51 replies to this topic

[StupendousMan]

It's time to put some numbers into the discussion. I've looked at the game-by-game records for all AL teams during the period 1961-2006, excluding strike years. For each team, I calculated two things:

a) overall winning percentage at end of season

b) longest consecutive streak of games with cumulative record of 0.500

The results can be expressed very clearly with a simple graph.



The 2007 Red Sox have, at the moment, a winning percentage of about 0.600. The graph shows that teams which end up with similar winning percentages usually have a streak of between 40 and 75 consecutive games during which they win and lose equal numbers of games.

If anyone would like the numbers in tabular form, just ask.

Edit: you can find the numbers for this graph, plus the additional analysis later in this thread, by going to

http://spiff.rit.edu...hmond/baseball/

and looking at the most recent item.

Edited by StupendousMan, 20 July 2007 - 06:48 PM.

[Fratboy]

The 2007 Red Sox have, at the moment, a winning percentage of about 0.600. The graph shows that teams which end up with similar winning percentages usually have a streak of between 40 and 75 consecutive games during which they win and lose equal numbers of games.

So somewhere between 25% and 50% of the season, less for teams that win more than 98 games.

Tremendously useful, my man. Stupendous, in fact.

EDIT: 2004 Red Sox had 82 games of .500. Games 22 to 107, between 4/30 and 8/6, one of the two dots near the 85 game mark and to the right of the .600 mark.

EDIT #2: Actually, the .500 stretch was 86 games. And it was 86 years since they'd won the World Series. Spooky.

Edited by Fratboy, 19 July 2007 - 06:46 PM.

[Worst Trade Evah]

It's time to put some numbers into the discussion. I've looked at the game-by-game records for all AL teams during the period 1961-2006, excluding strike years. For each team, I calculated two things:

a) overall winning percentage at end of season

b) longest consecutive streak of games with cumulative record of 0.500

The results can be expressed very clearly with a simple graph.



The 2007 Red Sox have, at the moment, a winning percentage of about 0.600. The graph shows that teams which end up with similar winning percentages usually have a streak of between 40 and 75 consecutive games during which they win and lose equal numbers of games.

If anyone would like the numbers in tabular form, just ask.

That one chart is pure genius. Talk about a picture worth a thousand words. That's just an outstanding contribution.

P.S. You guys mind toning it down with each other? It's fine arguing about the metric (which I see just as a hueristic to guide expectations and not as a serious research tool or something -- and it's fine at that), but mind ditching some of the nastiness?

P.P.S. Thinking about that chart some more -- check out the team with the second highest season winning % -- just over .700. And they STILL had a near 40 game stretch of .500 play. There's a team that finished @.635 that had a .500 streak of nearly 90 games. How frustrating must *that* have been.

Just a great chart -- should go up on top of the site as a banner or something.

Edited by Worst Trade Evah, 19 July 2007 - 07:07 PM.

[Deathofthebambino]

So if I'm reading you correctly, 3 of the 14 away series and 7 of the 16 home series have gone as you would predict; 8 of the away series and 5 of the home series have gone better than you would predict; and 3 of the away series and 4 of the home series have gone worse than you would predict. So your metric has been accurate for the Sox 33% of the time (10 out of 30 series). Not a bad batting average, but I don't see how it tells you anything about the Sox that you couldn't get just from looking at their overall record, or their home and road splits, or Pythagorean record, or whatever.

CC nails it for me. I just don't see what exactly the metric adds that isn't already readily identifiable in a number of other ways. But, whatever, Manila seems to like it.

It's time to put some numbers into the discussion. I've looked at the game-by-game records for all AL teams during the period 1961-2006, excluding strike years. For each team, I calculated two things:

a) overall winning percentage at end of season

b) longest consecutive streak of games with cumulative record of 0.500

The results can be expressed very clearly with a simple graph.




The 2007 Red Sox have, at the moment, a winning percentage of about 0.600. The graph shows that teams which end up with similar winning percentages usually have a streak of between 40 and 75 consecutive games during which they win and lose equal numbers of games.

If anyone would like the numbers in tabular form, just ask.

This right here, is pretty damn fantastic work though. Thank you SM.

Edited by Deathofthebambino, 19 July 2007 - 07:10 PM.

[Fratboy]

P.P.S. Thinking about that chart some more -- check out the team with the second highest season winning % -- just over .700. And they STILL had a near 40 game stretch of .500 play. There's a team that finished @.635 that had a .500 streak of nearly 90 games. How frustrating must *that* have been.

The .704 team is one familiar to us: the 1998 Yankees. They were a .500 from Games 119 through 156 (August 14 - September 21). Imagine that - the juggernaut Yankees a .500 team down the stretch with almost no momentum going into the playoffs (they won their last 7 against a Cleveland team resting up for the playoffs and expansion Rays)

The .630 team is also another one familiar to us: the 2001 A's. They were a .500 team from Game 1 through Game 86 (July 7, 2001), with a 43-43 record, 8 games behind the Red Sox in the Wild Card race, and 19 games behind the Mariners. We know what happened from here: they obliterated the competition the rest of the way, putting up a 59-17 record (a 0.776 winning percentage!!!) and beating the Twins for the Wild Card by a whopping 17 games.

EDIT:

It's fun to figure out what these other teams are, the outliers. The updated link has the team names, for those interested:

* The dots at the far left and far right are the 2003 Tigers and 2001 Mariners, respectively. We know what happened there.

* The 2004 Red Sox are the uppermost dot just to the right of .600

* The .370 team with about 90 days at .500 are the 1982 Twins. If you recall, they had an absolutely horrendous start through about the first 2.5 months, and then they were a .500 team for a little more than 3 months before fading.

