Posted by oldresorter on 1/20/2022 4:50:00 PM (view original):
shoe / cubcub - sorry that comment came off the way it did, I was impressed with this thread & how EVERYONE posted their pov. I should have emphasized that more than I did. having paticipated in 1000's of what if forum comments over the years, I was impressed with how much better this one was than the good old days. I was only warning how lucky you all were to have this level of discourse, and to appreciate it. I hope that makes sense?

anyhow, sorry for interrupting the flow, I really was trying to be positive and complimentary, not negative in any way - along with passing on my hope / wish that the game could catch up and provide more modern advanced stats to the users. The data/stats package was relevant for around the year 2000, but in 2020, there is so much more that college coaches can look at, I'd wish what if could spruce up the information provided.

Yea we all love each other here even if it don't sound like it.

Keep on posting!!

Posted by Benis on 1/21/2022 9:13:00 AM (view original):
When using TS%, I don't think it's obvious how good a player is at GETTING to the line. I think using eFG%, TS% and FT rate is necessary to get more of the picture.

I agree with that, using ftar as part of the formula,

I guess like to's, using the stats we get to play with, is a bit tricky, for at least two reasons, first off, end of game fouling skews fta's (does real life too) & second, when a coach sets a bad scorer, say like a 90/65/90 ASD 90 reb / 30 lp pf / c type to very low distro, they will have a very high ftar (since the formula is fta/fga's - yet one might not want to give that player more distro?

I suppose to that second point, 25 or 50% more of a low number, is still a low number.

I have a spreadsheet up right now using all 4 methods, I found ts% being factored by to's and now I added in a factor for ftar, to be the best. I wish I had done this on my last yrs team, as I lost to the eventual winner in my world in the round of 16 by 4 pts, and all season long I felt I coached my team poorly. Using this formula to guide distro would have changed at least one player (my leading scorer) reducing him by half, along with reducing by a 1-2 pts most of my bad players or low distro players.

one reason ts% matters, is it takes into account making the ft's, as well as taking them. james harden or steph for example, ts% or efg% is pretty identical for how they rank relative to the league relative to giannis or say lebron, who miss more ft's and are worse at ts% relative to everyone else.

Posted by oldresorter on 1/21/2022 7:42:00 AM (view original):
cubcub 2 questions (along with a little of my own POV) for you:

1 - why do you use efg% instead of ts%? I probably am not understanding all of your logic, but to my way of thinking when scouting & evaluating college talent real life, ts% adds the component of the ability to draw and make foul shots to the scout. I thought it did in what if too? this in particular, using my logic, would help differentiate (or even up the 'efficiency' fight) between premium star inside players set to -2 and premium outside players - wouldn't it? I must admit, I'm asking, not telling, you have thought this thru WAY WAY more than I have!

1a - buried in all of this fta's and -2 play, is the effect that drawing fouls has on fatigue and sometimes benching the other teams better players, efg% doesn't take that into account either.

2 - unless I'm missing this being included, and I know it's not knowable on face value looking at stats, cause different roles (like pg vs c) seem to make more or less turnovers, don't you have to include turnovers somehow into the 'what stats / equations do I use to determine the distro I set players at?' analysis? A very quick and dirty way to do this, would be to add to's to the (fga + .44 * fta) part of the ts% equation, but that is not quite right either, as it's unfair to pg's (and other assist players), who appear at least, to make more to's just like real life, due to making them both by assisting, and by shooing

1. YIkes OR, thanks for the call out. Whenever I said efg% I meant TS%.

Basically points*100 / FGA + .44*FTA

1a. Yes, but 3pt shots lead to more offensive rebounds, and then there is this "marginal shot" thing I detailed above that also gives a little more value to perimeter shooters (outweighing the fact they draw fewer fouls). So IMO, it's pretty even.

2. Turnover stats suck in this game from a player distro standpoint. I just try to not recruit guys who are turnover machines because they tend to be pretty overvalued. I do have a stat called PPPU (Points per possesion used) which is points / FGA + .44*FTA + TO but the0nly and Gil have told me I really needed to stop using that because TO stats do not work that way. Better to use efg% and try to adjust mentally for abnormally low/high BH/PA guys.

