The "Win Window"

[Young Drachma]

I was riffing earlier about the concept of something called a "win window." What is it? I'm not sure yet.

Here's how I came up with the thought.

My FOOL ballclub was an expansion team. I took a strategy pretty early on that would be a team that would be able to compete for a title from the outset pretty much. And short of a successful first half of that first season, we've been evading it ever since. A 3rd place finish, a 5th place and 4th place finish respectively have been the story of the day for us over the past three years.

I've felt like we're always just a player or two away. I recently flubbed up a free agent signing that was supposed to be the masterstroke in an otherwise brilliant off-season. Now, I'm looking at things and despite all I gave up (prospects for acquiring two pitchers to bolster my rotation and bullpen) I'm still just "one or two players" away.

I've realized in other online leagues too, that the problem is largely related to the inorganic nature of the online league trade market. You only have a few teams at any one time that are considered actually bothering to make any deals. Then you have inactive owners who sit on good to great talent.

Anyway, the point of this is, there has to be a way to ineffectively measure how much a particular player will "contribute" to a team that's looking to add some value to the ballclub.

I think VORP might be the easiest way to do it.

What are the thoughts on this?

[Alan T]

Originally Posted by Dark Cloud

I was riffing earlier about the concept of something called a "win window." What is it? I'm not sure yet.

Here's how I came up with the thought.

My FOOL ballclub was an expansion team. I took a strategy pretty early on that would be a team that would be able to compete for a title from the outset pretty much. And short of a successful first half of that first season, we've been evading it ever since. A 3rd place finish, a 5th place and 4th place finish respectively have been the story of the day for us over the past three years.

I've felt like we're always just a player or two away. I recently flubbed up a free agent signing that was supposed to be the masterstroke in an otherwise brilliant off-season. Now, I'm looking at things and despite all I gave up (prospects for acquiring two pitchers to bolster my rotation and bullpen) I'm still just "one or two players" away.

I've realized in other online leagues too, that the problem is largely related to the inorganic nature of the online league trade market. You only have a few teams at any one time that are considered actually bothering to make any deals. Then you have inactive owners who sit on good to great talent.

Anyway, the point of this is, there has to be a way to ineffectively measure how much a particular player will "contribute" to a team that's looking to add some value to the ballclub.

I think VORP might be the easiest way to do it.

What are the thoughts on this?

Isn't that what "Win Shares" was supposed to be calculated to try to measure?

[Young Drachma]

Win Shares does this:

Win shares are intended to represent player value (what they were responsible for) rather than player ability (what the player's true skill level is).


Whereas my "Win Window" is a team statistic aimed at evaluating a team's talent to determine whether the team really is a contender or a pretender. It's a fictional statistics for fictional games, as I doubt you could use it in a real life context given the variables that apply to the real world that don't apply to the fictional world.

[Alan T]

Originally Posted by Dark Cloud

Win Shares does this:

Win shares are intended to represent player value (what they were responsible for) rather than player ability (what the player's true skill level is).


Whereas my "Win Window" is a team statistic aimed at evaluating a team's talent to determine whether the team really is a contender or a pretender. It's a fictional statistics for fictional games, as I doubt you could use it in a real life context given the variables that apply to the real world that don't apply to the fictional world.

Hmm.. well theoretically if you believe that "Win Shares" is a definition of how many wins a player theoretically adds to a team, you should be able to work together and figure out how many "Win Shares" that were on the entire team for a cumulative "Win Shares" that theoretically should be pretty close to how many wins the actual team did win.

So then you can try to take that a step further, and figure out a way to forecast the "Win Shares" for each individual for a forecasted Win Shares of some sort. This is somewhat difficult as Win Shares is designed to be a reporting statistic and not a predictive statistic however. (Similar to your suggestion of VORP however). Perhaps you could trend a player's Win Share and then have comparison win shares for similar players of age and ability and create a forecasted win shares that way. Then add all of the forecasted Win shares for every individual on the team to create a total team forecasted win shares.

Then it is a case of determining if your team's entire total forecasted win shares ends up being in the range of where you feel a contending team should be or not to determine if they are in their win window...

This is probably alot of work and may not even be accurate however without testing or research, but it is the only suggestion I could think of.