* The .414 team with about 130 days at .500 is another one familiar to us: the 2002 Orioles. 63-63 through 86 games, and then the stellar 4-32 stretch drive. YIKES!

* The .562 team with about 140 days were the notoriously streaky 1987 Brewers.

Good stuff.

Edited by Fratboy, 19 July 2007 - 09:03 PM.

[paulftodd]

Interesting stuff. The length of 500 stretches seemed a bit longer than I would have expected, but thinking about it some more I can see a case where a team could have a hot May, followed by a 500 June and a bad July and then a hot August, and so there could be an 80 game stretch of 500 (3 months) with the hot streak in May cancelling out the slump in July.

Looking at it another way, and looking at only the Red Sox from 2001-2007 and 1978, I classified each Month of the season as A+ (>700 winning pct ) A (650-699), B (600-649), C (550-599), D (450-549), F ( 400-450), and F- (<400) and graded each seasons month from A+= 6 to F- + 0 and here are the monthly grades.

Win Distrbution

	2007	2006	2005	2004	2003	2002	2001	Avg Score	1978
March/April	A	C	D	A+	A	A	B	4.3	C
May	A+	A	C	D	D	A+	D	3.4	A+
June	D	A	A	F	B	F-	B	3	A+
July	F	C	D	D	C	C	D	2.3	D
Aug		F-	A	A+	D	F	F-	2.3	A+
Sept/Oct		D	C	A	A	A	F-	3.3	D
Wins	95 proj	86	95	98	95	93	82		99
Winning Pct	589	531	586	605	586	574	509		607
Record on July 19	56-39	57-36	51-42	50-41	56-40	57-37	55-40		62-28
Grade
	Score	Winning Pct
A+=	6	>700
A=	5	650-699
B=	4	600-649
C=	3	550-599
D=	2	450-549
F=	1	400-449
F- =	0	<400

So having 2-3 months of 500 ball, consecutive or not is not predicative. What is important is how good or bad those other months are. We have had 2 very good months already this season, we need 1 more good one and need to stay away from the disasters that were 2001 and 2006, both of which were due to injuries.

What it really comes down to is knowing is if the 500 run is a slump, in which case there is nothing to worry about, or is it a regression to the teams true talent level this year and the slump is yet to come. I can not believe this team will not score runs, there are just too many guys who are underperforming their career numbers, they simply can not all have declined at the same time.

In the last 42 games, this team scored more than 5 runs 11 times compared to their average of 50% from 2003-2005 (2006 was 41% due to the 2nd half bat failure). For the season, they have scored 2 runs or less almost 30% of the time. It can not last.

Edited by paulftodd, 20 July 2007 - 03:30 AM.

[ManilaSoxFAN]

The monthly breakdown helps, but let's take a closer look. Here is 2003's splits...though other methodologies of seperating the .500 periods from the winning periods are possible, I imagine.

(Sorry -- don't know how to plot for graphs, etc.)

2003 (chrono period==> record of period (season's to-date record... aggregate games over .500)

3/31~4/4==> 4-1 (4-1...+3)
4/5~4/15==> 4-4 (8-5...+3)

4/16~4/20==> 5-0 (13-5...+8)
4/21~6/12==> 23-23 (36-28...+8)

6/13~6/15==> 3-0 (39-28...+11)
6/16~6/24==> 4-4 (43-32...+11)

6/25~6/27==> 3-0 (46-32...+14)
6/28~7/8==> 5-5 (51-37...+14)

7/9~7/12==> 4-0 (55-37...+18)
7/13~8/22==> 18-18 (73-55...+18)

8/23~8/25==> 3-0 (76-55...+21)
8/25~9/1==> 3-3 (79-58...+21)

9/2~9/12==> 7-2 (86-60...+26)
9/13~9/20==> 4-4 (90-64...+26)

9/21~9/22==> 2-0 (92-64...+28)
(/23~9/28==> 3-3 (95-67...+28)


What jumps out is that there was much back-and-forthing between growth and .500-ball, with two big .500-ball stretches:

4/21~6/12===> 46 games
7/13~8/22===> 36 games

The split of "flat" .500-ball periods vs "winning/growth" periods = 128-34. About 80% of the season was played during "flat" periods...with many small bursts in-between. Here is the same batch of periods by total games played in each (flat or growth) mode.

5 games of growth
8 games of flat

5 g
46! f

3 g
8 f

3 g
10 f

4 g
36! f

3 g
6 f

9 g
8 f

2 g
6 f


Here's 2005, another 95-win season:


2005

4/3~4/29==> 11-11 (11...+0)

4/30~5/11==> 10-2 (21-13...+8)
5/12~6/19==> 17-17 (38-30...+8)

6/20~6/26==> 6-0 (44-30...+14)
6/27~7/31==> 15-15 (59-45...+14)

8/1~8/13==> 9-2 (68-47...+21)
8/14~8/29==> 7-7 (75-54...+21)

8/10~9/1==> 3-0 (78-54...+24)
9/2~9/21==> 10-10 (88-64...+24)

9/22~9/27==> 4-0 (92-64...+28)
9/27~10/2==> 3-3 (95-67...+28)


This time, there were fewer back-and forths and the "flat" periods were more equally distributed in size:

4/3~4/29==> 22games
5/12~6/19==> 34 games
6/27~7/31==> 30 games
8/14~8/29==> 14 games
9/2~9/21==> 20 games
9/27~10/2==> 6 games


But this season, there were bigger "growth" periods, especially:

4/30~5/11==> 12 games
8/1~8/13==> 11 games


Looks like there is more ways to skin a 95-win cat. Still the comparison between flat and growth = 126-36. Again, approximately 80% of the season was during flat periods. This basic ratio won't change between 95-win types of seasons. It's just how the flat/growth is distributed.

