in case you hadn't noticed on the landing page the little ribbon message

they did the thing

Ding Dong the assist witch is dead

RIP 65% from the field teams winning 75+ games. May you rot in hell.

I'm curious to see what happens and at what point it happens. I didn't see an actual explanation of the changes, just what the end result was meant to be.

I think they just capped the assist boost at some number (which I am guessing 100) so that cost structures no longer reward the strategy

Yeah, I just get nervous because whenever they do something, it always ends up with unintended consequences.

It's been tested in a relatively controlled environment by a group of competent owners. It works like it's supposed to. Rob make a spreadsheet to track real vs expected vs actual results, and everything is behaving within the expected error margins. For example, while everyone was running up-tempo (the first 38 games), actual efg% was within 1.5 of expected efg% for all 24 teams. Now that we've switched to half court, all 24 teams are within 1.1% of expected efg% (the up-tempo boost to fg% is VERY obvious once you run a pretty controlled environment).

Here's the league

Posted by copernicus on 6/30/2022 5:56:00 PM (view original):
I think they just capped the assist boost at some number (which I am guessing 100) so that cost structures no longer reward the strategy

This definitely is not the case. You still get a strong boost going beyond 100, but it's not so overpowering that it's just going to beat well built squads by outshooting them. Previously, a 190% ast team would boost fg% by around half (so 50% shot to go in would have like a 75% chance). Now, it's like a 10% boost (so a 50% shot has a 55% shot to go in).

All of the teams in the experiment have the same set of rebounds, defense, efg%, 3pm, usg... all of these important stats. The only real variable is ast%. The high ast% teams are not dominating. This is in a controlled environment. Now imagine trying to build a 180 ast% team within a salary cap against someone who goes with an old dominant board/def/efg strategy (Curry, Bron, Worm, Worm, etc).

It's fixed, and so far, nothing else seems broken. It's almost as if they just replaced a line in the code with a formula that has values that make more sense.

Is this real life? Could it be?! I haven't joined an open league in over a year because I hated that tired strategy just to get chips. I have no problem with people wanting to boost their efficiency, I just didn't like one strategy being so overly dominant that any other strategy would be an exercise in futility.

last thing I'll mention: the old engine used a mathematical model where the bonus increased almost exponentially the higher you went. Like, the bonus from going from 140-150 was smaller than the bonus going from 150-160. Like the difference between 140 & 150's bonus was like 2%. The difference between 150 & 160 was like an additional 3%! Then 160 & 170 was like 4%!! And so forth.

Now it's like more like, the difference between 140 & 150 is like an additional 1% (actually, probably less). Then 150 & 160 like 0.5%. Then 160 & 170 0.3%. So you're getting a bigger boost, but the amount of the boost is decreasing as you get higher and higher.

It's the same premise as the proposal I shared on these forums some time ago, but they're obviously using a less aggressive model, particularly on the low end.

The conclusion I and most of the other testers came to was this: it seems to function exactly as it did before, but the bonus begins to flatten around 130 where it used to ramp up dramatically there.

Oh, and they got rid of the step function. They confirmed that to us & testing bears that out.

Posted by ashamael on 6/30/2022 8:37:00 PM (view original):

Posted by copernicus on 6/30/2022 5:56:00 PM (view original):
I think they just capped the assist boost at some number (which I am guessing 100) so that cost structures no longer reward the strategy

This definitely is not the case. You still get a strong boost going beyond 100, but it's not so overpowering that it's just going to beat well built squads by outshooting them. Previously, a 190% ast team would boost fg% by around half (so 50% shot to go in would have like a 75% chance). Now, it's like a 10% boost (so a 50% shot has a 55% shot to go in).

All of the teams in the experiment have the same set of rebounds, defense, efg%, 3pm, usg... all of these important stats. The only real variable is ast%. The high ast% teams are not dominating. This is in a controlled environment. Now imagine trying to build a 180 ast% team within a salary cap against someone who goes with an old dominant board/def/efg strategy (Curry, Bron, Worm, Worm, etc).

