This Has My Attention: New Questions

I’m simply asking the question “Why would this be occurring?”

Three somewhat related reasons:
1. Strat on line is not real life
2. The erudite professors you quote could be full of bunkum (N.B. I had to look up what a Weibull distribution is - maybe Frankie can explain to me why its more appropriate than a Poisson or Gaussian)
3. This is a game of chance and the pythagorean exponents may be very different than what's being assumed.

Also, a question for the usual suspects who always show up in the conspiracy theory threads - is this really about a conspiracy theory or victimization?

Bc it clearly has nothing to do with wanting to game to function in a more consistent and logical fashion, yeah?
Wondering if one could’ve come up with a more misrepresentative question…

gk: I don’t think your use of the word “chance” is correct. Chance is simply the possibility of something happening. The game is statistically accurate, or at least as accurate as a simulation can be. The “chance” of a walk off home run in a real game is the same as the “chance” of rolling a walk off home run with dice. Whether you’re facing a Hall of Fame pitcher in real life or rolling dice is irrelevant. They’re both just statistical possibilities.

Further, although the Wins and Losses were perceived as reversed (a Loss was better, ie a “Win”), the stats in the Upside Down league were not reversed. Teams that had actually won more still tended to underperform their expected wins/run differentials, exactly as I have been trying to point out.

Lastly, this tendency for high run diff teams to underperform expectations is a recent phenomenon, at least as far as I can tell. When I look back at my older leagues, this was not the case. High run diff teams had (generally) the same wins as expected. These 10, 11, 12 game outliers were no where to be found.

dj: Whether it’s a real life game or a simulation, math is math. Yes, I guess that everyone in Sabermetrics might be wrong is a possibility. You can optimize the exponents (for the Strat 365 run scoring environment) to improve accuracy - but the general distributions would not change. Positive run diff teams should overperform expected wins, not the other way around.

Instead of all this blather, let’s do this.

Can someone show me a league where the highest positive run diff team OUTPERFORMED expectations by 10 games? Find me a well chosen example. We’ve already cited plenty here demonstrating exactly the opposite. (A recent team please, not a 2004 team.)

dj: Whether it’s a real life game or a simulation, math is math. Yes, I guess that everyone in Sabermetrics might be wrong is a possibility. You can optimize the exponents (for the Strat 365 run scoring environment) to improve accuracy - but the general distributions would not change. Positive run diff teams should overperform expected wins, not the other way around.

J-Pav, yes math is math, but you do not understand it. A smaller exponent would provide a smaller number of expected wins based on the same run differential.

Also, real life managers don't allow tired relievers to give up 13 runs in an inning. Since HAL does, would you admit there might be a difference in the run differentials for good teams in the on line game and in real life?

Sorry for the "blather"

Do only bad teams have tired relievers? What are you even talking about?

Beginning to have a headache after trying to read and understand all this. Yikes!

That genuinely interesting things can devolve so quickly has always been a downside of these forums.

That genuinely interesting things can devolve so quickly has always been a downside of these forums.

sigh...
one of many reasons I dont spend much time on the forums anymore, bc it
might be my fault.
some folks (not everyone) get offended by the assertion that there might possibly be something in the game engine that they are unaware of-
and apparently if I say it, then for sure, it will bring out some animus.
One of many reasons I miss Bruce, as well as, seeing Nev here.

My 95 completed teams underperform pythag by an average of 3.5 wins. I also checked the more precise pythagenpat which lowered the figure to 3.1 wins. Interestingly I found one instance where Strat straight up miscalculated pythag https://365.strat-o-matic.com/league/453619 My RS and RA in that league does not compute to 101 expected wins as displayed. Should be 93.

Anyway as I said at the very beginning of the thread I always expected my teams to underperform pythag due to cheap bullpens, so no real surprises here apart from Strat's miscalculation. Would be interesting for someone to undertake a more systematic analysis - I have a database with RS and RA data for 400 leagues but it doesn't log actual wins, so it won't be me.

I don't want to add further to my own blathering--but I will speak to the Weibull distribution and Weibull methods.

I use them, or rather, assess the adequacy of their usage by manufacturers as an occasional part of my work. Not sure how/why the technique would come up here, but it is used in reliability engineering to model the expected probability for a component or system of components related to their long term operability. Let's avoid talking about the selection of coefficients and keep it plain language.

So when you manufacture something and needs to have a minimum operable time of 5 years, you can never provide guarantees as an engineer because there is no such thing as 100% guarantees. So you provide a reliability probability function for the likelihood the component or system will be able to operate at time x. Then based on cost and schedule, you pick a threshold. So when designing a spec, you demand an 80% probability at time x of system success. This drives quality of parts, build practices etc. It gets quite complex for integrated systems. Anyone who uses it knows it is not absolute.

The methods are different depending if you are modeling something that has actual performance data in real life, or something new, but there are still no absolutes. It is a model.

So quoting a weibull here is strange. In fact it makes the whole argument moot because reliabilty engineering is about anything but absolute outcomes. You design to the non-absolute spec.

So, if a part or system fails before its minimum threshold fail date, people don't start questioning it. It is a reliability model.
You may discern root cause, but you don't throw away all data that doesn't support the hypothesis of "someone must be sabotaging it unintentionally or intentionally." And what is most different in this application is when assessing they are running the equivalent of years of operation to simulate their reliability. Not anecdotes.

Also, a question for the usual suspects who always show up in the conspiracy theory threads - is this really about a conspiracy theory or victimization?

It's about transparency and us knowing what we are paying for. I want to know the ingredients of the secret sauce. Because that will tell us if there is any hanky panky going on behind the curtain.