The '05 periods by amounts of games, btw:

22 games of flat

12 games of growth
34 f

6 g
30 f

11 g
14 f

3 g
20 f

4 g
6 f


Overall, 2005 spread it's flat periods out and depend on big growth spurts.

OTOH, 2003 won via many small spurts, while spending two long periods treading.

What will 2007 look like?

[Tim Lollard]

That chart is like the baseball version of a Judy Blume novel: "Don't worry, your team is normal."

[Fratboy]

The split of "flat" .500-ball periods vs "winning/growth" periods = 128-34. About 80% of the season was played during "flat" periods...with many small bursts in-between. Here is the same batch of periods by total games played in each (flat or growth) mode.

95-67 is 28 games over .500. Take out those 28 wins, and you get a 67-67 over the course of 134 games. Of course it's 80% of the season. (Actually 82%.) It's just about the definition of a 95-win season. This doesn't reveal anything meaningful, and only works for a 95-win season. You're cherry-picking here. This happens in EVERY 95-win season.

For an 81-win, 100% of it is .500, but you have a bunch of up stretches and down stretches that add to a net sum of 0. For a 95 win season, you'll have more up than down. This isn't a revelation.

[ragnarok725]

95-67 is 28 games over .500. Take out those 28 wins, and you get a 67-67 over the course of 134 games. Of course it's 80% of the season. (Actually 82%.) It's just about the definition of a 95-win season. This doesn't reveal anything meaningful, and only works for a 95-win season. You're cherry-picking here. This happens in EVERY 95-win season.

For an 81-win, 100% of it is .500, but you have a bunch of up stretches and down stretches that add to a net sum of 0. For a 95 win season, you'll have more up than down. This isn't a revelation.

This is true, but I think the point Manila is trying to make (and I don't know whether or not it is a valid one) is that those stretches of winning that define an above-average team are generally clumped in short stretches, surrounded by .500 ball. Yes, if you subtract the 28 games over .500 then you're left with 82% of the games at .500. We get that. That's not the point he's trying to make/prove (at least I think).

Simply put, Manila thinks that good teams are generally only good for short streaks and play .500 ball the rest of the time. That's the hypothesis. That the winning is generally pretty tightly clumped into streaks.

[Fratboy]

Simply put, Manila thinks that good teams are generally only good for short streaks and play .500 ball the rest of the time. That's the hypothesis. That the winning is generally pretty tightly clumped into streaks.

It's not a hypothesis. It's a definition of what a good team does. This isn't anything meaningful or revealing.

Conversely, bad teams are bad teams only for short streaks and play .500 ball the rest of the time. That's what happens when the vast majority (I can't put a percentage on it) of teams are clumped between .400 and .600.

Edited by Fratboy, 20 July 2007 - 08:59 AM.

[ManilaSoxFAN]

95-67 is 28 games over .500. Take out those 28 wins, and you get a 67-67 over the course of 134 games. Of course it's 80% of the season. (Actually 82%.) It's just about the definition of a 95-win season. This doesn't reveal anything meaningful, and only works for a 95-win season. You're cherry-picking here. This happens in EVERY 95-win season.

For an 81-win, 100% of it is .500, but you have a bunch of up stretches and down stretches that add to a net sum of 0. For a 95 win season, you'll have more up than down. This isn't a revelation.

True enough -- a 95-win season means that the team will spend the vast majority of the time playing .500 ball. Like right around the 80% rate which I've been citing for a while. Not a revelation; you're right.

24 hours ago, tho, that was repeatedly challenged by you (though you insisted that this statement somehow meant that the entire flat period will occur in one long continuous chunk of time. Which is ludicrous. Which is why I didn't claim that whatsoever in the first place.). Glad it's now all straightened out.


The question which can be asked in this thread -- with the helpful graph at our disposal -- is: Does win distribution matter? Is, for ex, the Sox '05 season a better way to get 95 wins...or is the 2003 approach/result superior, easier, whatever?

Tying this issue to present-day reality...does the 2007 approach/result (so far) tell us anything?

We went 8 flat, 9 growth, 8 flat and 16 growth. That was an outstandingly good flat/growth ratio (~40% flat only). Way above a 95-win team's ratio. But then we've found ourselves in a 54 game flat period. That's a long time, I admit (tho I've only disected two seasons for compative purposes).

But guess what? We're just approaching (in an ugly fashion, I concede) the ~80% flat:growth ratio. Will we plateau at 80%...or just keep slipping to an 85% flat mode...or worse?

So, are we "correcting" or "nose-diving"? Pretty much the question of the day.....

But playing .500 ball for an x period of time -- in and of itself -- isn't a very good measure to determine the answer. Playing .500 ball in DET, in Oak, Sea and LAA is surely different than playing .500 ball at home vs the KC, CWS, TB dregs of the league, tho. Right?

[Shelterdog]

First, Stupendous Man - thanks - this is really interesting and really great work.

Second, it is now clear that Manila is PhillipHughes2.50.

Third, whether you use the Manila Metric (which I'm not endorsing) or the data from Stupendous Man's chart, I think it's pretty clear that you an extended stretch of 500 ball is not enough to say that a good team is bad.

An interesting question would be whether there is some analysis that one could do about the quality of play during the 500 stretch that correlates to the end record. I would expect that you would evidence of bad luck during the 500 stretches for teams that ended up with good records (for example, they might have bad records in one run games, underperform their pythag, have problems with their "contextual runs" - although that statistic is a little strange to me, hit into inordinate numbers of double plays, leave lots of men on base), while the bad teams might show evidence of good luck during their 500 stretches.

