 2 of 3 • 1, 2, 3
by geekor » Tue Jul 31, 2007 11:37 pm
Look at this league (ATG3):
http://fantasygames.sportingnews.com/baseball/stratomatic/atg3/league/standings.html?group_id=10763
Notice the team with the best RD is at .500 (luckily in a bad division and leading the division).
I have the 2nd best RD and right now won't even make the playoffs (4thbest record.
Of the 2 teams in front of me in my division, one has the best record, the other is tied for 2nd best and the current WC leader. The person with the best record in the league has a NEGATIVE RD, the person with the 2nd, is even.
Once again, it doesn't work anymore, period, end of story. With salary so much closer to perfect, and the gain of experience and knowledge to everyone, Luck is a MUCH bigger factor than anyone gives credit for. Back in the day luck may have been 10-20%, now it's closer to 40% of if you win or lose.
http://fantasygames.sportingnews.com/baseball/stratomatic/atg3/league/standings.html?group_id=10763
Notice the team with the best RD is at .500 (luckily in a bad division and leading the division).
I have the 2nd best RD and right now won't even make the playoffs (4thbest record.
Of the 2 teams in front of me in my division, one has the best record, the other is tied for 2nd best and the current WC leader. The person with the best record in the league has a NEGATIVE RD, the person with the 2nd, is even.
Once again, it doesn't work anymore, period, end of story. With salary so much closer to perfect, and the gain of experience and knowledge to everyone, Luck is a MUCH bigger factor than anyone gives credit for. Back in the day luck may have been 10-20%, now it's closer to 40% of if you win or lose.
- geekor
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
by CHARLESBELL » Thu Oct 18, 2007 4:59 pm
Really, though, what does it matter one way or the other? RD can't be used in any predictive way anyway. It can't help in drafting a team, or setting lineups, or choosing a rotation. It can't even predict second half winning percentage based on first half winning percentage since the sample size would be too small.
It's a fun algorithm to play with, but even if it does provide an accurate estimate with a large enough sample size, what good is it for anything practical?
The only measure of how many games you win is - how many games you win.
It's a fun algorithm to play with, but even if it does provide an accurate estimate with a large enough sample size, what good is it for anything practical?
The only measure of how many games you win is - how many games you win.
- CHARLESBELL
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
by J-Pav » Fri Nov 23, 2007 12:13 pm
[b:4e32a3b3e0]Charlie[/b:4e32a3b3e0]:
That's not exactly true, because winning is still a function of net runs and you can (to a degree) build your team with that in mind. Think of the opposite...could you build a team that couldn't score and yielded lots of opposition runs? ([b:4e32a3b3e0]Cummings[/b:4e32a3b3e0] does this all the time! j/k!!) Could HAL "love" such a team into greatness? I doubt it. So managerial competence is important (to a certain extent) and your lineups can be predictive (to a certain extent).
I have experienced plenty of frustration this year and last, so I agree with [b:4e32a3b3e0]Terry[/b:4e32a3b3e0] and [b:4e32a3b3e0]geekor[/b:4e32a3b3e0] in that (1.) there are weird outliers, and (2.) they are always happening [i:4e32a3b3e0]against[/i:4e32a3b3e0] me!
But, trying to be more rational and objective, I suggest the following. Go to the expanded standings page and look at the pythag records. On a sheet of paper, do something like this:
-8(+) -7 -6 -5 -4 -3 -2 -1 0 +1 +2 +3 +4 +5 +6 +7 +8(+)
Then chart each of the 12 teams and put a check by their [i:4e32a3b3e0]net[/i:4e32a3b3e0] performance versus their pythag record. That is, if your team went 90-72 and the pythag record was 93-69, put a check by -3.
I did this for my last ten leagues and found it to resemble (more or less) a perfect bell curve.
What I found:
53% of 120 teams were within +/- 2 games of projection.
73% were within +/- 4 games.
90% were within +/- 6 games.
93% were within +/- 7 games.
This is a very small sample in the SOM world, but I think it makes the point. Run differential is still relevant.
