It would be nice if there was some sort of "getting hot" effect where if a player or team makes a few shots in a row, they heat up and their chance of making the next shot slightly increases. To counter this, it would be cool if the sim used timeouts to cool the other team/player off.

There are typically multiple timeouts that aren't used in a game, and it would be nice to see some better timeout logic to stop a run by the other team.

Posted by cubcub113 on 3/26/2021 11:11:00 AM (view original):

Posted by gillispie on 3/26/2021 11:07:00 AM (view original):
well i dont know about succinctly but i can explain what is going on.

feedback to reduce 'extreme outliers' was built into the engine ~8 years ago, and some post somewhere by seble i believe states which things it applies to, but it definitely includes shooting on a per player basis and foul committing on a per player basis, i believe there's about 5 things total. they may all operate at a player level; i don't recall. major team level stuff like team shooting, team rebounds - are NOT impacted.

basically, to avoid highly unlikely outcomes, and to reduce complaining about 'unrealistic' outcomes, seble put feedback on these items. so if a player misses their first 8 shots, they are pretty likely to make that 9th, more than they'd normally be.

feedback takes a little while to kick in for each item, but in general, feedback is 'engaged' for much of the game, there's nothing special about the last 5 minutes except that the effect of feedback is based on deviation from norm, and the more events that have happened, the bigger room for deviation. feedback won't act on too small a sample size.

basically, we simplify in our thinking about feedback by thinking things like 'if you should be even and are up 15 with 5m, feedback is probably kicking you pretty hard by the end'. that isn't how it actually works; its really a lot of small feedbacks on small items. however, its probably reasonable to think about it on a game level, just be aware its not really how it works. on the flip side, if a favorite team is up 5 when they should be up 20, they are less likely to get upset late, not more. so its not just like a general upset or comeback function; its just pushing for reversion to the mean, to the norm.

its a pretty controversial thing, for a number of reasons...

three big changes happened game play wise from my standpoint. first, the favorite wins more. if you are regularly upset that you got upset, you aren't coaching as well as you are thinking you are. its definitely significantly easier today for favorites to win titles than it was previously. also, individual shooters taking lead roles in the offense was always good, but because feedback works on a per-player level, higher end teams reap an additional theoretical volatility-based benefit from using higher scorers when possible. a big one for me was the per-player foul feedback, it means the 3rd and 4th fouls are definitely less likely to come - which helps press teams (many things were done to hurt press teams around the same time). i saw a significant reduction in early foul outs on my press teams after. also note this means the aggressive setting (leave in longer maybe it is called now?) on the distro page is even more attractive; if you still aren't using that as a default for every player on every team, you should.

i guess when i say big changes... those are ones i felt were pretty noticeable. i don't really know how any of that affects the planning of the average HD coach. i just figured it would be good to talk about the practical impacts of the feedback changes to help folks wrap their head around what feedback even means, i guess. i think the extent of what most people notice is the last 5m seems crazy sometimes, but frankly people used to talk about weird game outcomes at least as much, probably more.

A lot of these issues are kind of cosmetic, although I would argue that per-player foul feedback is the only thing that affects game-planning (making fb/press better and the "leave in longer" foul setting very powerful.

I've said this before in the discord server, but what's the last time you've seen a player go 3-3 from 3 against a roughly equal opponent? 4-5? What's the last time you've seen someone go 5-5 from the field against an even opponent? It's ridiculously rare. I don't think this affects game results though, since I think most people would agree the sim engine is pretty well calibrated from a macro-level (with Team As level and Team Bs total level the deviation in results has always felt completely right to me).

This is all very interesting. I knew there were 'corrections' in the engine to limit extreme outcomes but thought the effects were applied to the team as a whole. What you guys are saying is that this is really on an individual basis; so individual outcomes are not independent, but affected by past performance with an adjustment towards the mean. As the clock winds down there are fewer opportunities to mean correct and possibly more historical data which varies from the mean. This makes what look like overcorrections in the final 5 min much less surprising...But, knowing this takes a bit of 'fun' out of reading the PBP given there is a constant pull towards the mean. With that said, still agree with Cub that the variance in results seem fair.

