There is no code in the engine that adjusts anything before the beginning of the game (aside from player improvement). You're actually making my point for me - people who play this game don't like it when the inherent variance of the game is too large. The difference between these two results was a little bit of HCA and a bunch of luck. There is still a fairly broad distribution of outcomes possible for any given game. It looks like in your first game against Alaska Fairbanks you got about an 80th percentile offensive output and in the 2nd game you got a 5%. Those are still pretty far apart. My guess is if you simulated that game 100 times you probably score about 80 points on average.

I'm sorry then. I must be a complete idiot. You're right I don't like the inherent variance of the game whether it's game to game or in game. Isn't kinda of the same thing whether it's game to game or in game that the variance is too large. His 2 halves are kinda the same as my 2 games except that shoe claims in game that the code helps the losing team balance out the outcome of the game by adjusting the probability of the winning team to do worse and the losing team to better(far better) to make result closer to what the outcome should be. Maybe I'm just not understanding. Sorry if I'm being stupid

One thing to consider. In this game and in real life sports, whenever our team does well, we consider THAT the norm when it could very well be the outlier. So maybe its not that your team underperformed that much but that it happened to way overperform previously and that skewed your opinion of how good your team really is.

Posted by dahsdebater on 10/4/2018 11:12:00 AM (view original):

The game *should not be forcing or speeding regression* because that is a fundamental perversion of probability and statistics, and building it into the game means we are not dealing with a simulation, we’re dealing with a god mode.

This really isn't true since in effect this implementation preserves a predictable distribution of outcomes, it just narrows it. But you're free to have your own take on it.

One way of thinking about this is that rather than flipping a fair coin it's more as if the sim is randomly picking from a set of papers marked with 1s and 0s. In that case if you pick a significantly non-equilibrium number of 1s, you're going to naturally experience a correction in which subsequent draws have a greater probability of coming up 0 than the initial distribution. The more off-center your existing results are, the bigger the correction. The more you've drawn, the bigger the potential correction. This is still a random statistical process. The mathematics driving it are not identical but share a lot in common with the way I presume this regression correction would have been programmed.

It is true. You’re right that you experience a natural regression to the mean over time. When the parameters are preserved, the regression is natural, and there is no valid complaint. If an opponents 67% FT shooter goes 1-4 from the line in the first half, then 3-4 in the second half, that’s a natural “regression”. It’s possible (but less likely) he could also shoot 4-4 in the second half, and if regression is natural, there is no complaint. But when the system has weighted his second half performance differently, based on his first half performance, that’s where the problem is. No when he goes 4-4 in the second half because the sim is speeding up or forcing regression, and you lose by a point, it is correct to say you were actually screwed by the sim, because the sim has been engineered to change the parameters in the middle of the game.

The difference between the two scenarios is that with natural regression, the player is always flipping a 67% coin. In the second scenario, he’s flipping a coin much higher, because the sim is trying to get to the expected result before the end of the game. That’s not how probability works, that’s the whole point of this thread. So denying that it’s a perversion of probability is incorrect.

Posted by terps21234 on 10/4/2018 1:17:00 PM (view original):
I'm sorry then. I must be a complete idiot. You're right I don't like the inherent variance of the game whether it's game to game or in game. Isn't kinda of the same thing whether it's game to game or in game that the variance is too large. His 2 halves are kinda the same as my 2 games except that shoe claims in game that the code helps the losing team balance out the outcome of the game by adjusting the probability of the winning team to do worse and the losing team to better(far better) to make result closer to what the outcome should be. Maybe I'm just not understanding. Sorry if I'm being stupid

You’re not being stupid. That’s the long and short of it. If you have good luck and your opponent misses a higher than expected number of shots in the first half, the sim is designed to give that opponent a better shooting percentage baseline in the second half. So that 67% FT shooter isn’t always flipping a 67% coin. If he gets cold in the first half, he can be magically transformed into a much better shooter in the second half.

Posted by shoe3 on 10/4/2018 1:59:00 PM (view original):

Posted by dahsdebater on 10/4/2018 11:12:00 AM (view original):

The game *should not be forcing or speeding regression* because that is a fundamental perversion of probability and statistics, and building it into the game means we are not dealing with a simulation, we’re dealing with a god mode.

This really isn't true since in effect this implementation preserves a predictable distribution of outcomes, it just narrows it. But you're free to have your own take on it.