[Young Drachma]

Originally Posted by Alan T

Hmm.. well theoretically if you believe that "Win Shares" is a definition of how many wins a player theoretically adds to a team, you should be able to work together and figure out how many "Win Shares" that were on the entire team for a cumulative "Win Shares" that theoretically should be pretty close to how many wins the actual team did win.

So then you can try to take that a step further, and figure out a way to forecast the "Win Shares" for each individual for a forecasted Win Shares of some sort. This is somewhat difficult as Win Shares is designed to be a reporting statistic and not a predictive statistic however. (Similar to your suggestion of VORP however). Perhaps you could trend a player's Win Share and then have comparison win shares for similar players of age and ability and create a forecasted win shares that way. Then add all of the forecasted Win shares for every individual on the team to create a total team forecasted win shares.

Then it is a case of determining if your team's entire total forecasted win shares ends up being in the range of where you feel a contending team should be or not to determine if they are in their win window...

This is probably alot of work and may not even be accurate however without testing or research, but it is the only suggestion I could think of.

Well, you're certainly giving me stuff to think about. I'm going to generate some data in-game and see what I can mess around with. I never do download the CSVs, but I think I will so I can take a look at stuff more detailed, since I'm super curious about some things.

[ScottVib]

However... (and I could be wrong), couldn't using Win Shares to evaluate the teams "Win Window" be a bit incestuous? After all, isn't the number of total win shares assigned to a teams players is simply 3x the number of wins the team earned. The only reason for this to be different is roundoff error, or players who spent part of a season with one team and part with your team.

[Young Drachma]

Amazon.com: Diamond Dollars: The Economics of Winning in Baseball: Vince Gennaro: Books

Has a ton of theories and such, mostly about real baseball that I could, were I feeling ambitious come up with theories for fake baseball (and by proxy, could be used for real baseball) and statistical methods that measure a player's value.

They're not even THAT complicated when he breaks it all down. It's just taking the time to 1) invest in developing a measure and 2) creating a calculator for people to use it and actually find value in it.

[Young Drachma]

I've noticed that winning teams minimize the number of negative VORP guys they have. I haven't figured out what sort of index I want to create yet, mostly because I don't feel like hand calculating a lot of stats, but...that's the thing I'm seeing universally as I sim in my test league./

[Young Drachma]

Win Window, Redux

I've figured it out. OOTP calculated the total VORP for batting and pitching of each team in the history section, which makes all of this work.

So the new stat is called VOCL or Value Over Championship Level. (Pronounced VO-CAL) It's not that beautiful of a stat, but it works. It's only measure is to help you determine whether your team is at championship level. Doesn't mean you're anymore likely to win or anything like that.

It just means that your team is officially within its "win window" I've seen teams win at a lower rate than championship level. It's just a helpful tool.

The thing is, in each league, championship level is different. In some leagues, you can get as high as 600 for some years, in playoff leagues, it can be as low as 450. I've seen VOCLs into the 800s on occasion.

We're not trying to calculate how low a team can go to get to championship level, instead, the goal is to try to figure out how to adjudge when your team has "made it" and so, the formula for VOCL is:

(Team Batting VORP + Team Pitching VORP) - 450 = VOCL

I'll keep tweaking it, as well as trying to integrate other measures into it, but...this is a good start.

[Young Drachma]

I just ran a test in OOTP9 using an MLB standard league.

The VOCL for an MLB league is lower than a fictional league. Doesn't surprise me, because the league VOCL is going to really be determined by the talent level that you're playing against, etc.

So we need to figure out how to get the VOCL modifier for each league. That 450 number in the test equation I ran was based on fictional online leagues that I'd done some research in.

After looking at some MLB numbers, I think the number needs to come down to about 400. That's about the threshold for a 100 win team in the majors. Now there are going to be playoff teams and thus, champions with a VOCL in the 300s. It's just inevitable because divisional play waters things down considerably. But the measure is still effective.

So I've decided to tweak things a bit and create a new statistic that feeds off the VOCL, because it makes more sense to me and is a better measure of what this whole stat was supposed to be.

The VOCL modifier is going to be lowered to 400.

Here are the factors for our new stat calculating the win window.