[Fratboy]

24 hours ago, tho, that was repeatedly challenged by you (though you insisted that this statement somehow meant that the entire flat period will occur in one long continuous chunk of time. Which is ludicrous. Which is why I didn't claim that whatsoever in the first place.). Glad it's now all straightened out.

It's not my fault you can't write clearly. Thanks for muddying up this otherwise awesome thread by bringing up crap from a now-locked thread.

But guess what? We're just approaching (in an ugly fashion, I concede) the ~80% flat:growth ratio. Will we plateau at 80%...or just keep slipping to an 85% flat mode...or worse?

Why do you keep insisting on this 80% nonsense? It's relevant only to 95-win seasons, and has absolutely nothing to do with StupendousMan's thesis, which is that for .600 teams, they'll have continuous stretches of 40-75 games of .500 performance, and that means that if the Red Sox indeed are a .600 team, what's happened isn't anything unusual. It could almost certainly be said that they're not a .700 team or a .350 team.

I know you're obsessed with 95-win seasons, but please, for the love of humanity, it's outside the scope of this discussion. SSS of 3-5 games are indicative of nothing, because it's cherry-picking. I don't know why you can't grasp this.

[paulftodd]

True enough -- a 95-win season means that the team will spend the vast majority of the time playing .500 ball. Like right around the 80% rate which I've been citing for a while. Not a revelation; you're right.

Manila, my previous post above clearly shows the exact opposite, A 95 win team will spend about 50 - 66% % of the time playing above 500 when measured on a monthly basis.

[StupendousMan]

Time for more graphs!

First, let's look at the longest winning streak during a season, as a function of team's final winning percentage. The correlation isn't as strong as one might expect ...



Next, the flip side: the length of the longest losing streak during a season. It is, not surprisingly, nearly a mirror image of the previous graph. Those are the 1988 Orioles with the 21-game losing streak, of course.



Finally, since someone asked for it:

... good teams are generally only good for short streaks and play .500 ball the rest of the time. That's the hypothesis. That the winning is generally pretty tightly clumped into streaks.

What fraction of a team's games occurs during winning streaks? It is _possible_ for a team to end up with a very good record -- 0.750 -- without a single such streak, by winning 3 in a row, losing 1, winning 3 in a row, losing 1, etc. But that seems unlikely. Or a team might have a single winning streak of 80 games, then alternate wins and losses 41 times, ending up with the same great record, but playing half the season in a winning streak.

So, to answer the question, I arbitrarily picked 5 as the minimum length of "a winning streak". What fraction of team's season occurs during winning streaks of at least this length?



I could pick other lengths (3 wins? 7 wins?) and repeat the analysis, if it seems interesting.

[Rasputin]

Stupendous Man, outstanding contribution.

I remember in 2004 we were having this conversation and I went back and looked at the WS winners to find ones that didn't have a long stretch of mediocre results and stopped after about four years of all of them having a stretch of mediocre results.

When there is as much luck in the game as there is in baseball you're going to have to ride out the rough spots.

An interesting question would be whether there is some analysis that one could do about the quality of play during the 500 stretch that correlates to the end record. I would expect that you would evidence of bad luck during the 500 stretches for teams that ended up with good records (for example, they might have bad records in one run games, underperform their pythag, have problems with their "contextual runs" - although that statistic is a little strange to me, hit into inordinate numbers of double plays, leave lots of men on base), while the bad teams might show evidence of good luck during their 500 stretches.

I would love to see this.

[ragnarok725]

It's not a hypothesis. It's a definition of what a good team does. This isn't anything meaningful or revealing.

Conversely, bad teams are bad teams only for short streaks and play .500 ball the rest of the time. That's what happens when the vast majority (I can't put a percentage on it) of teams are clumped between .400 and .600.

The definition of a good team is one that goes on a few torid stretches and otherwise plays .500 ball? I'm not sure that's true at all. Some good teams might fit that characterization, but I'm betting there are a lot of good teams that are much more consistent than that.

Is it unreasonable to think that some good teams could have very few, very short periods of .500 ball, the rest of the time playing .625 ball consistently to wind up at .600?

I think another really interesting graph to look at would be all of the team's 10+ (mutually exclusive) game streaks of .500 ball added up. How many streaks like that does the team have? How many total games? I think that might be a better indicator of how often (and for how long) a team is capable of playing above average baseball during the course of a year.

[jayhoz]

It's time to put some numbers into the discussion. I've looked at the game-by-game records for all AL teams during the period 1961-2006, excluding strike years. For each team, I calculated two things:

a) overall winning percentage at end of season

b) longest consecutive streak of games with cumulative record of 0.500

The results can be expressed very clearly with a simple graph.

The 2007 Red Sox have, at the moment, a winning percentage of about 0.600. The graph shows that teams which end up with similar winning percentages usually have a streak of between 40 and 75 consecutive games during which they win and lose equal numbers of games.

If anyone would like the numbers in tabular form, just ask.

How is a "consecutive streak of games with cumulative record of 0.500" defined. Would a month of of .600 and a month of .400 ball be considered a streak of .500 ball?

[Jethro Q. Walrustitty]

Is it unreasonable to think that some good teams could have very few, very short periods of .500 ball, the rest of the time playing .625 ball consistently to wind up at .600?

The difference between playing .600 ball and .500 ball over 50 games is a whopping 5 games. It is absolutely believable that a true .600 talent level team might play .500 ball over some 50 game stretch of any given season. It would be highly unlikely for them not to.