1. But what is your formula? You’re still not getting the numbers I would expect based on the formulas I’ve seen. Is this something you’ve created yourself, or with the0nlyis? It’s important, because you’re coming up with a much wider difference than I would expect based on the accepted formulas I’ve seen.

2. It’s interesting (and telling) that you dive into the stat analysis behind the curtain of the part of the algorithm you’ve clearly worked very hard on, but in response to an issue lying outside the scope of that algorithm, you ignore statistical discrepancy. If you’re going to do scientific stat analytics, it’s very important that you go in without the sort of biases you’re showing here (“12 BH in a big is roughly meaningless). So *points* matter to you, but *possessions* don’t? Brooks *should* have rebounded as well, and *shouldn’t have* turned the ball over more, because you don’t know how to account for them, so they’re just not part of efficiency I guess. Hard pass. :l

3. This is where your GI/GO takes you completely off the rails. Brooks will score more points, for sure, given a higher distribution. That’s not in doubt, although the claim that 2.5x is optimal is a severe stretch. But again, it’s not about two players in a vacuum, it’s about whole team. More on that below.

4. See #2 I guess. Just a bad take.

5. See #1. I question the validity of that 3% difference in the first place.

*From basketball-reference: “Effective Field Goal Percentage; the formula is (FG + 0.5 * 3P) / FGA. This statistic adjusts for the fact that a 3-point field goal is worth one more point than a 2-point field goal. For example, suppose Player A goes 4 for 10 with 2 threes, while Player B goes 5 for 10 with 0 threes. Each player would have 10 points from field goals, and thus would have the same effective field goal percentage (50%).”

This is clearly not what you’re doing, because if it was, the difference you’d get would just be the difference in FG% (neither player shoots 3s). I understand you are trying to add FT into equation, but that tells me you’re trying to make this tell us something entirely different from what efg% is designed to do. So again, let’s see what that formula is, show your work.

6. Respectfully, I am pretty sure I am thinking just fine. No team needs 5 starters scoring 15 points to be optimal. To say Wisconsin was ever “desperate” for scoring was an overstatement, but certainly by tournament time, with the freshmen on the bench, they were scoring plenty of points. In that situation, having some scorers on the bench with them is actually preferable. In summary, players bring more to the +/- than scoring.

6 part 2. Stamina doesn’t make the player better. It’s not a modifier (like amphetamine) that spikes performance. It just tells us how long the player will be able to sustain the level of established goodness. It’s feeling like you’re trying to treat it like another athleticism, and it’s not that. Deficiencies can get amplified in some settings when teams try to do things without having enough of it, but once you have enough in the right spots, more isn’t better. So to use Wisconsin, if they were set on running uptempo, having Brooks start might be a slightly better option; but only for a couple minutes per game. Pacquet would have no problems getting 20-22 minutes per game, even uptempo, if settings were right (I suspect Rowle may have been using minutes, or had him set to getting tired, not sure though). IOW, 71-75 stamina does not prohibit getting 20+ minutes of elite rebounding and defense, if that’s what a coach decides they want from their post player.

Look, I’m not going to tell you to stop using your algorithm. You’ve spent a lot of time on it, and if it works for you, use it. Lots of ways to build a team.

But calling Pacquet “unstartable” on an elite team is just incorrect. You are *vastly* overstating your case, and overestimating the scope of your tool. Elite teams can win with pitching and defense.

ETA—thanks for clarifying on the ts%

wow

"say like a 90/65/90 ASD 90 reb / 30 lp pf / c type to very low distro, they will have a very high ftar (since the formula is fta/fga's - yet one might not want to give that player more distro?"

Yeah you def need to consider the usage when comparing players stats - whether it's efg% or FT rate or anything else. Smaller sample size will likely skew it positively in that players favor.

But I do think FT rate is something that shouldn't be ignored because fouling can be so important in HD. But of course you need to also consider team make up, play style, etc etc

Do you guys usually just look at TO% when adjusting distro, or is there a better advanced metric that I'm not aware of. Maybe something like TO% adjusted by usage?