In this very short micro sample, I noted one league where four teams were +/- 8 (two on each side). So [i:4e32a3b3e0]one half[/i:4e32a3b3e0] of the outliers came from [i:4e32a3b3e0]one league[/i:4e32a3b3e0] ( :shock: ).
But, being fair and selecting a well chosen example, another league had ten teams all within +/- 2 games of projection.
Remember too that gifts from HAL are divided evenly among good teams and bad. There are plenty of 60 win teams that were projected to lose [i:4e32a3b3e0]even more[/i:4e32a3b3e0], and there were plenty of 90 win teams that should have won 100.
Conclusion:
In the grand scheme, run differential still matters. But as I wrote in [i:4e32a3b3e0]The Secret Formula[/i:4e32a3b3e0] post, close games are coin flips and coin flips are luck. Better pricing equals more close games, so it will continue to [i:4e32a3b3e0]appear[/i:4e32a3b3e0] that the whole thing is approaching [i:4e32a3b3e0][b:4e32a3b3e0]all luck[/b:4e32a3b3e0][/i:4e32a3b3e0] (but it really isn't).
That's not exactly true, because winning is still a function of net runs and you can (to a degree) build your team with that in mind. Think of the opposite...could you build a team that couldn't score and yielded lots of opposition runs? ([b:4e32a3b3e0]Cummings[/b:4e32a3b3e0] does this all the time! j/k!!) Could HAL "love" such a team into greatness? I doubt it. So managerial competence is important (to a certain extent) and your lineups can be predictive (to a certain extent).
I have experienced plenty of frustration this year and last, so I agree with [b:4e32a3b3e0]Terry[/b:4e32a3b3e0] and [b:4e32a3b3e0]geekor[/b:4e32a3b3e0] in that (1.) there are weird outliers, and (2.) they are always happening [i:4e32a3b3e0]against[/i:4e32a3b3e0] me!
But, trying to be more rational and objective, I suggest the following. Go to the expanded standings page and look at the pythag records. On a sheet of paper, do something like this:
-8(+) -7 -6 -5 -4 -3 -2 -1 0 +1 +2 +3 +4 +5 +6 +7 +8(+)
Then chart each of the 12 teams and put a check by their [i:4e32a3b3e0]net[/i:4e32a3b3e0] performance versus their pythag record. That is, if your team went 90-72 and the pythag record was 93-69, put a check by -3.
I did this for my last ten leagues and found it to resemble (more or less) a perfect bell curve.
What I found:
53% of 120 teams were within +/- 2 games of projection.
73% were within +/- 4 games.
90% were within +/- 6 games.
93% were within +/- 7 games.
This is a very small sample in the SOM world, but I think it makes the point. Run differential is still relevant.
In this very short micro sample, I noted one league where four teams were +/- 8 (two on each side). So [i:4e32a3b3e0]one half[/i:4e32a3b3e0] of the outliers came from [i:4e32a3b3e0]one league[/i:4e32a3b3e0] ( :shock: ).
But, being fair and selecting a well chosen example, another league had ten teams all within +/- 2 games of projection.
Remember too that gifts from HAL are divided evenly among good teams and bad. There are plenty of 60 win teams that were projected to lose [i:4e32a3b3e0]even more[/i:4e32a3b3e0], and there were plenty of 90 win teams that should have won 100.
Conclusion:
In the grand scheme, run differential still matters. But as I wrote in [i:4e32a3b3e0]The Secret Formula[/i:4e32a3b3e0] post, close games are coin flips and coin flips are luck. Better pricing equals more close games, so it will continue to [i:4e32a3b3e0]appear[/i:4e32a3b3e0] that the whole thing is approaching [i:4e32a3b3e0][b:4e32a3b3e0]all luck[/b:4e32a3b3e0][/i:4e32a3b3e0] (but it really isn't).
- J-Pav
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
by CHARLESBELL » Tue Dec 11, 2007 5:02 pm
J-Pav,
Thanks, I agree with everything you said. I agree that run differential can, given enough samples, closely match W-L record. The question that I don't think has been answered, though, is so what? It doesn't take RD to tell me that the more runs I score than my opponent the more games I am likely to win.
