THE STREAKS ARE JUST CRAZY NOW THE WINS & LOSES Topic

dahsdebater
Posts: 10036 (0)
Hall of Famer
284 270 18
Pro
Which I find particularly ironic since, as best I recall, it was about 2 years ago that a thread like this last went 8 or 10 pages.
10/23/2010 2:36 AM
just4me
Posts: 4313 (0)
Hall of Famer
892 1 4 2 2
Hall of Famer
Contrarian23 ran a 24 identical team and settings (though he used 6 ballparks with 4 teams using each), well, I don't know what stats he still has access to, but here's some of his thoughts and stats:

Quote post by contrarian23 on 8/9/2009 7:56:00 AM:
In a separate thread ("Triples in Petco"), I've been discussing a league I recently created (MLB85209) to study the effect of parks on triples.The league is set up essentially as an OL with no AAA. I manage all of the teams. They have the exact same roster, managerial settings, lineup, rotation, advanced settings...basically EVERYTHING is exactly the same except the ballparks. Six parks are in use. Four teams each in AT&T, BOB, Comerica, Petco, Oriole Park IV, and WIS Park.The other thread has some early stats on how home park affects triples. But there are some other things that I think can be studied with this kind of league, including:-- How the performance of hitters from different eras (with different XBH norms) varies. Take 2 hitters with similar numbers of triples, one of whom comes from a high triples era and one from a low triples era. We would all expect the player from a low triples era to hit more triples in WIS - but how many more? Is the 3B/100PA# rating an accurate estimate?-- Same thing for pitchers.-- How much does the performance of a single player vary, both within ballparks and across ballparks. This last one is what I what I want to explore in this thread. It is my belief that simulation algorithms can produce enormous single-season variation in player performance.I will provide periodic updates, and (hopefully) complete end-of-season numbers on each of the players I'm using (except for the 200K scrubs).Here's an interesting example, through 91 games. One of the 4 starting pitchers is 64 Wally Bunker. For each team he has started anywhere between 21-24 games, and pitched 128 2/3 - 143 2/3 innings. For 21 of the 24 teams he has started exactly 23 games. As mentioned above, each team has the same rotation in the same order, but each pitcher is available to come in as an RP if needed. This accounts for the slight variation in # of starts. I believe (without carefully checking) that Bunker has never appeared in a game below 97 fatigue, so fatigue should not account for any of the variation seen below.Additionally, his MPC/TPC are set to account for both seasonal fatigue and in-game fatigue, so in-game fatigue should not be a factor either.Of the 24 Bunkers, the "best performer" is:11-8, 2.44 ERA, .196 OAV, 1.02 WHIP. (This Bunker plays in Petco, as perhaps expected.)The worst performer is tough to identify; there are a couple who are similar. I'll go with:9-8, 4.61 ERA, .287 OAV, 1.48 WHIP. (This Bunker plays in BOB, again perhaps as expected.)I've not yet calculated average performance, weighted average performance, standard deviation, etc....that can all wait for end of season. But I think it's interesting that you can observe radically different outcomes for the same pitcher. One can imagine a thread discussing Bunker with owner #1 saying "great bargain, best pitcher on my recent OL team" and owner #2 saying "not worth the money, unless maybe you're in a pitcher's park."That's variation across all parks. Lest anyone think this is entirely due to the difference between Petco and BOB, here's the worst performing Petco Bunker:6-8, 4.07 ERA, .256 OAV, 1.30 WHIP.Here's the best performing BOB Bunker:12-7, 2.53, .228 OAV, 1.10 WHIP.If ranked by OAV, the Bunkers who pitch in BOB rank 5th (.222), 7th (.228), 19th (.268), and 24th (.287).The Bunkers who pitch in Petco rank 1st (.196), 6th (.225), 14th (.251), and 16th (.256).In each case we're looking at an absolute performance differential of .060 to .065 on OAV.More to come on this topic as the season progresses.
Quote post by contrarian23 on 9/1/2009 2:11:00 PM:
I'll be updating this thread across the next few days...interestingly 2 of the 4 divisions were blowouts. A Comerica team won its division by 17 games (had a lead of 20-21 games for a while) and a Petco team won by 8 (after being up by 15+). The other 4 divisions were decided by 2 games or less, and 2 have ended in a tie after 162 games. Oriole Park (IV) and WIS Park will have playoff games to determine the division champs.More to come....
10/23/2010 3:28 AM (edited)
just4me
Posts: 4313 (0)
Hall of Famer
892 1 4 2 2
Hall of Famer
He had a thread with tons of data on the different variations (as I once did with my league as well), but it looks like his thread, like mine, has disappeared... Though, the following quotes from another thread on this topic from last year are probably relevant:
 
