Fantasy Basketball Categories

[DSMok1]

I was recently experimenting with the categories possible for ESPN's fantasy basketball leagues to try to better match a player's fantasy value to their actual value. I am looking at fantasy value from a Roto perspective, or H2H categories (not points).

I chose Real Plus/Minus Wins as my target, and I only looked at the top 50 players in 2017 as measured by ESPN's Player Rater for the given scoring setup. I measured the correlation of those top 50 players to their RPM Win ratings in 2017.

The standard 8 scoring categories are: FG% FT% 3PM REB AST STL BLK PTS. Some of these only have tenuous connections to actual productivity--FT% is probably the worst offender of the group. The categories must be chosen so that the sum of the whole does not favor particular positions too much.

The standard 8 categories has a correlation of 0.67 to RPM Wins for the top 50 players.

Here is an image with a variety of category choices evaluated (the conventional 8 categories are in orange):



There is a very clear winner here: replacing FG%, FT%, and 3PM with AdjustedFG%, 3Pt%, and Assist/TO Ratio gives a far, far better correlation to Real Plus/Minus Wins, and generally gives a much better top players list. On the other hand, Jae Crowder is now a better fantasy player than Demar Derozan... Is that bad?

EDIT: I have looked at this further and it is actually better to have 3PM than 3Pt%. See Below.

Here are the top 50 players with these new categories.


Any thoughts to the impacts of using these new categories? Will the metagames of fantasy basketball still be the same?

[Crow]

The wide gap between roto-value and RPM is the main reason I've never played roto and likely never will. Your revised categories might resonate with some expert leagues. Some depart from the standard categories... but most never will.

NBA.com has some new category weights are seem a bit off to me or way worse. 1.2 for any rebound? Not for me.

[J.E.]

Pretty interesting. The images aren't showing up for me though

[DSMok1]

UPDATE:

I realized that I should have been looking at the correlation to RPM Wins, not plain RPM.

Interestingly, the winning category setup remains the same. I will update the original post with corrected images and comments.

[DSMok1]

Recently someone asked me about the best points configuration for ESPN fantasy basketball.

I thought I would update this fantasy basketball scoring thread with an answer to this question.

The standard configuration is:
FGM 1
FTM 1
FTA -1
PTS 1
FGA -1
REB 1
AST 1
STL 1
BLK 1
TO -1

Those categories lead to the following top 20 scorers in fantasy, per game, for the 2018-19 season:

Giannis Antetokounmpo
Anthony Davis
James Harden
Joel Embiid
Karl-Anthony Towns
LeBron James
Nikola Jokic
Russell Westbrook
Nikola Vucevic
Kevin Durant
Paul George
Andre Drummond
Rudy Gobert
Kawhi Leonard
Clint Capela
Stephen Curry
Kyrie Irving
Damian Lillard
Ben Simmons
Bradley Beal

As is pretty obvious, bigs are overrated in this metric--particularly big rebounders. KAT, Capela, and Ben Simmons are particularly egregiously high (vs. metrics like PIPM).

So, I took all of the available ESPN categories and evaluated what combination of 10 categories and scoring values provided the best fit onto PIPM.

The results:
PTS 2
FGA -2
3PM 3
REB 1.5
AST 2.5
STL 4
BLK 3
TO -5
MIN -0.5
Team Wins 7

This adjustment brings the R^2 onto PIPM up from 0.56 with standard scoring to 0.77 with the new scoring. And the top 20 (per game) looks far more realistic:

Player
James Harden
Giannis Antetokounmpo
Anthony Davis
Stephen Curry
Paul George
Joel Embiid
LeBron James
Rudy Gobert
Nikola Jokic
Kawhi Leonard
Russell Westbrook
Karl-Anthony Towns
Kevin Durant
Kyrie Irving
Chris Paul
Damian Lillard
Nikola Vucevic
Andre Drummond
Kyle Lowry
Clint Capela

It's certainly not perfect (and underrates KD somehow), but that top 5 would have been a perfectly legitimate MVP ballot last year.

[Crow]

MIN -0.5? Was that for just the top 50 again or everybody? What is your interpretation of that? Top guys playing a little too much / losing efficiency or what?

A/TO less than 2 hurts fantasy score but does it hurt team?

Simmons and Beal replaced by Paul and Lowry. All others stay on list even though position may have changed.

[DSMok1]

MIN -0.5? Was that for just the top 50 again or everybody? What is your interpretation of that? Top guys playing a little too much / losing efficiency or what?

A/TO less than 2 hurts fantasy score but does it hurt team?

Simmons and Beal replaced by Paul and Lowry. All others stay on list even though position may have changed.

Minutes -0.5. This factor is basically saying if you're playing and not compiling any stats, you're a net negative.

A/TO less than 2.0...every assist and turnover has a different context. Related to a simple metric like this (and I'm obviously rounding), that was the break even point that best matched PIPM.

Ben Simmons dropped to 29th. Beal to 27th. CP3 had been 39th in std categories and Lowry had been 50th (!!). Other big movers: Iguodala jumped from 177th to 77th. Brook Lopez jumped from 103rd to 41st. Patrick Beverly jumped from 147 to 63, Robert Covington from 92 to 40.

Big declines: Devin Booker went from 25 to 118. Zack LaVine from 41 to 175. Deandre Ayton from 28 to 112. DeMar DeRozan from 28 to 89. Brandon Ingram from 85 to 313.

The biggest mover in a positive sense was Stephen Curry by a mile (Harden and Paul were next). In a negative sense it was Collin Sexton, Brandon Ingram, and Devin Booker.