One way of thinking about this is that rather than flipping a fair coin it's more as if the sim is randomly picking from a set of papers marked with 1s and 0s. In that case if you pick a significantly non-equilibrium number of 1s, you're going to naturally experience a correction in which subsequent draws have a greater probability of coming up 0 than the initial distribution. The more off-center your existing results are, the bigger the correction. The more you've drawn, the bigger the potential correction. This is still a random statistical process. The mathematics driving it are not identical but share a lot in common with the way I presume this regression correction would have been programmed.

It is true. You’re right that you experience a natural regression to the mean over time. When the parameters are preserved, the regression is natural, and there is no valid complaint. If an opponents 67% FT shooter goes 1-4 from the line in the first half, then 3-4 in the second half, that’s a natural “regression”. It’s possible (but less likely) he could also shoot 4-4 in the second half, and if regression is natural, there is no complaint. But when the system has weighted his second half performance differently, based on his first half performance, that’s where the problem is. No when he goes 4-4 in the second half because the sim is speeding up or forcing regression, and you lose by a point, it is correct to say you were actually screwed by the sim, because the sim has been engineered to change the parameters in the middle of the game.

The difference between the two scenarios is that with natural regression, the player is always flipping a 67% coin. In the second scenario, he’s flipping a coin much higher, because the sim is trying to get to the expected result before the end of the game. That’s not how probability works, that’s the whole point of this thread. So denying that it’s a perversion of probability is incorrect.

This perception that nothing with a moving target can be probabilistic is absurd. I outlined an example of this and you entirely ignored it. I guess because you had to to fit your narrative.

Posted by dahsdebater on 10/4/2018 2:27:00 PM (view original):

Posted by shoe3 on 10/4/2018 1:59:00 PM (view original):

Posted by dahsdebater on 10/4/2018 11:12:00 AM (view original):

The game *should not be forcing or speeding regression* because that is a fundamental perversion of probability and statistics, and building it into the game means we are not dealing with a simulation, we’re dealing with a god mode.

This really isn't true since in effect this implementation preserves a predictable distribution of outcomes, it just narrows it. But you're free to have your own take on it.

One way of thinking about this is that rather than flipping a fair coin it's more as if the sim is randomly picking from a set of papers marked with 1s and 0s. In that case if you pick a significantly non-equilibrium number of 1s, you're going to naturally experience a correction in which subsequent draws have a greater probability of coming up 0 than the initial distribution. The more off-center your existing results are, the bigger the correction. The more you've drawn, the bigger the potential correction. This is still a random statistical process. The mathematics driving it are not identical but share a lot in common with the way I presume this regression correction would have been programmed.

It is true. You’re right that you experience a natural regression to the mean over time. When the parameters are preserved, the regression is natural, and there is no valid complaint. If an opponents 67% FT shooter goes 1-4 from the line in the first half, then 3-4 in the second half, that’s a natural “regression”. It’s possible (but less likely) he could also shoot 4-4 in the second half, and if regression is natural, there is no complaint. But when the system has weighted his second half performance differently, based on his first half performance, that’s where the problem is. No when he goes 4-4 in the second half because the sim is speeding up or forcing regression, and you lose by a point, it is correct to say you were actually screwed by the sim, because the sim has been engineered to change the parameters in the middle of the game.

The difference between the two scenarios is that with natural regression, the player is always flipping a 67% coin. In the second scenario, he’s flipping a coin much higher, because the sim is trying to get to the expected result before the end of the game. That’s not how probability works, that’s the whole point of this thread. So denying that it’s a perversion of probability is incorrect.

This perception that nothing with a moving target can be probabilistic is absurd. I outlined an example of this and you entirely ignored it. I guess because you had to to fit your narrative.

I ignored your example because it is not meaningful to the discussion. A “moving target” is a perversion of probability based simulation. You can find expressions to make it look like some kind of statistical paradigm, no doubt. But it’s inconsequential, essentially a red herring, because apparently you’d rather divert folks attention away from the fact that the sim is designed to screw them if they enjoy good luck in the first half. I guess.

The simple fact is, when you flip a fair coin, and it lands heads 3 times in a row, your odds for flipping tails next are still 50%. Engineering different odds for the coin flip based on what has happened is a perversion of a probability based simulation.

Dahs wins.

/thread.