[ragnarok725]

The difference between playing .600 ball and .500 ball over 50 games is a whopping 5 games. It is absolutely believable that a true .600 talent level team might play .500 ball over some 50 game stretch of any given season. It would be highly unlikely for them not to.

Right, they might. But they don't have to. Fratboy's post said that every .600 team follows that pattern, that it's a definition, and I don't think it is. SM's first graph showed as much - a good number of .600 teams never had a .500 stretch of even 40 games, much less 50. And even if they have 50+, that's less than a third of the season. This idea that a team would play most of the season at .500 with a few really hot stretches hasn't been proven or disproven so far, so let's not call it a definition.

Edited by ragnarok725, 20 July 2007 - 09:54 AM.

[StupendousMan]

How is a "consecutive streak of games with cumulative record of 0.500" defined. Would a month of of .600 and a month of .400 ball be considered a streak of .500 ball?

Pick two games during the season. Add up all the wins, and all the losses, between those two games. If the number of wins equals the number of losses, then the stretch of games between those two points has a cumulative record of 0.500.

In other words, it is

a consecutive streak of games with cumulative record of 0.500

:-)

Edited by StupendousMan, 20 July 2007 - 09:55 AM.

[jayhoz]

Pick two games during the season. Add up all the wins, and all the losses, between those two games. If the number of wins equals the number of losses, then the stretch of games between those two points has a cumulative record of 0.500.

In other words, it is
:-)

So in your model a team could go 81 - 0 in the first half and 0 - 81 in the second and they would be considered a club that played .500 ball over the length of the season rather than the best team in history during the 1st half and the worst in the 2nd half. This seems like flawed logic to me.

[StupendousMan]

So in your model a team could go 81 - 0 in the first half and 0 - 81 in the second and they would be considered a club that played .500 ball over the length of the season rather than the best team in history during the 1st half and the worst in the 2nd half. This seems like flawed logic to me.

Hey, I'm just perfoming mathematical calculations. Don't blame the messenger, blame the fact that (81 / 162) = 0.500.

You can interpret the results any way you like.

[jayhoz]

Hey, I'm just perfoming mathematical calculations. Don't blame the messenger, blame the fact that (81 / 162) = 0.500.

You can interpret the results any way you like.

I asked the question so that I could more accurately interpret the results. Now that I have the answer I am much less inclined to form opinions based on the data.

I appreciate all the work that goes into these types of analysis and generally really like the fact that it provides me with a perspective on the game that I can't readily formulate myself.

That being said; people will look to you as an "expert" and generally accept your premise because they do not have the time and/or skill to sift through the data themselves. To say that the math is the math regardless of how it impacts your premise is a bit of a cop out don't you think?

[StupendousMan]

I appreciate all the work that goes into these types of analysis and generally really like the fact that it provides me with a perspective on the game that I can't readily formulate myself.

Thanks. It's fun for me to do this work, and I generally learn something through the experience.

That being said; people will look to you as an "expert" and generally accept your premise because they do not have the time and/or skill to sift through the data themselves. To say that the math is the math regardless of how it impacts your premise is a bit of a cop out don't you think?

My premise? I have made no premises, nor have I made any interpretations. Please go back and read through my posts in this thread. All I have done is listen to what other people have said, and tried to collect information and present it for them to see. I suspect that you may have differences with some of the posts in this thread which attempt to come to some conclusions, based either on the information I've provided, or other sources.

[Fratboy]

The definition of a good team is one that goes on a few torid stretches and otherwise plays .500 ball? I'm not sure that's true at all. Some good teams might fit that characterization, but I'm betting there are a lot of good teams that are much more consistent than that.

No, I was placing it within the context of 80% of a 95-win season being a .500 record, basically a 95-win team is defined by going 67-67 in 134 games and 28-0 in all remaining games. That's literally what a 95-win season is, without respect to the quality of the team involved.

In retrospect, saing "that's what a good team does" was an exceedingly poor choice of words, because I don't really believe that. Rather, it's how an XX-win season could be broken down into wins and losses, but shouldn't.

I agree with you I think there's much more consistency than that, being "warm" most of the time instead of "hot" and "cold".

Is it unreasonable to think that some good teams could have very few, very short periods of .500 ball, the rest of the time playing .625 ball consistently to wind up at .600?

Not in the least. I think a scatterplot would reflect that.

<cough>

To point to a recent example, since I'm re-reading the Bill James 1992 Baseball Book, take a look at the 1991 Toronto Blue Jays. They finished 91-71. Check out their monthly results:

Month by Month
Split		   W   L	  RS	  RA	 WP
April		  12   9	  96	  89	.571
May			15  12	 105	  88	.556
June		   16  12	 115	  99	.571
July		   15  11	  99	 101	.577
August		 15  14	 125	 128	.517
September	  14  12	 125	 104	.538
October		 4   1	  19	  13	.800

For five full months, no more than 16 wins, no fewer than 14 wins. THAT is the model of winning consistently.

Edited by Fratboy, 20 July 2007 - 10:27 AM.

[jayhoz]

My premise? I have made no premises, nor have I made any interpretations.

My apologies then. I do think it is a little dangerous (too strong a word) to present data that leads people to an obvious yet quite possibly flawed conclusion though.

[ManilaSoxFAN]

Manila, my previous post above clearly shows the exact opposite, A 95 win team will spend about 50 - 66% % of the time playing above 500 when measured on a monthly basis.

paul, I used the same data you did. Just broke it up according to "flat/growth" streaks rather than monthly divisions. Clearly, and logically, a 95 win team will spend ~80% of the time playing ~.500 ball...and ~20% above 500.