Quote post by tzentmeyer on 9/21/2009 9:58:00 AM:
 Quote: Originally posted by mudbone1969 on 9/20/2009

Maybe so, but I don't know all the particulars I guess.
In any case, it's real and don't let anyone tell you different.
 
 
This is most definitely NOT TRUE. There is no adjustment logic of any kind. 
Whether you're a new user, incredibly helpful veteran user or the largest troublemaker on our site, all simulations are straight up.

Quote post by zubinsum on 9/21/2009 11:26:00 AM:
Is there anything in the algorithms that adjusts players or teams up or down to simulate "hot" and "cold" streaks?

Quote post by tzentmeyer on 9/21/2009 3:53:00 PM:
No, there is not. But you'll "see it" from time to time, just like you'll see 7 straight coin flips turn up heads.
10/23/2010 3:22 AM
crazystengel
Posts: 11942 (0)
Hall of Famer
685
Hall of Famer
In WIS the very good teams win more than 7 out of 10 and the very bad teams fall short of 3 out out of 10, a sufficiently wide variance that in any given league, one or or more teams may be pushing real-world historical performance records.

Winning more than 7 out of 10 means 114+ wins; winning fewer than 3 out of 10 means 0-48 wins.  Other than prog leagues (where everyone's team salary is different) I almost never see teams with that many or that few wins.

I've kept pretty detailed records of a league I've run for 13 seasons (Random Captain).  Of the 312 teams that have played in that league, none has won 7 out of 10 games over a full season.  Only one had a winning % under .300 (.296, to be exact).  By your definition, that league's had no very good teams and one very bad team in 13 seasons.

Take a look at everyone's profile.  How many owners do you see whose winning % is less than .400 or greater than .600? 

10/23/2010 4:59 AM (edited)
crazystengel
Posts: 11942 (0)
Hall of Famer
685
Hall of Famer
Posted by amycox67 on 10/23/2010 6:07:00 AM (view original):
And may I ask this Crazy would you or anyone one get more better owners in a theme league then you would in a open league because I just looked at some league I was in that closed in the last month 2 had 2 teams & 1 with 1 team that win 38 games or less ( all rookie players) and 4 team with 115 + wins ( All HOF players)
 
amy, I've only played 2 OLs in the past 2 years so I'm not the guy to ask about OLs vs. themes.  But I imagine that you would find more completely hapless owners in OLs than in other leagues,and that would skewer the records for everyone (and of course lead to more "streakiness").

10/23/2010 6:44 AM
napolean
Posts: 666 (0)
Hall of Famer
587 7 184
Rookie
You're a conspiracy theorist or you're not, you're paranoid or you're not.  After reading all the posts and all the info some people took a LOT of time to compile and post, (and we owe them a thank you) I think what we have is simple probabilities expressing themselves in strange winning and losing streaks over the millions of games played on this site. I therefore now believe the game is straight and no one is monkeying around with it. Let's take our knocks and luck like men and roll on. After all who could blow a 14 game league in August and not win the Pennant?  The 1978 Red Sox that's who.... Who could be up 2 runs with 2 outs, no one on base in the 9th INNING to WIN the WORLD SERIES and lose, the 1986 Red Sox that's who, .....(I'll never get over it). So much for shoulda, woulda, coulda..but probabilities suggest it CAN happen.
10/23/2010 1:01 PM
jvt87
Posts: 268 (0)
Hall of Famer
237 2 5
Prospect
Posted by crazystengel on 10/23/2010 4:59:00 AM (view original):
In WIS the very good teams win more than 7 out of 10 and the very bad teams fall short of 3 out out of 10, a sufficiently wide variance that in any given league, one or or more teams may be pushing real-world historical performance records.