It's fixed, and so far, nothing else seems broken. It's almost as if they just replaced a line in the code with a formula that has values that make more sense.

so they didnt wind up going with Rob's 100 hard cap? huh

Nope

Posted by copernicus on 6/30/2022 8:48:00 PM (view original):

Posted by ashamael on 6/30/2022 8:37:00 PM (view original):

Posted by copernicus on 6/30/2022 5:56:00 PM (view original):
I think they just capped the assist boost at some number (which I am guessing 100) so that cost structures no longer reward the strategy

This definitely is not the case. You still get a strong boost going beyond 100, but it's not so overpowering that it's just going to beat well built squads by outshooting them. Previously, a 190% ast team would boost fg% by around half (so 50% shot to go in would have like a 75% chance). Now, it's like a 10% boost (so a 50% shot has a 55% shot to go in).

All of the teams in the experiment have the same set of rebounds, defense, efg%, 3pm, usg... all of these important stats. The only real variable is ast%. The high ast% teams are not dominating. This is in a controlled environment. Now imagine trying to build a 180 ast% team within a salary cap against someone who goes with an old dominant board/def/efg strategy (Curry, Bron, Worm, Worm, etc).

It's fixed, and so far, nothing else seems broken. It's almost as if they just replaced a line in the code with a formula that has values that make more sense.

so they didnt wind up going with Rob's 100 hard cap? huh

Ben was the one that put forth the hard cap. I'm pretty sure Rob was the one who told them how to go about doing it as close to the way it was done previously but without the glitchy **** happening at 130.

Posted by copernicus on 6/30/2022 8:48:00 PM (view original):

Posted by ashamael on 6/30/2022 8:37:00 PM (view original):

Posted by copernicus on 6/30/2022 5:56:00 PM (view original):
I think they just capped the assist boost at some number (which I am guessing 100) so that cost structures no longer reward the strategy

This definitely is not the case. You still get a strong boost going beyond 100, but it's not so overpowering that it's just going to beat well built squads by outshooting them. Previously, a 190% ast team would boost fg% by around half (so 50% shot to go in would have like a 75% chance). Now, it's like a 10% boost (so a 50% shot has a 55% shot to go in).

All of the teams in the experiment have the same set of rebounds, defense, efg%, 3pm, usg... all of these important stats. The only real variable is ast%. The high ast% teams are not dominating. This is in a controlled environment. Now imagine trying to build a 180 ast% team within a salary cap against someone who goes with an old dominant board/def/efg strategy (Curry, Bron, Worm, Worm, etc).

It's fixed, and so far, nothing else seems broken. It's almost as if they just replaced a line in the code with a formula that has values that make more sense.

so they didnt wind up going with Rob's 100 hard cap? huh

Hard cap doesn't sound like me. I was all about modeling diminishing returns, similar to what appears to have been implemented.

Awesome! Big thanks to new management, as well as Rob, Ash, Ben, and everyone else involved in helping get this fixed.

Posted by Kona1 on 7/1/2022 12:02:00 AM (view original):
Awesome! Big thanks to new management, as well as Rob, Ash, Ben, and everyone else involved in helping get this fixed.

Thanks, man. I know this particular fix is something I've wanted for a long, long time. I would have preferred a better boost than we're seeing at lower ast% totals, but ultimately, this fixes the biggest flaw in the sim in my eyes, and that trumps any desire I might have had to see my own model be the one they ultimately went with. I know Rob did a **** ton of leg work on this one, and beyond that, he organized our test group as well as doing what he does with data & spreadsheets.

Huge props to the new management! They've done due diligence in making changes to the sim. They haven't rushed just throwing things in a mess but instead listened to a variety of relatively smart people - from different backgrounds with different preferences - and have been very open with their communication and vetting the changes before pushing them live. I couldn't be happier, even if it sometimes seems like it takes a while to get these things through!

I look forward to seeing what's next. I know that the group of us who have been involved in this all have different ideas as to what should be the biggest priority, and I don't want to tease something that ends up not being what WIS chooses to focus on next, as I have no idea what their timetable or priorities are, but I think that almost everybody is going to enjoy whatever comes next.