[Clears Cleaver]

Can someone tell me how often an 80-85 win team plays .600 ball for two months? I mean, that is what we could also be seeing here, right?

[Rudy Pemberton]

Can someone tell me how often an 80-85 win team plays .600 ball for two months? I mean, that is what we could also be seeing here, right?

Or it could be 95 win team playing .500 ball for two months. We really don't know yet.

[jayhoz]

Or it could be 95 win team playing .500 ball for two months. We really don't know yet.

When will we know? I suspect the answer to that question is after approximately 162 games.

[ManilaSoxFAN]

How is a "consecutive streak of games with cumulative record of 0.500" defined. Would a month of of .600 and a month of .400 ball be considered a streak of .500 ball?

Pursuant to your (and ragnarok's) query, I did use a definition when breaking doen the flat/growth streaks of the 2003 and 2005 Sox (above). Just not sure how to explain it.

See the 2005 record. The Sox went up and down until they reached 11-11. After that day, they would never break even again. End of .500 ball at that point.

They then climbed up to a +8 level (via 10-2). End of the growth period? -- The first time they reached +8.

After that, they got stuck (varied slightly up and down) at that +8 mark for a long time (34 games of 500 ball). That's the flat period. Until once again, the team broke the +8 mark and went up to the first time a +14 mark was reached. That was the growth period. And so on.

How to define this method is beyond me. It's systematic, though, I believe.

[RingoOSU]

paul, I used the same data you did. Just broke it up according to "flat/growth" streaks rather than monthly divisions. Clearly, and logically, a 95 win team will spend ~80% of the time playing ~.500 ball...and ~20% above 500.

A 67-95 team will also spend 80% of the time playing .500 ball.

[ManilaSoxFAN]

A 67-95 team will also spend 80% of the time playing .500 ball.

True. But, on average, ~20% of the time will be BELOW .500 ball. One would thus measure the losing periods, I guess...not the winning periods.

The original question/doubt was that/if a 95-win team would actually play ~80% of it's games at a .500 level. But, of course, it will. The context was our record during this season. But I don't get where your point fits in. Seriously.

[Fratboy]

But I don't get where your point fits in. Seriously.

His point is this, which I'll phrase in the form of a question:

Which is more meaningful: the 80% of the season at .500 or the 20% of the season at 1.000 or .000?

[sibpin]

So in your model a team could go 81 - 0 in the first half and 0 - 81 in the second and they would be considered a club that played .500 ball over the length of the season rather than the best team in history during the 1st half and the worst in the 2nd half. This seems like flawed logic to me.

Perhaps it's worth drawing lines on the first graph from this thread that has the upper and lower bounds for the data.

For example, a 81-81 team has both an upper and lower bound at 162. All 81-81 teams are by definition at 162. Similarly, a 162-0 team by definition is at 0.

But an 82-80 (.506) team has an upper bound of 160 (assuming, for example, that the team went 80-80 through the first 160 games and then 2-0 the rest of the season). Similarly, a 60-102 (.370) team has an upper bound of 120.

I'm having a difficult time calculating lower bounds, though. It's clear to me that 1-161 and 2-160 teams has a lower bound of 2, but I can't quite figure out everything in between.

[durandal1707]

Trying to define ".500 stretches" is difficult because it's based entirely on how large of an interval you're looking over and at what point in the season you're looking. As it's been mentioned, a team can win its first 81 games and then lose the remaining 81. Are they a .500 team? Of course. They're also a .500 team if you take a sample out of the middle of the season. But of course you can sample from the first half and show they're a 1.000 team, or you can show they're a 0.000 team, and a whole range of Win% values by favoring one side or the other.

I like using the monthly method but the problem with that is that there's not going to be the same amount of games in each month. I'd suggest looking at the season using two intervals - an 18 game interval and a 6 game interval. Both of these numbers are nice because they're factors of 162 and they're even - meaning you can have .500 intervals. The 6-game interval would obviously be a bit more "volatile" as factors like strength of opposition creep in, but it can capture more "local" data than the 18-game swath - for instance, if a team wins 9 in a row and then loses 9 in a row, the 18-game interval shows them at .500, but the three six-game intervals show them as having one winning interval, a .500 interval, and a losing interval.

I don't have a lot of time for breaking up previous seasons at the moment, but here's a quick look at the '07 Sox season thus far:

@KC, @TEX - SEA, LAA, @TOR - @TOR, NYY, TOR
6-Game: 3-3, 4-2, 5-1
18-Game: 12-6

TOR, @BAL, @NYY - OAK, SEA, @MIN - @TOR, BAL
6-Game: 4-2, 4-2, 5-1
18-Game: 13-5

DET, ATL - ATL, @NYY, @TEX - @TEX, CLE, NYY
6-Game: 4-2, 5-1, 4-2
18-Game: 13-5

NYY, @OAK, @ARI - @ARI, COL, SF - SF, @ATL, @SD
6-Game: 2-4, 3-3, 5-1
18-Game: 10-8

@SD, @SEA, TEX - TEX, TB - @DET, TOR
6-Game: 2-4, 4-2, 2-4
18-Game: 8-10

In-Progress:
TOR, KC, CWS
5-Game: 1-4

6-Game Totals:
6-0: 0
5-1: 4
4-2: 6
3-3: 2
2-4: 3
1-5: 0
0-6: 0

I'm not quite sure what to make of this. I'd look at the 6-Game Totals and say that the '07 Red Sox look very good - a majority of their intervals are above .500. I'd say that the team's problem hasn't been playing .500, it's been playing poorly (three 2-4 intervals since June, and the best they can hope for of the current interval is 2-4) with some sprinklings of decent intervals in there (thank you D-Rays, Rangers and Giants). The results may produce more 3-3 (and 9-9) intervals if you adjust the "phase" or perhaps move postponed games into their original intervals, but it seems there hasn't been a consistent stretch of .500 play, just a few good intervals mixed in well with some equally bad intervals.