Run differential is based on games already played. All it really "predicts" is that if the same ratio of runs scored to runs allowed persists one should see a similar match in W-L.
I even get it that you could use RD, sort of, to determine if your team is under or over performing and perhaps make current season adjustments. You'd have to believe that the differential is accurate and not a fluke based on the sample size, but it could be used that way.
Where I still scratch my head a bit is how it would be used in a predictive way to put your "next" team together. The same logic for in-season adjustments doesn't really apply to subsequent teams. I think there are too many variables from league to league to be able to use run differential in pre-played predictive way, with one exception, and that is if you want to build the same team as before, in a similar league, and you use RD to see if you should buy more run producers or more run denyers the next time. Beyond that you are trying to adjust to too many variables and RD isn't an effective predictor in the majority of situations.
Run differential still matters, sure, but only after the fact. To me, it is a post season stat where I can say Wow! I really got lucky with this team, or Crap! HAL screwed me again! :-)
Thanks, I agree with everything you said. I agree that run differential can, given enough samples, closely match W-L record. The question that I don't think has been answered, though, is so what? It doesn't take RD to tell me that the more runs I score than my opponent the more games I am likely to win.
Run differential is based on games already played. All it really "predicts" is that if the same ratio of runs scored to runs allowed persists one should see a similar match in W-L.
I even get it that you could use RD, sort of, to determine if your team is under or over performing and perhaps make current season adjustments. You'd have to believe that the differential is accurate and not a fluke based on the sample size, but it could be used that way.
Where I still scratch my head a bit is how it would be used in a predictive way to put your "next" team together. The same logic for in-season adjustments doesn't really apply to subsequent teams. I think there are too many variables from league to league to be able to use run differential in pre-played predictive way, with one exception, and that is if you want to build the same team as before, in a similar league, and you use RD to see if you should buy more run producers or more run denyers the next time. Beyond that you are trying to adjust to too many variables and RD isn't an effective predictor in the majority of situations.
Run differential still matters, sure, but only after the fact. To me, it is a post season stat where I can say Wow! I really got lucky with this team, or Crap! HAL screwed me again! :-)
- CHARLESBELL
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
by maligned » Wed Dec 12, 2007 8:10 am
I agree with geekor that there is more luck involved now than ever, but run differential has not improved or worsened in its ability to predict a team's record.
You can find just as many results not reflecting our theorem if you look back at your 2004 or 2005 results as you find now--it's just that the teams with the best run differential won more often because they typically had so much more run differential (and thus, on average, more victories).
You can find just as many results not reflecting our theorem if you look back at your 2004 or 2005 results as you find now--it's just that the teams with the best run differential won more often because they typically had so much more run differential (and thus, on average, more victories).
- maligned
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
by J-Pav » Fri Dec 14, 2007 10:14 am
[quote:40d0630656="charliewb"]
Where I still scratch my head a bit is how it would be used in a predictive way to put your "next" team together. The same logic for in-season adjustments doesn't really apply to subsequent teams. I think there are too many variables from league to league to be able to use run differential in pre-played predictive way, with one exception, and that is if you want to build the same team as before, in a similar league, and you use RD to see if you should buy more run producers or more run deniers the next time. Beyond that you are trying to adjust to too many variables and RD isn't an effective predictor in the majority of situations.
Run differential still matters, sure, but only after the fact. To me, it is a post season stat where I can say Wow! I really got lucky with this team, or Crap! HAL screwed me again! [/quote:40d0630656]
My last finished team just missed the Finals. I wanted them to go all the way because it would have been a fun team to post. They finished backwards in just about every category of importance (even though I wasn't [i:40d0630656]trying[/i:40d0630656] to accomplish this)!
Sixth in offense (.247 BA, .315 OBP, .469 SLG)
Seventh in pitching (4.64 ERA, 1.33 WHIP)
Negative 27 run differential!
Many players with poor individual stats, but...
10 pythag wins (Brilliant coaching :wink: !!) and a trip to the playoffs!
I understand where you're coming from with the prediction frustration. I've had many teams just "not show up" for me in '07 despite my best laid plans. I think you have two points that cloud one issue: "run differential" and "predictiveness" versus "wins."