To determine a team's Win Window, you calculate the VOCL + Win Differential (W - L) / Number of Games in the season = Win Window

[Alan T]

Looking at what you are trying to do here with "VOCL" to compute how close someone is to a "win window", I think you were going on the correct path in understanding how a lower VOCL can win a championship in MLB play. I think that is mostly due to the higher percentage chance of getting into the post season.

In MLB, you have 8 out of 30 teams or roughly 27% of them make the postseason, but your fictional online league that you used only 2 of 16 teams make the post season or roughly 12-13%.

With the randomness in both baseball in general as well as OOTP, even the worst team in all of baseball is going to beat the best team in all of baseball on average at least once or twice in a 10 game span. That means the more rounds that are inserted into the playoffs, the more likely you will have a team "overachieve" or pull off an upset.

It is probably far more likely that a 2 team playoff that uses 100+ games to decide whom those two teams are will see a better team as the league champion than an 8 team playoff system due to that.

So that makes the required VOCL level needed to possibly win the championship different based on how easy it is to make the playoffs. I would wager even if you changed your fictional league test that you did to include more teams in playoffs, it would have the effect of lowering the average VOCL of the league champions despite the Win differential or number of games in the season remaining fairly constant.

[Young Drachma]

Well I used FOBL too. That league has playoffs similar to MLB.

And then I used an MLB league, as well, to do testing. That's why I lowered the threshold. I looked at all of the playoff teams and the numbers were relatively consistent. But what I was trying to do was concoct what makes a "better" team versus not. That said, you make a good point that playoffs make the threshold lower and because of that, it'd be wiser to consider lowering the number. But I didn't want to make the number so low that it was important to focus on JUST making the playoffs, but maybe I ought to consider that, since well...if you make the playoffs, you can win and that's what we're trying to measure.

Right now, it measures a team that's pretty much title worthy at the level of 90+ wins or so. But using the VOCL as the championship level threshold and the win window as the measure for how close your team is to being "in the hunt" of things, makes it work pretty well.

Thanks for checking it out.

[Alan T]

Originally Posted by Dark Cloud

Well I used FOBL too. That league has playoffs similar to MLB.

And then I used an MLB league, as well, to do testing. That's why I lowered the threshold. I looked at all of the playoff teams and the numbers were relatively consistent. But what I was trying to do was concoct what makes a "better" team versus not. That said, you make a good point that playoffs make the threshold lower and because of that, it'd be wiser to consider lowering the number. But I didn't want to make the number so low that it was important to focus on JUST making the playoffs, but maybe I ought to consider that, since well...if you make the playoffs, you can win and that's what we're trying to measure.

Thanks for checking it out.

Yeah, I don't think my argument is necessarily as true in other sports such as say Hockey or Basketball which includes a fairly large amount of teams. In those sports it just seems less likely to have massive upsets round after round to a championship. In baseball though there just seems to me to be a much higher chance of a "worse" team beating a "better" team due to hot streaks, randomness or who knows what though.

I haven't honestly done research on this, and may just be assuming things incorrectly, but I wonder since football and baseball went to a wildcard format, what the win% is for the team with the better record in the playoffs for Hockey, Football, Basketball and Baseball.

[Young Drachma]

Yeah, I think this is totally a baseball thing. I don't think it'd work in other sports, because the threshold is so low to make the playoffs and the barriers to entry for bad teams winning it all is..well...not as easy as it is for a hot baseball team to get going.

[Young Drachma]

Predicting the Playoffs

After the 2003 regular season ended, the time before the divisional series was filled by "experts" forecasting the outcome of the four divisional series. This phenomenon will be repeated before the League Championship Series, and again before the World Series. These same pundits will look back after each series to pat themselves on the back, make excuses or explain how they went wrong. They believe, or at least pretend, that postseason results can be accurately predicted.

Others believe that the postseason is essentially a crapshoot, that any club can win a succession of short series among eight clubs which all finished within 10-15 games of one another during the regular season. This group includes Billy Beane, quoted in Moneyball as saying: "My s*** doesn't work in the playoffs. My job is to get us to the playoffs. What happens after that is f****** luck." Those in the first group have criticized Beane's Oakland A's and Bobby Cox's Atlanta Braves as teams that "can't win the big ones"; those in the second think "clutch postseason performance" is as real as "clutch hitting," or the Easter Bunny.