[ManilaSoxFAN]

His point is this, which I'll phrase in the form of a question:

Which is more meaningful: the 80% of the season at .500 or the 20% of the season at 1.000 or .000?

Over the course of the last many decades in baseball, the difference between say, batting .300 and not (see the famous Bull Durham soliloquy), has been discussed as much as the difference between winning, losing or splitting those last 28 games. So you know the answer.

Still not getting any point here.

I'm wondering if you have an issue somehow with my response to paulftodd -- who claimed that a 95-win team would spend 50~66% of the time playing .500 ball..when measured on a monthly basis. I've explained that -- by breaking it down in smaller pieces (than month-by-month), one can see that it's really more like ~80% at .500 ball.

Is there something inaccurate about my response to paulftodd's question/statement?

[Saints Rest]

paul, I used the same data you did. Just broke it up according to "flat/growth" streaks rather than monthly divisions. Clearly, and logically, a 95 win team will spend ~80% of the time playing ~.500 ball...and ~20% above 500.

Someone once had (and this is a paraphrase) that in a typical season, you can pretty much gurantee 60 wins and 60 losses, it's what happens in the other 40 games that counts.

In other words, relative to the context of this thread, a typical team plays .500 ball for 120 games or 75% of the season. That's almost exactly what Manila says here.

[StupendousMan]

I'm back with another view of the data.

But first, note that you can find a collection of my graphs -- and data in tabular form -- from this thread at

http://spiff.rit.edu...rr/streaks.html

Okay, some people have asked how important "consistency" is. I address this (very big) question in one (small) way in the second half of the document mentioned above. In brief, I broke seasons into "chunks" of 10 non-overlapping games, calculated the winning percentage within each chunk, then computed the standard deviation from the mean of those percentages. Teams which played at the same level all the time -- winning 6 of 10, 6 of 10, 6 of 10, etc., -- would have SMALL values of standard deviation; teams which bounced around in performance -- winning 4 of 10, then 8 of 10, then 4 of 10, etc., -- woul d have LARGE values of standard deviation. I compared this metric of a team's "consistency" to its record at the end of the season.

Then I repeated the analysis for chunks of 20 games at a time, and again for chunks of 30 games at a time. The results are pretty much the same for all three chunk sizes, so I will show here only two of the graphs. First, for 10-game chunks:



and now for 30-game chunks:




I will presume to state a conclusion myself this time: there is little correlation between a team's consistency and its ultimate record. Historically, some teams have managed to win a lot of (or few) games while being relatively consistent, while others have managed to win a lot of (or few) games while running hot and cold.

[ManilaSoxFAN]

Trying to define ".500 stretches" is difficult because it's based entirely on how large of an interval you're looking over and at what point in the season you're looking. As it's been mentioned, a team can win its first 81 games and then lose the remaining 81. Are they a .500 team? Of course. They're also a .500 team if you take a sample out of the middle of the season. But of course you can sample from the first half and show they're a 1.000 team, or you can show they're a 0.000 team, and a whole range of Win% values by favoring one side or the other.

I like using the monthly method but the problem with that is that there's not going to be the same amount of games in each month. I'd suggest looking at the season using two intervals - an 18 game interval and a 6 game interval. Both of these numbers are nice because they're factors of 162 and they're even - meaning you can have .500 intervals. The 6-game interval would obviously be a bit more "volatile" as factors like strength of opposition creep in, but it can capture more "local" data than the 18-game swath - for instance, if a team wins 9 in a row and then loses 9 in a row, the 18-game interval shows them at .500, but the three six-game intervals show them as having one winning interval, a .500 interval, and a losing interval.

I don't have a lot of time for breaking up previous seasons at the moment, but here's a quick look at the '07 Sox season thus far:

@KC, @TEX - SEA, LAA, @TOR - @TOR, NYY, TOR
6-Game: 3-3, 4-2, 5-1
18-Game: 12-6

TOR, @BAL, @NYY - OAK, SEA, @MIN - @TOR, BAL
6-Game: 4-2, 4-2, 5-1
18-Game: 13-5

DET, ATL - ATL, @NYY, @TEX - @TEX, CLE, NYY
6-Game: 4-2, 5-1, 4-2
18-Game: 13-5

NYY, @OAK, @ARI - @ARI, COL, SF - SF, @ATL, @SD
6-Game: 2-4, 3-3, 5-1
18-Game: 10-8

@SD, @SEA, TEX - TEX, TB - @DET, TOR
6-Game: 2-4, 4-2, 2-4
18-Game: 8-10

In-Progress:
TOR, KC, CWS
5-Game: 1-4

6-Game Totals:
6-0: 0
5-1: 4
4-2: 6
3-3: 2
2-4: 3
1-5: 0
0-6: 0

I'm not quite sure what to make of this. I'd look at the 6-Game Totals and say that the '07 Red Sox look very good - a majority of their intervals are above .500. I'd say that the team's problem hasn't been playing .500, it's been playing poorly (three 2-4 intervals since June, and the best they can hope for of the current interval is 2-4) with some sprinklings of decent intervals in there (thank you D-Rays, Rangers and Giants). The results may produce more 3-3 (and 9-9) intervals if you adjust the "phase" or perhaps move postponed games into their original intervals, but it seems there hasn't been a consistent stretch of .500 play, just a few good intervals mixed in well with some equally bad intervals.