Posted by shoe3 on 1/21/2022 10:10:00 AM (view original):
1. But what is your formula? You’re still not getting the numbers I would expect based on the formulas I’ve seen. Is this something you’ve created yourself, or with the0nlyis? It’s important, because you’re coming up with a much wider difference than I would expect based on the accepted formulas I’ve seen.

2. It’s interesting (and telling) that you dive into the stat analysis behind the curtain of the part of the algorithm you’ve clearly worked very hard on, but in response to an issue lying outside the scope of that algorithm, you ignore statistical discrepancy. If you’re going to do scientific stat analytics, it’s very important that you go in without the sort of biases you’re showing here (“12 BH in a big is roughly meaningless). So *points* matter to you, but *possessions* don’t? Brooks *should* have rebounded as well, and *shouldn’t have* turned the ball over more, because you don’t know how to account for them, so they’re just not part of efficiency I guess. Hard pass. :l

3. This is where your GI/GO takes you completely off the rails. Brooks will score more points, for sure, given a higher distribution. That’s not in doubt, although the claim that 2.5x is optimal is a severe stretch. But again, it’s not about two players in a vacuum, it’s about whole team. More on that below.

4. See #2 I guess. Just a bad take.

5. See #1. I question the validity of that 3% difference in the first place.

*From basketball-reference: “Effective Field Goal Percentage; the formula is (FG + 0.5 * 3P) / FGA. This statistic adjusts for the fact that a 3-point field goal is worth one more point than a 2-point field goal. For example, suppose Player A goes 4 for 10 with 2 threes, while Player B goes 5 for 10 with 0 threes. Each player would have 10 points from field goals, and thus would have the same effective field goal percentage (50%).”

This is clearly not what you’re doing, because if it was, the difference you’d get would just be the difference in FG% (neither player shoots 3s). I understand you are trying to add FT into equation, but that tells me you’re trying to make this tell us something entirely different from what efg% is designed to do. So again, let’s see what that formula is, show your work.

6. Respectfully, I am pretty sure I am thinking just fine. No team needs 5 starters scoring 15 points to be optimal. To say Wisconsin was ever “desperate” for scoring was an overstatement, but certainly by tournament time, with the freshmen on the bench, they were scoring plenty of points. In that situation, having some scorers on the bench with them is actually preferable. In summary, players bring more to the +/- than scoring.

6 part 2. Stamina doesn’t make the player better. It’s not a modifier (like amphetamine) that spikes performance. It just tells us how long the player will be able to sustain the level of established goodness. It’s feeling like you’re trying to treat it like another athleticism, and it’s not that. Deficiencies can get amplified in some settings when teams try to do things without having enough of it, but once you have enough in the right spots, more isn’t better. So to use Wisconsin, if they were set on running uptempo, having Brooks start might be a slightly better option; but only for a couple minutes per game. Pacquet would have no problems getting 20-22 minutes per game, even uptempo, if settings were right (I suspect Rowle may have been using minutes, or had him set to getting tired, not sure though). IOW, 71-75 stamina does not prohibit getting 20+ minutes of elite rebounding and defense, if that’s what a coach decides they want from their post player.

Look, I’m not going to tell you to stop using your algorithm. You’ve spent a lot of time on it, and if it works for you, use it. Lots of ways to build a team.

But calling Pacquet “unstartable” on an elite team is just incorrect. You are *vastly* overstating your case, and overestimating the scope of your tool. Elite teams can win with pitching and defense.

ETA—thanks for clarifying on the ts%

Pacquet isn’t unstartable on an elite team. He’s just unstartable on Wisconsin given that Brooks is an option and the team isn’t exactly flush with scoring (no truly elite scorers). If I had a PG, 2 elite shooters at the 2/3, and a nice LP scoring 4 I would kill to have Pacquet at the 5 to round out my team and be heavy title favourites.

Ratings over stats any day. To me, it is just abundantly clear that Brooks is the option here over Pacquet, because the gap is scoring is so massive (Brooks I expect to be able to score 10 PPG and Pacquet more like 5 at team efficiency). Turns out when you look at the stats the gap was even wider then I might have surmised (Brooks 15 points per 40 and Pacquet only 6).