Who's right? Let's look at the past century of postseason play. Since 1903, there have been exactly 200 postseason championship series of best-of-five or longer. This includes 94 best-of-seven World Series, four best-of-nine World Series (1903, 1919-21), 34 best-of-seven League Championship Series (LCS), 32 best-of-five LCS, 32 best-of-five divisional series, and four best-of-five divisional playoff series following the 1981 strike-induced split season. That's a sizable data set.

Absent an easily accessible, generally accepted way to compare clubs in different leagues all the way back to the deadball era, I'll define the "better team" as the one with the better regular season record. Four of these 200 series matched clubs with identical records, leaving 196 with a "better team" that, at least in theory, "should have won." The better team has won 106 of these series and lost 90, for a .541 winning percentage. These 196 series break down as follows:

World Series: Better teams are 52-44 (.542), with two evenly matched Series.
Seven-game LCS: Better teams are 19-14 (.576), with one evenly matched Series.
All best-of-seven or -nine series: Better teams are 71-58 (.550).
Five-game LCS and divisional series: Better teams are 35-32 (.532), with one evenly matched series.

As the tables below show, the better club is more likely to win blowout series, less likely to win those decided by a single game. Better clubs lose more deciding fifth or seventh games than they win.


Results for Team with Better Record, World Series
Win by 4 (10): 1907 1927 1928 1932 1939 1950 1966 1976
1989 1998
Win by 3 (14): 1905 1908 1910 1915 1920 1929 1937 1938
1941 1942 1961 1970 1983 1984
Win by 2 (13): 1903 1911 1917 1919* 1923 1930 1936 1944
1948 1951 1977 1978 1981
Win by 1 (15): 1909 1912 1925 1940 1947 1955 1956 1967
1968 1973 1975 1986 1991 1997 2002

Lose by 1(19): 1924 1926 1931 1934 1945 1946 1952 1957
1960 1962 1964 1965 1971 1972 1979 1982
1985 1987 2001
Lose by 2(11): 1906 1918 1921 1935 1953 1959 1980 1992
1993 1995 1996
Lose by 3 (8): 1913 1916 1933 1943 1969 1974 1988 2000
Lose by 4 (6): 1914 1922 1954 1963 1990 1999

Note: In 1949 and 1958, the World Series opponents had the same regular season record.
* - The 1919 Reds may well have won even a non-fixed World Series. Their .686 regular-season winning percentage is the NL's second highest since 1910.

The 4-0 losses include three of the biggest World Series surprises: the 1914 Braves over Philadelphia, Dusty Rhodes and the 1954 Giants over the 111-43 Indians, and the 1990 Reds sweep of Oakland, which I won't discuss further because a crying Gary Huckabay is not a pretty sight. The four recognized "miracle" pennant winners, the 1914 Braves, 1961 Reds, 1967 Red Sox and 1969 Mets, went 2-2 in the World Series, suggesting that "magic" loses its strength when confronted by Bob Gibson or Mickey Mantle/Roger Maris.

Note that the 1999 and 2000 Yankees, Exhibit A to Commissioner Selig's competitive balance laments, both beat clubs with better regular-season records. In fact, the 2000 Yankees had only the ninth-best record in the majors, fifth-best in the American League. That's the same as the 1987 Twins, who won the World Series despite being outscored during the regular season. At the other end of the spectrum, the 4-0 wins include six Yankee demolitions of inferior opponents, and one humbling of the Yankees at the hands of the Big Red Machine.


Results for Team with Better Record, Best-of-7 LCS
Win by 4 (3): 1988A 1990A 1995N
Win by 3 (6): 1989A 1991A 1996A 1999A 2001N 2002A
Win by 2 (6): 1985N 1986N 1993N 1995N 1998A 1999N
Win by 1 (4): 1986A 1987N 1992N 1996N

Lose by 1 (3): 1985A 1988N 1991N
Lose by 2 (6): 1990N 1993N 1997A 1997N 1998N 2000A
Lose by 3 (5): 1987A 1989N 2000N 2001A 2002N
Lose by 4 (0):

Note: In 1992, both ALCS opponents had the same record.