Winning more than 7 out of 10 means 114+ wins; winning fewer than 3 out of 10 means 0-48 wins.  Other than prog leagues (where everyone's team salary is different) I almost never see teams with that many or that few wins.

I've kept pretty detailed records of a league I've run for 13 seasons (Random Captain).  Of the 312 teams that have played in that league, none has won 7 out of 10 games over a full season.  Only one had a winning % under .300 (.296, to be exact).  By your definition, that league's had no very good teams and one very bad team in 13 seasons.

Take a look at everyone's profile.  How many owners do you see whose winning % is less than .400 or greater than .600? 

It's hardly necessary to point out the logical error in extrapolating from "leagues I've run," with a limited number of owners pursuing a particular theme, to the larger WIS universe. Nor the basic mathematical flaw in assuming that a regression to mean of .400-.600, among those customers who regularly buy WIS products means none of their teams, or those of one-time or occasional users, surpass or fall short of those numbers.

Again, none of this answers Amy's original question about observed streakiness now compared to two years ago. The answer to that is yes, anecdotally.
I've certainly observed that, as well as comments by other plays complaining about the phenomenon. But that's strictly anecdotal, we lack sufficient and sufficiently rigorous data to reach a definite conclusion. Despite the best efforts of some customers, there just aren't enough numbers available yet to say, although presumably administration could provide more.

In turn, that would not answer the question of whether increased streakiness is a bad thing. Two other alternatives have been offered: that WIS would never do such a thing, or that streakiness is a normal part of real baseball. I tend to agree with the latter, but that does not bring us any closer to a comparison between WIS streakiness and real-life streakiness.

The flaws in the arguments advanced so far in defense are that it's "random" and "anything can happen." But of course, if the simulation is in any way accurate, it can't. This is not a random universe. The data sets drawn from historical baseball remove that element. We don't need to approximate the performance of the 1916 Giants, we know exactly what they did. Yes, running their data sets against those of every other professional team greatly increases the complexity and the variability of the possible outcomes. But "anything" is not possible. To the extent the algorithms allow for completely random outcomes, they would fail to model the results of real-life baseball adequately.

That's what is suggested by some of the admittedly limited data presented here. If a sampling of a "Tim Jordan" data set, run against the same opposition under the same circumstances, produces outcomes ranging from 51 to 121 runs, it suggests significant problems with the modeling. And also that, from a consumer perspective, among games of chance the odds are significantly better putting money on "black" than investing in the WIS "Tim Jordan."





10/24/2010 2:52 PM
skunk206
Posts: 11011 (0)
Hall of Famer
1248
Hall of Famer
10/24/2010 3:59 PM
crazystengel
Posts: 11942 (0)
Hall of Famer
685
Hall of Famer
The flaws in the arguments advanced so far in defense are that it's "random" and "anything can happen." But of course, if the simulation is in any way accurate, it can't. This is not a random universe.

As I understand it, there is a random element in literally every event.  There's no way to avoid that -- if 1985 Willie McGee is facing 1994 Greg Maddux in an at-bat, how do you determine a result without using an element of chance? 

That doesn't mean that "anything" can happen, but there are a wide variety of things that could happen, and some are more and some are less likely.  And sometimes the things that are less likely to happen happen more than they "should" because each event is treated as a unique occurrence which is not influenced by what happened before.  So there will be times when a 1988 Milacki gives up 4 straight hits to mediocre hitters and blows a save; and times when a 2006 Josh Towers strikes out 3 straight Hall of Famers with the bases loaded.  That doesn't mean there's something wrong with the engine, necessarily, or that a player has been "hot" or "cold" coded.

