Hey everyone,
General: To start, I am actually not sure there is any market for what I have been working on since getting my new computer. I believe from play stats that 2K19 isn't as popular now as say 2K14 was post server death, but I'm about 3 teams (2 classics and 1 modern) in on a few tendencies, heights, positions, peaks, and intangible/potential rating changes for the most up-to-date standard 2K rosters for 2K19. Since the servers went down and this computer had to install the game all my previous work for rosters and drafts are completely gone and corrupted now (even when trying to grab them from the old CPU). As a warning for anyone in a similar boat, unless you have to be sure not to transition the game install without backups on your files. Also as a heads up, I only have time nowadays to really edit ~15 players a day (AKA 1 team per day) with the rest of my time going to work and family. So I can edit everything but Free Agency ideally by next month. My hope is to get these completed not long after that point entirely with Free Agency included.
About the rosters: I think it is important to note before getting into my philosophy for changes, I have been testing my changes with the 2K Popboy sliders from YouTube for 2K19 and they have been excellent for me thus far. I'm hoping for other players who are coming back to the game it would help them as well to know how my rosters are working and what slider set I used to work with them. Especially when they are completed to date. I'd recommend using his sliders in general (though some tendencies get weird) and his MyLeague settings that he made for 2K20 (with the caveat of leaving the money alone). I'll also be planning on updating post-every Playoff and Regular Season as long as I continue to play the game with these roster sets.
This leads to my first philosophy with my rosters which people may not like: I base Potential on real life production. I do not randomly assume potential based on assumed progression because of the "Ben Simmons" comps. To clarify what I mean, Ben Simmons to this point in his career has had his best overall season (measured by a few impact metrics) his rookie season. He has yet to quite get back to that point with his measurable production (though he got close in 2020 as his per 48 numbers were comparable) so to assume he should hit around a 90 in his prime just isn't accurate so far in his career. The league caught up to him by figuring out his shooting preferences and thus made him less effective. (Also the additions of Tobias and Butler didn't help his immediate growth with him needing more on-ball roles for his high production). With all of that said, in my roster Ben is an 85 Overall with an 87 Potential (his height in his rookie season by my metrics). He hasn't yet proven beyond that level.
Which leads to my other philosophies about player development which is tied to Peaks, Playoff impact, and Prime. I average all peak ages first to be 25-28 unless a player proves beyond that scale. The reason I do this is because of a report I did in College and a meeting I had then with an ESPN Rep. I took ~550 of the most popular NBA players in NBA history and tracked their impact year by year to see their highest peaks and ideal prime by their impact. To do this, I used a combination of Win Shares (only as a catch), BPM, VORP to track actual production and scaled for minutes as best I could. It measured the major players from '74 to now and discovered the two highest peaks by far either occurred at 25 or 27 and most "primes" actually cut around the latter gradually or drop off quickly from there. Now it is more commonly reported with ESPN openly with their reports regarding data. Though, I will also clarify, it was actually well known before they began having it on TV (at least among NBA analysts). In my single meeting with that ESPN rep, she indicated they had people a few years prior to me discover similar phenomenon as the "athletic peak" around 24-25 and "mental peak" at 27-28. Though they also still overvalue high profile players post-common peaks more than what my metrics revealed historically to be ideally true. So, long story short, I use this as my average just to prevent more editing for me for dual peaks and/or to catch high peak players like Derrick Rose who had their high points cut due to injury but who has shown high enough production late in age to not have "fallen off a cliff." Basically trying to make sure those types of players don't decline too much by age 27. JJ Redick, as an example, had his highest impact season at 31 for the Clippers. His peak in my roster is 25-32 because he reached his athletic peak by 25, but continued working on his skill to his highest point at age 31. This means his potential of 81 in my roster can be achieved at the same time he did in real life. It also could occur earlier with my logic (which he could have done with better playmakers and schemes using him in Orlando).
With Playoff Impact, I don't count Playoffs in my ratings unless the player has significant minutes. I actually use a catch with VORP to have a 2+ minimum to count any Playoff stats for that player. The reason for this is to prevent the "Robert Horry" effect. Essentially what that is, is using Win Shares or BPM to track production and you will see Robert Horry actually had his by far best time of tracked play in the playoffs in 2005. He had a BPM of 7.5 and almost had a .200 Win Share / 48. That is a ridiculous level of impact given his minutes, but he was also playing 27 minutes a game against often other role players coming off the bench. Due to this, using these numbers unchecked that would make his potential and overall in my ratings around a 95 in 2005. I don't know about you, but while Horry was good, he was just not that level of good. It was an anomaly that occurred with a great scheme and rotational game plan that the Spurs had used during that year's Playoff run which boosted his impact ridiculously vs his career average. I found from doing these before that VORP actually does a good job catching these types of players as it requires a consistency and a bigger sample size to hit a 2 VORP in the playoffs.
Moving from player progressions, my next major point of philosophy would be regarding tendencies. This will be my last point I really talk about regarding the rosters as they kind of lump together. I use per possessions and per 36 shot attempts to find out my shot tendencies. I also use Shot Location charts 1 for 1 to have the player's shot location data. Moving to touches, right now I'm using USG% to figure out my touches rating. It isn't ideal as there are better shot creation metrics (like Ben Taylor's aka Thinking Basketball's metrics), but they are behind paywalls and I'm not dedicated enough to the realism to get that granular. The cost there just isn't worth it to me. The same thing is true with Dish to Open Man tendency. I actually don't touch this tendency because I only have access to AST% and it wrongly assumes a few things and has a weird scale with high volume on the ball easy passers getting good numbers but shot creators who are also willing passers are too low with the number (discovered this with Bird having on average 4 assists in test but shot 24 times and missed looks he would normally pass to as the CPU decided to shoot now even over 2 defenders). So I just leave it for now to 2K's default as that seems to perform well enough so far. Finally I do edit fouls. I use similar metrics as I do for Shot attempts and edit fouls that way to give a more realistic high number for most players.
Questions: First, do you guys know where I could host this roster to share with everyone who's interested? I'm older, so file hubs I used to know are primarily behind paywalls now or just entirely gone. I also am wondering if people know of good draft classes for 2K19 that I can use as a base to help keep these going with realism? With the servers gone Steam has no connection to anything so I'm without anything as a base. Also, I am admittedly not good at balancing rosters beyond tendencies and have very little skills in editing attributes (last time I tried, my scale was so different from 2K it broke the game). If anyone has those skills and good solid realistic drafts (or even a better base than 2K's standard rosters for 2018-19) I can use to base similar to my rosters, I'd really appreciate it!
TLDR: I will be posting realistic 2018-19 rosters mostly changing Peaks, Tendencies, Heights, Positions to real life levels as I develop them. I would like to have realistic draft classes to also add and edit to have similar structure if someone knows where these would be / has sets they work on (which all don't mind that I use for the base of this project). I'm using 2K Roster defaults for this project, and if anyone has a better base roster to use with attributes that they don't mind me using as a base, I'd also appreciate that! Thanks!