Compared to the World Series, the seven-game LCS is significantly less likely to end up as a four-game sweep--but in non-swept Series, the club with the better record holds only a 16-14 edge. The big surprises here include easy wins by the 1987 Twins over the Tigers and the 2001 Yankees over the 116-46 Mariners. The late-80s Oakland juggernaut barely broke a sweat, going 12-1 in three consecutive LCS.


Results for Team with Better Record, Best-of-5 LCS and Div Series
Win by 3 (19): 1969A 1969N 1970A 1970N 1971A 1975N 1976N
1979N 1981AE 1982N 1984A 1995A1 1995N2 1996N1
1997N1 1997N2 1998A1 1998N1 1999A1
Win by 2 (9): 1971N 1974N 1978A 1978N 1979A 1995N1 1996A1
1997A1 1999N2
Win by 1 (7): 1972A 1976A 1981NE 1981NW 1981N 1982A 2001A1

Lose by 1(15): 1972N 1973A 1973N 1977A 1980N 1981AE 1984N
1995A2 1997A2 1999A2 2000A1 2001A2 2001N2 2002A1
2002N2
Lose by 2(10): 1974A 1977N 1983A 1983N 1996A2 1998A2 1998N2
1999N1 2000N2 2002A2
Lose by 3 (7): 1975A 1980N 1981A 1996N2 2000A2 2001N1 2002N1

Note: In 2000, one of the NL divisional series matched opponents with identical records.

In all best-of-five postseason series, the team with the better record is 19-7 in sweeps, 16-25 in 3-1 or 3-2 series. In divisional series since 1995, the team with the better record is just 13-18. Since 2000, 10 of 11 divisional series have been won by the team with the worse record.

Does this foretell wins by the Red Sox, Twins, Marlins and Cubs? No. Overall, as one would expect from series between relatively evenly matched teams, the teams with the better record hold a small but consistent edge. But while the playoffs aren't quite random, when the better team can be expected to win just 54% of the time, anyone who claims to know what will happen is only fooling himself.

[Alan T]

On the same subject (of the chance of upsets in the playoffs), I did a web search and came across this interesting article that I think pretty much supports the point being made.. The more teams in the playoffs, the harder it is for the best team to win, even when a much more superior team than others.

Diamond Mind Baseball

[Young Drachma]

I think that I'd add "especially in baseball."

I don't know what about the game that lends itself to that, but...it seems like that'd be the case far more in baseball than in other sports, where playoffs actually make some sense. It could be the excessively long regular season.

[Alan T]

I think it is because of the game and not the length of the season or anything else.

In this game, if you do things at a very low percentage of success, you are considered great.

If you get a hit 33% of the time, you are one of the best contact hitters in the game.

If you throw out a base stealer 30% of the time, you have a good arm at catcher

If you hit a home run 15% of the time, you might end up in the hall of fame.

That doesn't carry over to other sports, where you have to be a 75% free throw shooter, or you need to hit at least 60% of your shots from the court in basketball.. or in football where you need to complete at least 50-60% of your passes to keep a QB job, or connect on 85% or more of your field goal attempts.

The difference between hitting .320 and .270 for a season is what? 1 hit a week, or maybe 5 hits a month difference between two players? I think when dealing with such small differences meaning such a great difference of perception, it just lends to a game where upsets are more frequent..

I could be way off base though as I don't really have much statistical proof here on this, just trying to figure it out in my head.

[Young Drachma]

That was a really incisive article.

[Young Drachma]

Here's a spreadsheet that will calculate VOCL and Win Window for teams. WW is still an ugly stat, but I dunno what to do it to make it "prettier" yet. Or what to add to it to make it look better.

It seems to be effective, though and that's all I care about.

It's based on FOOL at the moment, but you can totally adapt it for another league.

[Commo_Soldier]

I think you are on the right track as well. I personally feel that the formula may be a better predictor if it could say take the past two years of results at a 25% and 35% weight each for the VORP's and take a predicted batting and pitching VORP for the remaining 40%. Here are the following reasons I beleive this would be need to be factored in.