Thanks for playing Earl.

Look, I think it's pretty clear that you're not going to change your mind here. That's fine. I don't think it's important for you to do so. I don't think you're leaving the game. So I'm probably about done with this thread. The main thing is that in HD forums that are already often quite toxic, I did think it was important after seeing you blast the central element of the game - simulation of game results - that somebody come in and at least provide a counter-perspective. At least let new players see that there is a viable argument that, hey, maybe the game isn't so unfair as somebody said.

I do think that your perspective here is based on a very narrow definition of the words "probability" and "fair." Using the same equations to determine the range of outcomes of each event in each game is certainly an easy way to be fair and probabilistic. But I don't think it's the only way. The sim engine as it is now determines outcomes based on fixed equations, which determine an expectation value, and then a random number generator output is combined with that expectation value to determine the outcome of each event, resulting in a predictable distribution of outcomes. This is basically the textbook definition of a probabilistic simulation. Every team is subject to the same set of equations, they aren't changed, they don't discriminate against any team, any coach, any type of team, any type of coach, or as far as I can tell any system or strategy. This is basically the textbook definition of fair. Each team in each game has an equivalent probability of being positively or negatively adjusted based on their early-game performance. Nobody is being given an inherent advantage.

One could, if one wanted, think of the difference between the sim you are advocating for and the sim we have as the difference between a Monte Carlo simulation of a basketball game and a what if simulation of a basketball game. We are intentionally sampling more in the regions of high probability and largely ignoring the best- and worst-case scenarios in this engine, narrowing the distribution of outcomes. This is classic Monte Carlo behavior. Now, one could also argue that "WhatIfSports" should be using the "what if" simulation model. But Monte Carlo sampling is, in general, a more efficient way of describing probability distributions of reasonably normally-distributed variables.

Posted by Benis on 10/4/2018 1:46:00 PM (view original):
One thing to consider. In this game and in real life sports, whenever our team does well, we consider THAT the norm when it could very well be the outlier. So maybe its not that your team underperformed that much but that it happened to way overperform previously and that skewed your opinion of how good your team really is.

Sorry dude that is way not true. My team didn't over perform the first game against AK FB. My team is a good team, beat very good teams. I have a whole season to prove that. In a game that I win by 29 then lose by 5 against a SIM is ridiculous. If you want to say they overperformed and I only won by 2 or 3 fine, but not lose outright to an inferior team that I already beat and the SIM didn't make any adjustments.

Posted by dahsdebater on 10/4/2018 4:18:00 PM (view original):
Look, I think it's pretty clear that you're not going to change your mind here. That's fine. I don't think it's important for you to do so. I don't think you're leaving the game. So I'm probably about done with this thread. The main thing is that in HD forums that are already often quite toxic, I did think it was important after seeing you blast the central element of the game - simulation of game results - that somebody come in and at least provide a counter-perspective. At least let new players see that there is a viable argument that, hey, maybe the game isn't so unfair as somebody said.

I do think that your perspective here is based on a very narrow definition of the words "probability" and "fair." Using the same equations to determine the range of outcomes of each event in each game is certainly an easy way to be fair and probabilistic. But I don't think it's the only way. The sim engine as it is now determines outcomes based on fixed equations, which determine an expectation value, and then a random number generator output is combined with that expectation value to determine the outcome of each event, resulting in a predictable distribution of outcomes. This is basically the textbook definition of a probabilistic simulation. Every team is subject to the same set of equations, they aren't changed, they don't discriminate against any team, any coach, any type of team, any type of coach, or as far as I can tell any system or strategy. This is basically the textbook definition of fair. Each team in each game has an equivalent probability of being positively or negatively adjusted based on their early-game performance. Nobody is being given an inherent advantage.

One could, if one wanted, think of the difference between the sim you are advocating for and the sim we have as the difference between a Monte Carlo simulation of a basketball game and a what if simulation of a basketball game. We are intentionally sampling more in the regions of high probability and largely ignoring the best- and worst-case scenarios in this engine, narrowing the distribution of outcomes. This is classic Monte Carlo behavior. Now, one could also argue that "WhatIfSports" should be using the "what if" simulation model. But Monte Carlo sampling is, in general, a more efficient way of describing probability distributions of reasonably normally-distributed variables.

I mean... you’re not delusional enough to think that anyone thought there was a chance *you* might change your mind... right?