As for how differently one Tim Jordan performed in the same conditions, I don't think even a full season of plate appearances can be considered a large sample size.  If you could set up a season to last not 162 games but 162 million games, I'd be willing to bet those 24 Tim Jordans would wind up with virtually identical stats. 
10/24/2010 4:45 PM
just4me
Posts: 4313 (0)
Hall of Famer
892 1 4 2 2
Hall of Famer
Well, to answer the original question would be entirely impossible unless someone has data from both two years ago and from now to show the levels of streakiness then and now. From a purely subjective point of view though I can say that my observation show no more streakiness now than at any other point and the fact that people have been complaining of increased streakiness for at least the last 3-4 years would indicate that people just complain about observed streaks (confirmation bias) than any actual increase in streakiness (i.e., no more streaky now than 2 years ago). 

I believe the examples from RL baseball weren't necessarily used as an alternative theory, but as an example of how streaks are a normal part of any statistical venture - with baseball being used since this is a baseball simulation. As for the speculation of whether WIS would or wouldn't do such a thing, the fact that they've been flat out denying it for at least 3-4 years would indicate that no such thing exists (especially since other simulation do such a thing and advertise such a streak algorithm as a feature). 

The arguments aren't saying it's "random" or "anything can happen" quite in the way that you are suggesting. I would argue that it is random and that anything can happen, but this is also bound by reasonable statistical variance. 

As for the variance in the Tim Jordans; first off, runs are a horrible stat to use as an example as that is team dependent, and relies on multiple independent variables all producing -- or not producing -- in succession. If you want to use the Tim Jordan example above I would focus on one of his rate stats, and likewise for the Webb  example. 

In running (through random.org) twenty 500 AB seasons of 1906 Jordan vs. 2006 Webb, for AVG the twenty seasons together (10,000 ABs) yield an AVG of .2566 (where the Log5 for this matchup yields a .2533). Yet, in individual "seasons (500 AB)" Jordan hit as low as .212 (twice) and as high as .298. That's a variance of .086 from low to high, or .34 and .32, respectively, from the average yield (a variation of within 13% of the average). Eight of the twenty seasons finished with an AVG of between .266 and .272 (a difference of 3 hits), and all but the three outlier seasons mentioned above (85% of the twenty seasons) finished between .244 and .272 (14 hits - and a 5.8% spread from the average), and yet the hit spread from the worst to the best season was 43 hits. 

As far as streaks go, three seasons in row had Jordan hitting exactly .268 (134 hits).  He had only 5+ hit streaks 13 times, twice in his best season, and his best streak of 8 straight hits occurred in one of his two worst seasons. 

That's all with the same hitter against the same pitcher over and over again, and in just one stat category. As you introduce more categories, and more pitchers, plus ballpark effects and fatigue, and then manager settings, it's only going to exacerbate the effects and the variations in single player performance. It also increases the possibility of streaks as you go from a less probable occurrence (like a coin flip) to a more probable occurrence (like an out in a pitchers park against a good pitcher, or a hit in Coors against Denny Stark).

So, anything can happen, and it certainly is random, but it's not totally up to chance and there are limits, but even the unlikely is still possible. 

A great analysis on streaks (hitting streaks, in particular) was done a year or so ago, let me see if I can find it....






10/24/2010 5:34 PM
just4me
Posts: 4313 (0)
Hall of Famer
892 1 4 2 2
Hall of Famer
The link to the article on hitting streaks... not quite what I remembered, but very useful nonetheless.
10/24/2010 5:50 PM
napolean
Posts: 666 (0)
Hall of Famer
587 7 184
Rookie
Outstanding article just4me.......
10/25/2010 10:36 AM
believeit
Posts: 234 (0)
Hall of Famer
183
Prospect
Amy, that sounds like a proposal.  (For what?  Only Grizzly knows.) 

10/27/2010 3:05 AM
This post has a rating of , which is below the default threshold.
Show
This post has a rating of , which is below the default threshold.
Show
◂ Prev 1234 Next ▸
THE STREAKS ARE JUST CRAZY NOW THE WINS & LOSES Topic
△Top△Bottom

Search Criteria

Terms of Use Customer Support Privacy Statement

© 1999-2024 WhatIfSports.com, Inc. All rights reserved. WhatIfSports is a trademark of WhatIfSports.com, Inc. SimLeague, SimMatchup and iSimNow are trademarks or registered trademarks of Electronic Arts, Inc. Used under license. The names of actual companies and products mentioned herein may be the trademarks of their respective owners.