6b. You do not understand what I’m saying. Read the stamina section I wrote again shoe, it’s perfectly clear. But the summary is a guy with 90 STA is going to be more fresh way more than a guy at 75 STA in slowdown flex/man, and given rowle loves to play uptempo we see an even larger difference than expected between Pacquet and Brooks.

And about the modeling (and logic) I have to analyze scoring stats, I almost never use it anymore because I’ve played this game enough and used my model enough in the past I can tell that a 13 point LP buff and 15 points STA buff pushes a player from a non-scorer to a scorer.

And I’m not overstating my case at all. The coaches who don’t optimise their teams’ distro and are just like “screw it dude, Pacquet can be my third scorer” and then just ignore the fact he’s shooting 50% from the field with poor FTA/FGA rate are the ones who tend to struggle to achieve success in the NT significantly.

“screw it dude, Pacquet can be my third scorer”

- rowle 2022

Cub, what it comes down to is that I think your algorithm overvalues stamina (and yes I understand your point, I just have a different view), and completely ignores possessions and defense, and both are leading you to the absurd conclusion that Brooks brought significantly more value to Wisconsin as a starter. Even if I was being generous and granted the 2.5 optimization figure, he’s still not the best or even second best scoring option in the starting lineup - arguably he’s the worst in the last lineup, running a flex offense, opponent neutral. No team needs 5 guys scoring 15 pts/40 to win games. And anyway, the team optimization number is not a fixed standard, and you’re assuming the whole team was optimized and not… having their 55per PF shoot 3s because someone convinced him to bump up attempts.

I suspect a big part of the disconnect on stamina is that you’re assuming a larger drain rate, ie when a player is on the low end of fresh, or getting into fairly fresh rather than entirely fresh, they play significantly worse, to a larger degree than I assume. And to the games credit, that’s one area they have smartly left up for user interpretation, how fast the performance deteriorates, and how severe it is. In FB/P, I see that deterioration have significant effect. In my other sets, the real value high stamina has is that it keeps worse players off the floor. It does also let your excellent scorers take on a higher distribution load (high distribution does drain it faster) so it can be helpful there if you want to push the envelope with a guy. But again, I don’t think that would be advisable with Brooks in the starting lineup, not with all the better options in with him.

Tactically, my approach would be to start Brooks with Han when the opponent’s defensive strength is on the perimeter (either weaker man or press interior defenders, or a 3-2 zone base). That’s when his scoring advantages over Pacquet have real impact. Apart from those situations, my tendency is going to be to get him (or maybe one of the other starters, but then the question is who?) on the bench, where they are fully utilized. You say there are no “truly elite” scorers there, and ok fine, no 100-across-the-board guys, but there are 4 players with 93+ LP, and 3 of them pair that with 95+ athleticism. It’s not like Brooks is doing something no one else on the team can do (score in the paint).

TLDR, repressing your opponent’s fg% and improving possessions through rebounds and avoiding turnovers are all part of the value picture, and from a tactical standpoint, there are usually going to be better ways to utilize a player like Brooks rather than as the 5th best scoring option in the starting lineup.

But again, I see these two as pretty much a wash, from a value standpoint. Both are bringing immense value to the team, and can be utilized to help the team win at the end of the year from the starting lineup, or the bench.

It’s interesting that you bring up “volatility” gil. I’ve been suspicious for the past few months (1-12 depending on the topic) that the post-seble handlers have been tinkering with volatility in various areas. First injuries, then grades, and more recently I’ve wondered if parts of the game engine haven’t been touched (probably temporary, or intermittently) as well. Like a FG% probability variable for a game widens by a percentage point or 3, for a few games at a time, and this gets felt a little more by teams running uptempo, or FB/P (and especially both). Outside of that context, of course, from a statistical standpoint, we would expect running uptempo FB/P to increase possessions, which should *decrease* volatility. As the sample size increases, the outcomes should head back toward the mean, right? Which should be advantageous for the better teams under normal circumstances (ie, if we could assume normal probability variables).

*raises hand quietly*
*whispers*

What about kimball's Stanford in Naismith...?