1. A team will most likely build up to success in the past two years or rebound from an off year so I think more than one season's of results is needed.
2. Conversely one must not look to far back at past success as it is hard to sustain a championship level of success.
3. Predicted VORP's (would have to be based on an accurate forumla) is needed on the basis that turnover happens. Teams that had success the past few years will drop off because of aging, losing key FA's. Where as teams that have been losing may jump up because of young talent peaking and gaining key FA's.

Overall as stated I think this is a good place to start on if a team has a reasonable shot at competing in the upcoming year.

[Young Drachma]

Predicted VORP wouldn't really work since PECOTA isn't really in the game, it's just projected as the season goes along, which doesn't help for something like this. It would, however, make it a stat that indeed has some real world value. W00t. Awesome.

It's not really intended to be a predictive statistic, it's meant to be an assessment tool of where you were with what you had and to tell you cereris paribus how your team would be able to perform the following year, were you to add more VORP and subsequently more wins.

That said, I like the idea of weighting the past two years of results. Maybe even going to three years, going 15, 25, 60 for them in order from 2 yrs ago to present.

[Young Drachma]

Originally Posted by Commo_Soldier

I think you are on the right track as well. I personally feel that the formula may be a better predictor if it could say take the past two years of results at a 25% and 35% weight each for the VORP's and take a predicted batting and pitching VORP for the remaining 40%. Here are the following reasons I beleive this would be need to be factored in.

1. A team will most likely build up to success in the past two years or rebound from an off year so I think more than one season's of results is needed.
2. Conversely one must not look to far back at past success as it is hard to sustain a championship level of success.
3. Predicted VORP's (would have to be based on an accurate forumla) is needed on the basis that turnover happens. Teams that had success the past few years will drop off because of aging, losing key FA's. Where as teams that have been losing may jump up because of young talent peaking and gaining key FA's.

Overall as stated I think this is a good place to start on if a team has a reasonable shot at competing in the upcoming year.

I've created a stat here called the Right Track Index (RTI). RTI takes a team's VOCL over a three year period, weighting the most recent year higher, the next year less and the 3rd year less than that (55-30-15) to evaluate whether a team is on the 'right track' or not. The higher the number, the more 'on track' they are.

Our numbers are skewed in the sense that Valdosta has won three titles in a row and are clearly a juggernaut for the ages.

But that said...

Here's the numbers for the most recent 3 seasons of FOOL.

Team			RTI
Valdosta Peanuts	667.2
Colorado Rancheros	90.7
Boston Settlers		-114.6
Brooklyn Brownstones	66.7
Rio Grande Roadrunners	66.3
Atlanta Firecrackers	-33.5
Quad City Thunderstorms	-342.7
Compton Brothers	-23.1
Ann Arbor Wolverines	-95.0
Baltimore Terrapins	-440.7
Toronto Atlantics	-564.9
New York Highlanders	-557.2
Hartford Harpooners	-731.7
El Paso Alligators	-721.9
Columbus Crusaders	-727.4
Texas Copperheads	-901.0

RTI tells us that at the end of 1972, Valdosta is clearly on a track all their own. This measure would be coupled with a RTI Chg that would reflect from year to year how much a team gained or lost in RTI, to really be able to track their performance over a particular period of time.

For me personally, it tells me that even though I know my team is getting out of a crazy situation, having just had our first two winning seasons in franchise history in the last two years, that we've still got a lot of work to do to be amongst the contenders like Colorado, Rio Grande, Brooklyn and Valdosta, all teams that have won titles in recent years.

I'm attached a spreadsheet that calculates this, along with Retained and Predictive Value. To refresh those are:

RVe (Retained Value) is the amount of VORP a team gained/lost during an off-season.

PVe (Predicted Value) is the measure of the team’s VOCL from the previous season added to their RVe.

[Alan T]

DC, The last three seasons probably aren't the best to try to test this with index with simply because there has been very little change at the top of the standings over the past three seasons.

What would be really interesting is to measure the 1965-1969 time period perhaps, as there was a different champion each year, and you saw the start of the Valdosta dynasty during this time. It would be interesting to see what happened post St.Louis being folded and watch the growth of Colorado, Valdosta, Ann Arbor, and a few others during this time and see what your index tells us about these teams.

[Young Drachma]

Good call. I'll take a look at that and see what we have.