An example of a dominant game with this scoring:

https://www.basketball-reference.com/bo ... 40IND.html
Bojan Boganovic would have scored a massive 67.5 points in his 26 minutes of play!

[Crow]

Minutes -0.5. This factor is basically saying if you're playing and not compiling any stats, you're a net negative.

Ok, thanks for reply.

[DSMok1]

UPDATE:

I just looked at this in more detail (through 2020), with additional data. My previous recommended 8 categories was not ideal.

This time I looked at a composite of PIPM, BPM, RPM, and RAPTOR as the target (Thanks BBall-Index!). I looked at fit in a variety of ways and came to the following 8 categories:

Adjusted FG%, 3PM, REB, AST, AST/TO, STL, BLK, and Pts. This is like my original recommendation, but with 3PM instead of 3P%. That change helps the fit quite a bit.

For 9 categories, use those same 8 categories and add in Team Wins.

I did not find a 10th category that would actually improve fit over those 9 categories.

[DSMok1]

UPDATE for Points-based leagues:

I revisited my points assessment, looking at the average of PIPM, RAPTOR, BPM, and RPM. The regression target was a VORP value based on the average of those 4 and using a replacement level of -2.0.

I set Points at a value 1.0 and worked off of that baseline, stepping through and rounding variables to round numbers. The end result has a correlation of over 0.9 with the VORP target.

Here are the values:

Points 1.0
FGA -1.0
3PM 1.0
TRB 0.5
AST 1.0
TO -2.0
STL 3.0
BLK 1.5
PF -0.5
Minutes -0.2

With these values, here are the top 30 players in Fantasy Points per Game in the last 3 seasons:

Code: Select all

╔══════╦═══════════════════════╦═════════╦═════╦═══════════╗
║ Rank ║        Player         ║ Season  ║ Tm  ║ Fantasy/G ║
╠══════╬═══════════════════════╬═════════╬═════╬═══════════╣
║    1 ║ James Harden          ║ 2019-20 ║ HOU ║ 16.1      ║
║    2 ║ James Harden          ║ 2018-19 ║ HOU ║ 15.7      ║
║    3 ║ Anthony Davis         ║ 2018-19 ║ NOP ║ 14.9      ║
║    4 ║ James Harden          ║ 2017-18 ║ HOU ║ 14.7      ║
║    5 ║ Giannis Antetokounmpo ║ 2018-19 ║ MIL ║ 13.7      ║
║    6 ║ Stephen Curry         ║ 2017-18 ║ GSW ║ 13.7      ║
║    7 ║ Damian Lillard        ║ 2019-20 ║ POR ║ 13.2      ║
║    8 ║ Giannis Antetokounmpo ║ 2019-20 ║ MIL ║ 13.2      ║
║    9 ║ Anthony Davis         ║ 2017-18 ║ NOP ║ 13.0      ║
║   10 ║ LeBron James          ║ 2017-18 ║ CLE ║ 12.3      ║
║   11 ║ Paul George           ║ 2018-19 ║ OKC ║ 12.3      ║
║   12 ║ Anthony Davis         ║ 2019-20 ║ LAL ║ 12.2      ║
║   13 ║ Stephen Curry         ║ 2018-19 ║ GSW ║ 11.9      ║
║   14 ║ LeBron James          ║ 2018-19 ║ LAL ║ 11.7      ║
║   15 ║ Karl-Anthony Towns    ║ 2019-20 ║ MIN ║ 11.6      ║
║   16 ║ Luka Dončić           ║ 2019-20 ║ DAL ║ 11.5      ║
║   17 ║ Kawhi Leonard         ║ 2019-20 ║ LAC ║ 11.4      ║
║   18 ║ Chris Paul            ║ 2017-18 ║ HOU ║ 11.3      ║
║   19 ║ Jimmy Butler          ║ 2019-20 ║ MIA ║ 11.1      ║
║   20 ║ Kawhi Leonard         ║ 2018-19 ║ TOR ║ 11.0      ║
║   21 ║ LeBron James          ║ 2019-20 ║ LAL ║ 10.9      ║
║   22 ║ Kevin Durant          ║ 2017-18 ║ GSW ║ 10.6      ║
║   23 ║ Rudy Gobert           ║ 2018-19 ║ UTA ║ 10.4      ║
║   24 ║ Giannis Antetokounmpo ║ 2017-18 ║ MIL ║ 10.2      ║
║   25 ║ Jimmy Butler          ║ 2017-18 ║ MIN ║ 10.1      ║
║   26 ║ Kyrie Irving          ║ 2019-20 ║ BRK ║ 10.1      ║
║   27 ║ Joel Embiid           ║ 2018-19 ║ PHI ║ 10.1      ║
║   28 ║ Nikola Jokić          ║ 2018-19 ║ DEN ║ 9.9       ║
║   29 ║ Kyrie Irving          ║ 2018-19 ║ BOS ║ 9.7       ║
║   30 ║ Trae Young            ║ 2019-20 ║ ATL ║ 9.6       ║
╚══════╩═══════════════════════╩═════════╩═════╩═══════════╝

I did not include any bonus for team wins, but that could be added.

[DSMok1]

And another update for points leagues:

I realized that the points above over-value free throws (no penalty for FTA).

After experimenting, the updated points would be:

Points 1.0
FGA -1.0
FTA -0.3
3PM 1.0
TRB 0.5
AST 1.0
TO -2.0
STL 3.0
BLK 1.5
PF -0.3
Minutes -0.2

Those minor tweaks will improve gaming the system by drawing a ton of free throws... I can't imagine who would do that...