How do you build a team that (you can predict) will score more runs and give up less? This is what we're [i:40d0630656]all[/i:40d0630656] trying to do as managers, right? You can go high dollar starters in a pitchers park w/ strong defense, you can go Big Bats in The Cell w/ no BP HR pitchers, and on and on. You can build an RC27 formula or an OPS formula or count card points for hits or walks or home runs or whatever you decide is important.
But in the end it's [i:40d0630656]dice rolls[/i:40d0630656], and once in a while a team of nine Albert Pujols' will lose to five Boof Bonsers (like the Geico gecko says, "that's a complete dramatization, but you get my point"). That doesn't mean you don't want a lineup of nine Pujols' going in, though.
So run differential doesn't only matter after the fact. But [b:40d0630656]wins[/b:40d0630656] only matter after the dice rolls (luck), and the exact same teams can finish plus or minus 10 wins (or more!) with each simulation season. So you can (somewhat) predict how many runs your team (might? should?) score, but you can expect to be wrong some part of the time. So I wholeheartedly agree that doing everything right and still losing is completely aggravating, but you can still try to build a better mousetrap. Heck, if it worked out exactly as we planned each time would that really be much fun? :D
Where I still scratch my head a bit is how it would be used in a predictive way to put your "next" team together. The same logic for in-season adjustments doesn't really apply to subsequent teams. I think there are too many variables from league to league to be able to use run differential in pre-played predictive way, with one exception, and that is if you want to build the same team as before, in a similar league, and you use RD to see if you should buy more run producers or more run deniers the next time. Beyond that you are trying to adjust to too many variables and RD isn't an effective predictor in the majority of situations.
Run differential still matters, sure, but only after the fact. To me, it is a post season stat where I can say Wow! I really got lucky with this team, or Crap! HAL screwed me again! [/quote:40d0630656]
My last finished team just missed the Finals. I wanted them to go all the way because it would have been a fun team to post. They finished backwards in just about every category of importance (even though I wasn't [i:40d0630656]trying[/i:40d0630656] to accomplish this)!
Sixth in offense (.247 BA, .315 OBP, .469 SLG)
Seventh in pitching (4.64 ERA, 1.33 WHIP)
Negative 27 run differential!
Many players with poor individual stats, but...
10 pythag wins (Brilliant coaching :wink: !!) and a trip to the playoffs!
I understand where you're coming from with the prediction frustration. I've had many teams just "not show up" for me in '07 despite my best laid plans. I think you have two points that cloud one issue: "run differential" and "predictiveness" versus "wins."
How do you build a team that (you can predict) will score more runs and give up less? This is what we're [i:40d0630656]all[/i:40d0630656] trying to do as managers, right? You can go high dollar starters in a pitchers park w/ strong defense, you can go Big Bats in The Cell w/ no BP HR pitchers, and on and on. You can build an RC27 formula or an OPS formula or count card points for hits or walks or home runs or whatever you decide is important.
But in the end it's [i:40d0630656]dice rolls[/i:40d0630656], and once in a while a team of nine Albert Pujols' will lose to five Boof Bonsers (like the Geico gecko says, "that's a complete dramatization, but you get my point"). That doesn't mean you don't want a lineup of nine Pujols' going in, though.
So run differential doesn't only matter after the fact. But [b:40d0630656]wins[/b:40d0630656] only matter after the dice rolls (luck), and the exact same teams can finish plus or minus 10 wins (or more!) with each simulation season. So you can (somewhat) predict how many runs your team (might? should?) score, but you can expect to be wrong some part of the time. So I wholeheartedly agree that doing everything right and still losing is completely aggravating, but you can still try to build a better mousetrap. Heck, if it worked out exactly as we planned each time would that really be much fun? :D
Last edited by J-Pav on Fri Dec 14, 2007 10:17 am, edited 1 time in total.
- J-Pav
- Posts: 55
- Joined: Tue Jul 03, 2012 2:34 pm
22 posts
• Page 2 of 3 • 1, 2, 3
Who is online
Users browsing this forum: No registered users and 2 guests