0.9.12 game engine

[jwr38]

Looking through the version history for 0.9.12 I saw you didn't work on the game engine. Is the game engine fixed and not changing? I don't feel like running tests and finding the same problems and reporting the same things over again.

Is the game engine pretty much set for final release?

[SimonT]

Not Changing.

Final.

Reason : Nothing I can change without redoing 40 hours of tweaking again.

[jwr38]

I would have to say I am disappointed Simon. I hope you will answer my question on how to create players that work. Stated in my 9.11 with questions post.

Personally I believe you will find a lot of people not impressed with the sim after they have a season with it. But this is your product and you can do as you please. I disagree that the engine works in a reliable manner but that is my own personal opinion.

[SimonT]

I try to please all my customer and I know it's impossible to have a 100% approuval rating.

[Rimsky]

Simon,

Am helping Ynohtna in his BFHL league. First, thank you for making a very interesting simulation and we appreciate all the time you spend on this project. JWR has raised some issues about the engine but instead of dwelling on that, I wanted to understand how the engine was stress tested and how the system works at a high level.

Question 1: If all teams are equal (same players etc) - should the teams all be playing at around 0.500 (within a few standard deviations due to normal distribution probabilities?). If not, why, and how does the simulator determine the random elements?

Question 2: If all teams are equal, but one team has a player that is MUCH better - should that team be OVER 0.500 (again +/- a few standard deviations) while every other team should be <0.500 (again +/- a few standard deviations)? Would that better player be the top player (from a statistic perspective, either goals, assists, etc) in his position (again, +/- a few standard deviations, but he should be a top player)?

I ask this, because question 1 and 2 are fundamental to whether the engine "works".

[OFF-TOPIC] But does question 1 and 2 also apply for version 1?

Again, appreciate your work and time - we just needed to get an understanding of the new engine before we dive right into it.

[SimonT]

The engine was stress by simulating over 1 million seasons using different set of rating.

1. No. The randomness strength is different in every formula that kind mess up the normal distribution probabilities. Morale has an effect also unless disabled.
2. Hockey is a team sport, one individual player effect in the result could be impossible to see. I always taught and still think that a good players can have bad stat and a bad players can have good stat.

The engine of version 2 is based on version 1.

A problem that I had in version 1 was the effect on game slider in the game engine because I was unable to test multiple seasons with different slider automatically. The version 2 engine was received a lot of attention in that area because I’m now able test multiple seasons with different slider automatically, running test 24-48 hours in row.

[Rimsky]

Thanks for the insight. We've disabled morale last season as well, exactly to eliminate that random aspect.

I understand if the randomness for shots on goal may be slightly different than the randomness for a giveaway...but that should impact all teams the same. Therefore, if all teams are subject the the same different randomness, it should (over multiple simulations) get close to .500. Could you clarify how this isn't the case?

We love what you've done with the program...given that it's a similar engine, we'll stick with v2 for now and play with the sliders. Wouldn't mind at least point 1 to be considered for V3, since I think that makes sense from a big picture perspective. [Edit: To be more precise, for V3, if Team A plays identical team ALOT, I'd like to see the algorithm to come out such that the win percentage is close to 0.500] Hope I'm not being too much of a problem

[SimonT]

If you have a 2 identical team play 1,000,000,000,000,000,000 times against each other, I would think you get a win percentage close to 0.500 in version 2.

But I would not be surprise at all if in those games, you see numerous winning streaks for one team. That is simply how randomness is. 

Let me ask you a question, if you flip a coin, how many flip do you need to make to be TOTALLY sure to get a 50% result on each side?

[Rimsky]

Ok, I'm fine with the engine. I was just concerned b/c previously you answered "no" to my question 1 above - I wasn't clear enough so you probably misunderstood my question.

I'm ok with numerous win streaks etc - like you said that's part of randomness - I was more concerned that you answered no for my question 1, which you've clarified hahaha.
I didn't mean exactly 50% - i meant 0.500 +/- std deviation.

Like I said, thanks for your input - because it's hard for us to know exactly what goes on behind the software. Just need to know big picture what it does to make sure the GMs can manage around that! Looking forward to trying out V2 season start!

[jwr38]

The engine was stress by simulating over 1 million seasons using different set of rating.

1. No. The randomness strength is different in every formula that kind mess up the normal distribution probabilities. Morale has an effect also unless disabled.
2. Hockey is a team sport, one individual player effect in the result could be impossible to see. I always taught and still think that a good players can have bad stat and a bad players can have good stat.

The engine of version 2 is based on version 1.

A problem that I had in version 1 was the effect on game slider in the game engine because I was unable to test multiple seasons with different slider automatically. The version 2 engine was received a lot of attention in that area because I’m now able test multiple seasons with different slider automatically, running test 24-48 hours in row.

well simon I agree with you that the good players can't be good all the time BUT it is the degree is it happening in version 2. I am finding good players do not preform to their stats very often while mid range players seem to be able to do well on a consistent basis. I was also taught good players will do well more often than an average player.

I have also had the problem of making one team 10% better in all stats. Now 10% in our world of sim management is huge. More often than not this team WILL NOT make the playoffs. Now i don't know about you but most leagues run online and take a YEAR to run a sim. It takes hours or trading to get a better team but with version 2 it seems it doesn't matter if your team is better but how lucky you are. Luck (randomness) should have a minor role not a large role as you currently have it. remember this is with all random extras turned off. Hockey is not luck based game but a team skills event.

I am not the first person to point these same faults to you. I have been sent emails from many people who own the sim and have told me they have told you the same thing for months (and that I was wasting my time trying to show you faults in the engine as you would do nothing). You should have been more clear from the start saying that the game engine was pretty much fixed when you rolled into beta (as I was told by a tester). It would have saved a lot of wasted time on my part.