There were a lot of folks reaching for the Xanax when 3.0 came out. I understand the impulse to get another view out there “for the new folks”, when it looks like the system is being painted a certain way by the people who don’t like it. A lot of folks have a very specific view of me because that’s exactly what I did. As I said from the outset, people like what they like. But we should be honest about it, and not try cover it up, or rationalize it with advanced looking statistics. A flipped fair coin is always going to have the same odds, no matter how it flipped the last time. A simulation that deviates from that is perverting probability. Your models and bracketed formulas don’t change that.

Another reason why I appreciate mully, while I think he’s completely wrong about half the time (including this one), he’s upfront. He’s ok with the system extra-favoring the favorite. He’s ok with god mode, as long as it’s on the side of the favorite. He likes what he likes. Cool.

Posted by terps21234 on 10/4/2018 4:29:00 PM (view original):

Posted by Benis on 10/4/2018 1:46:00 PM (view original):
One thing to consider. In this game and in real life sports, whenever our team does well, we consider THAT the norm when it could very well be the outlier. So maybe its not that your team underperformed that much but that it happened to way overperform previously and that skewed your opinion of how good your team really is.

Sorry dude that is way not true. My team didn't over perform the first game against AK FB. My team is a good team, beat very good teams. I have a whole season to prove that. In a game that I win by 29 then lose by 5 against a SIM is ridiculous. If you want to say they overperformed and I only won by 2 or 3 fine, but not lose outright to an inferior team that I already beat and the SIM didn't make any adjustments.

I don't think you have a bad team. I think your team is pretty good. I was just speaking in generalities about everyone's biased perceptions about their own team - whether it's an HD team or your favorite college basketball team. You can pinpoint the games where they played really well and say 'see, they're really good!'. The problem is that you're just looking at it in a vacuum and not the whole picture. The great teams are the ones who play well every night and not just occasionally.

In your case, you did beat a couple of very good teams and based upon that, you can say that your team is also good. But then again your team struggled outright against some very weak competition a couple times - only a 7 pt win against a sim team with an RPI of 200+ or having to beat the 175 RPI team in OT.

IMO, the reality of your team is that it's probably somewhere in the middle of those two games against Alaska Fairbanks. You're definitely better than that team and 'shouldn't' lose to them like you did. But on the other hand maybe you're not 29pts better. Maybe the 29pt win is farther from the mean (or norm) than the 5pt loss. It's hard to make that determination based upon just the two games though.

I dunno, just spitballing here.

For what it’s worth, if it’s really important to know why I say the 0s and 1s paradigm is a red herring (now that I have a few extra minutes)...

First of all, there is no reason to think the HD game engine works like this. It doesn’t work like this in determining recruiting, or EEs. But even if it does, there is a *fair* way to do it, and it would not be too complicated to change. Drawing 0s and 1s from an *unlimited* set is a probability based simulation, and is fair. To do this, you would be constantly replacing the drawn number, not removing the number from the table and the possibility of being drawn again.

For example, say a player has a C FT grade corresponding to a hidden “hard” probability of 67%. I use FT shooting so we don’t get lost in the weeds of all the variables going into individual shooting percentages, like defense, defender, settings, double teams, etc. Every time this player shoots a free throw, using the 0s and 1s paradigm, he should be drawing a number from a table filled with 100 cards; 67 “1s” (made FT) and 33 “0s” (missed FTs). The table should have the same number of cards every single time he draws. That is a fair, unbiased system. If he makes 10 in a row, he should still have a table with 67 1s and 33 0s on it.

The difference between a fair system described above and the system dahsdebater is defending is that in the dahsdebater system, once the player makes 10 in a row, he is now drawing from a table with 57 1s and 33 0s. His effective FT ability has magically been turned from 67% to 63%. The simulation is biasing itself toward expected results. This doesn’t have massive effect (although the principle is the same) when limited to one player’s free throws, and when the total is as high as 100; but when applied to the entire team’s FG%, the effect can be quite significant.

And obviously, a table with 100 cards is pretty generous, because as he admits, there is still a lot of variance when the total is that far above what a player will use in a game. If we take it down to 9 cards total, 6 1s and 3 0s, then your 67% FT shooter can be magically turned into a 0% shooter after 6 made FT. That’s the point with crossing over from probability into god mode, and why I draw a very clear distinction between them.