On the 'leave in longer' setting - that should be toggled regardless of the feedback mechanic. If there is no efficiency impact to a player with 4 fouls versus 1 foul, you are only hurting yourself by putting the backup in with fewer fouls. I think of the foul toggle is a real life option which is not relevant to HD. Coaches IRL take out their best players in foul trouble to preserve them for the most important minutes of the game. If every minute is equally important in HD, there is no value in sitting your players in foul trouble.

I guess my question on feedback is how it operates on a per player basis (would be great to see the list of 5 from Seble you referenced gil). If a team is up 15 with 5 minutes to go thanks to uncharacteristically great 3 point shooting, will there be a mean correction to their overall offensive efficiency or just their 3-point shooting? They are unlikely to take more 3s with that game-script and more likely to have more time at the FT line.

Edit: gave the above example of a team outperforming it's mean - but question still stands for individual outperformance (if a team is up 15 and their guards are both shooting 50% from 3, will their overall offensive performance be mean adjusted or just 3P)

Real life experience:

I played at D3 U. of the Ozarks in AR. We played a few non-con games vs D2s, and should've beaten the Southern Arkansas Muleriders in both my freshman & sophomore seasons. My freshman year we were up 4 with about a minute left, and had two dumpster fire turnovers en route to snatching defeat from the jaws of victory.

But my sophomore year...

We were ahead by 24 with 10 minutes left. They came back (we stopped doing everything that got us there, and did a terrible job trying to milk the clock on offense), they hit two free throws to tie it at the end of regulation, and they beat us in OT.

swany - this is the best i could find. i know there were more discussions but still:

1) seble's 'development blog' post about the change that added feedback

https://www.whatifsports.com/forums/Posts.aspx?ForumID=30&TopicID=421688&ThreadID=9639684

I've made some good progress this week on the engine tweaks. Below is a list of the changes that are included. This will probably be released next week at some point.
Reduced game-to-game variance in rebounding
Reduced game-to-game variance in shooting
Reduced game-to-game variance in fouls
Reduced game-to-game variance in turnovers
Reduced game-to-game variance on FT shooting
Changed the performance decrease from fatigue so that there will be less of a decrease in general due to fatigue, but a greater decrease than before when a player gets into the range of Tired/Very Tired
Reduced offensive rebound odds on free throw misses
Various minor logic improvements
The net result of these changes will be a reduction in upsets and more consistent results game to game. You should also see fewer of the really extreme results. Upsets will still happen though, as they should, so don't expect to not see them at all after the release.

2) a dev chat that says FT was taken out
https://www.whatifsports.com/devchat/devchat.asp?chatid=88
One of the adjustments on 8/28 was “reduced single-game variance on free throw shooting.” But now it appears it is impossible for a player to start a game 4 for 4 from the line. Each player is forced to miss at least once in their first four attempts. Why was this done? I would much rather have a 72% free throw shooter simply have a 72% chance to make a free throw. (drsnell - Hall of Famer - 1:32 PM)

Actually, starting this afternoon, it will begin to work that way. There was a bug found and fixed related to FT shooting in intentional foul situations. As part of that fix, the variance reduction was removed from FT shooting completely. It's one area I debated about for the 8/28 release anyway.


3) i dont have anything to post but wanted to mention i was pretty much positive in other discussions seble made it clear the foul and shooting effects listed in #1 were on a per-player basis. the reb and TO stuff is almost definitely on a team level because those are team level activities, even though seble added a matchup component to rebounding at some point. which means my earlier post where i said specifically team level macros like rebounding aren't part of it, was wrong! whoops. i'm not sure why the team effects loom smaller in my mind, i think they were both more mild and less situational (in terms of stuff like how they impacted a press vs a man team or a consolidated scoring vs spread out team, etc)

4) about 3s vs 2s and feedback therein, i'm not sure if they are looked at as two categories or not. my guess is that they are not split - that a player has a calculated odds of each shot, and these odds are summed over time and compared to the makes, and deviations result in feedback after a certain sample size (4-5 is my guess). i don't think players have the volume for seble to have split 2s/3s.