The first question is: Doesn't it matter how well a team theoretically "should do" during an arbitrary stretch of time...And then see how well it has actually done?

After all, perhaps it's not if .500 was "good" or not over a period of time...but more importantly, whether it was good vs a specific category of team...and whether it was on the "lesser odds of winning" road or at "higher odds of winning" home turf? Going 5-5 on a West Coast trip is much different than going 5-5 at home against weak teams, no?

Rather than divide by fixed time frames, maybe it'd be more useful to divide by opponent and location. In that way, one might gain more insight into whether a team is kinda ahead of the winning curve, or behind.

[Fratboy]

there is little correlation between a team's consistency and its ultimate record. Historically, some teams have managed to win a lot of (or few) games while being relatively consistent, while others have managed to win a lot of (or few) games while running hot and cold.

Finally, the first sound conclusion reached in this mess. Amen, brotha.

Edited by Fratboy, 20 July 2007 - 12:16 PM.

[ManilaSoxFAN]

Finally, the first sound conclusion reached in this mess. Amen, brotha.

Welcome on board, kid. Perhaps this non-revelation earlier eluded you -- "Looks like there is more ways to skin a 95-win cat...........This basic ratio won't change between 95-win types of seasons. It's just how the flat/growth is distributed."

So now the question is: How does one know if a team is playing a stint of .500 ball...which is somehow indicative of a 95-win season...or whether it's really just a part of an 85-win season? If so much of the time is spent in a "flat" mode, how can anyone ever recognize the indicators that the ability to enter a "growth" period are there?

[absintheofmalaise]

Welcome on board, kid. Perhaps this non-revelation earlier eluded you -- "Looks like there is more ways to skin a 95-win cat...........This basic ratio won't change between 95-win types of seasons. It's just how the flat/growth is distributed."

So now the question is: How does one know if a team is playing a stint of .500 ball...which is somehow indicative of a 95-win season...or whether it's really just a part of an 85-win season? If so much of the time is spent in a "flat" mode, how can anyone ever recognize the indicators that the ability to enter a "growth" period are there?

By seeing what the record is at the end of the season. You can't predict a record based on how a team plays when they are playing .500. Look up the definition of average.
Oh, and frat didn't agree with you.

edit:added the frat thing

Edited by absintheofmalaise, 20 July 2007 - 01:09 PM.

[Fratboy]

I am wrong. Manila's right. It is good and righteous to give Minallah thanks and praise. Surely he shall guide us through these dark times towards 80% of a season filled with the life-nurturing goodness of a .500 winning percentage. My ignorance in these matters is truly astounding, and I am thankful that a shining beacon such as our friend overseas has shown me the path of useful statistical analysis.

[JimBoSox9]

The first question is: Doesn't it matter how well a team theoretically "should do" during an arbitrary stretch of time...And then see how well it has actually done?

Not really. I don't see how beating the Royals and losing to the Tigers would be particularly more predictive than the other way around.

Edit: The thing that I take away from Stupendous Man and Durandel's works is that the breakdowns show that the measure of a team is how bad they are when they're bad, and the Sox haven't had any really bad stretches this year.

Edited by JimBoSox9, 20 July 2007 - 01:45 PM.

[ManilaSoxFAN]

By seeing what the record is at the end of the season. You can't predict a record based on how a team plays when they are playing .500. Look up the definition of average.
Oh, and frat didn't agree with you.

edit:added the frat thing

Actually, I thought frat was agreeing with:

"There is little correlation between a team's consistency and its ultimate record. Historically, some teams have managed to win a lot of (or few) games while being relatively consistent, while others have managed to win a lot of (or few) games while running hot and cold."

Which is an equivalent of:

"Looks like there is more ways to skin a 95-win cat...........This basic ratio won't change between 95-win types of seasons. It's just how the flat/growth is distributed."

BTW, a specific two-season, 95-win comparison followed...detailing the very specific differences in the manner each team achieved it's results.
-------------------------------------------------------------------------------------------------------------------------

Back to -- "So now the question is: How does one know if a team is playing a stint of .500 ball...which is somehow indicative of a 95-win season...or whether it's really just a part of an 85-win season? If so much of the time is spent in a "flat" mode, how can anyone ever recognize the indicators that the ability to enter a "growth" period are there?"

Your offering, absinthe was -- "By seeing what the record is at the end of the season. You can't predict a record based on how a team plays when they are playing .500. Look up the definition of average."

What if the said team had already rung up a +17 record? But was playing 500 ball for 50+days already? Could you predict an eventual outcome? If so...based on what, exactly? Good predictions need a factual base, no?

Also, are you saying that all .500 ball is the same? No diff tween going 5-5 at home vs KC, CHI and TB and going 5-5 in LAA, OAK and SEA? ?

[ManilaSoxFAN]

Not really. I don't see how beating the Royals and losing to the Tigers would be particularly more predictive than the other way around.

You are right -- one good weekend in one place and against one team...vs one bad one elsewhere...probably doesn't say much. And trying to be predictive is a often sucker's game.

That said, establishing a way in which one can measure oneself is a tried-and-true beneficial activity. Quality Circle/Deming Management and all that.

The idea wouldn't be to become accurately predictive; it'd be to keep tabs on how the performance was so far...in order to have a better understanding. One might even use some info gathered to make some modifications in behavior/programs.

Is that a waste of time?

[absintheofmalaise]

I was saying that you can't predict a W-L record based an a team playing .500 ball. No more and no less.