https://xrapm.com/
is now live and is being updated regularly
For now its' main focus is impact metrics of various kinds.
It also has shot charts, and other charts showing a player's development across standard BoxScore stats
"Three Factor" RAPM is available, as is coach RAPM, 29-year RAPM etc.
More stuff will be added over time
New advanced stats site - xRAPM
Re: New advanced stats site - xRAPM
This is a welcome development in general and pair data is back.
George - Leonard rates as best pair at +3.7... but sum of xRAPM is +7.1? Different degrees of regression or not playing well together or ?
Pair data is all-time (including playoffs?) or x years? Will it be updated for '24-25? '24-25 separate at some point? Playoffs separate? Yearly data could show progresions.
4 of top 12 broken up by player movement, that original teams did not prevent and may not have tried
Column sorting... not desired, computationally expensive or just not yet? I could download as desired I guess.
7 of 10 top Coaches by RAPM walked away or were ignored for re-hire at a point for reasons? Could look at that more.
Will eventually get to comparing factor data with other source.
And should compare overall RAPMs more, with time and motivation.
George - Leonard rates as best pair at +3.7... but sum of xRAPM is +7.1? Different degrees of regression or not playing well together or ?
Pair data is all-time (including playoffs?) or x years? Will it be updated for '24-25? '24-25 separate at some point? Playoffs separate? Yearly data could show progresions.
4 of top 12 broken up by player movement, that original teams did not prevent and may not have tried
Column sorting... not desired, computationally expensive or just not yet? I could download as desired I guess.
7 of 10 top Coaches by RAPM walked away or were ignored for re-hire at a point for reasons? Could look at that more.
Will eventually get to comparing factor data with other source.
And should compare overall RAPMs more, with time and motivation.
Re: New advanced stats site - xRAPM
Pair RAPM is not influenced by other data. It's just RAPM ran with only pairs, on both offense and defense.
It will correlate, but probably never equal the number that one would get from summing individual impact from elsewhere
Pair RAPM, in contrast to standard RAPM, has 20 entities on the court at each time, so the results will always be more flat
Works nowColumn sorting... not desired, computationally expensive or just not yet? I could download as desired I guess.
Re: New advanced stats site - xRAPM
139 pairs +1 or better on offense. 17 on defense.
Nature of synergies?
Top 17 on defense only had 4 centers and hardly any top names. Randomness is probably a big factor. More elements as you say.
10 players and 20 pairs model might get at size of synergies better? Step forward, step back? Would be another set to compare to, "interpret".
Other ways to rollup:
All player pairs for a player, minutes weighted
All player pairs for a lineup.
New ways to check, interpret, guess.
Player pair partners compared for lineup construction. Quality rank useful. Comparison to raw pairs and their average construction / results.
Agreement between raw and adjusted that Horford is best with Tatum, least good with Brown among 4 main partners. The middle is the middle though the exact levels between them vary. Useful imo.
SGA pairs with JDub and Holmgren real close on raw, pretty clear separation of RAPM. Errors in both, can't say for sure, but lean toward "taking" the RAPM distinction? Maybe. But maybe not as strongly as if raw and adjusted agreed.
4same raw could be brought in several ways. With and without level comparisons. 4same adjusted could be considered, along with many other RAPM variations. Guessing but maybe better guessing. Coach guessing patterns compared to analytic data and analytic guessing.
Datasets can be used in many ways for a long time. Can. Are anywhere close to can? Want analysis and guidance? Pay for analytic production and analysis of analytic production.
Nature of synergies?
Top 17 on defense only had 4 centers and hardly any top names. Randomness is probably a big factor. More elements as you say.
10 players and 20 pairs model might get at size of synergies better? Step forward, step back? Would be another set to compare to, "interpret".
Other ways to rollup:
All player pairs for a player, minutes weighted
All player pairs for a lineup.
New ways to check, interpret, guess.
Player pair partners compared for lineup construction. Quality rank useful. Comparison to raw pairs and their average construction / results.
Agreement between raw and adjusted that Horford is best with Tatum, least good with Brown among 4 main partners. The middle is the middle though the exact levels between them vary. Useful imo.
SGA pairs with JDub and Holmgren real close on raw, pretty clear separation of RAPM. Errors in both, can't say for sure, but lean toward "taking" the RAPM distinction? Maybe. But maybe not as strongly as if raw and adjusted agreed.
4same raw could be brought in several ways. With and without level comparisons. 4same adjusted could be considered, along with many other RAPM variations. Guessing but maybe better guessing. Coach guessing patterns compared to analytic data and analytic guessing.
Datasets can be used in many ways for a long time. Can. Are anywhere close to can? Want analysis and guidance? Pay for analytic production and analysis of analytic production.
Re: New advanced stats site - xRAPM
Player pairs listed only once, so check both name columns.
Player detail found with click on name.
Player detail found with click on name.
Re: New advanced stats site - xRAPM
Banchero 3 factors data totally different than the overall RAPMs.
Probably need metric blends but what to include / exclude and how to weight are obviously important decisions.
Probably need metric blends but what to include / exclude and how to weight are obviously important decisions.
Re: New advanced stats site - xRAPM
Here are some things I see on 3 factor data (of unknown timeperiod):
Joe Ingles 5th on offensive PPS factor. Dort 22nd. .1 and 1 place behind SGA. Jabari Smith 6th lowest. Both Thompson twins in the bottom 15.
Sexton 3rd worst on opponent PPS. Scoot Henderson 4th, Trae Young 9th. Gobert best, Holmgren 4th.
Eason 3rd best on offensive rebound impact. M Turner dead last.
Anthony Black tied for worst on team turnover rate impact. Wembanyama tied for 3rd worst. SGA first, DeRozan 2nd, Brunson best.
Would help in column headers always appeared or at least occasionally.
Joe Ingles 5th on offensive PPS factor. Dort 22nd. .1 and 1 place behind SGA. Jabari Smith 6th lowest. Both Thompson twins in the bottom 15.
Sexton 3rd worst on opponent PPS. Scoot Henderson 4th, Trae Young 9th. Gobert best, Holmgren 4th.
Eason 3rd best on offensive rebound impact. M Turner dead last.
Anthony Black tied for worst on team turnover rate impact. Wembanyama tied for 3rd worst. SGA first, DeRozan 2nd, Brunson best.
Would help in column headers always appeared or at least occasionally.
Re: New advanced stats site - xRAPM
This season, Spurs' TO% is 0.1 higher with Wemby on the floor; but opponent TO% is 1.3 higher.
Last year, it was 1.8 worse with him, opponents only 0.7 higher.
https://www.basketball-reference.com/te ... 24/on-off/
So the trend is from Net +1.1 to net -1.2, where negative is better.
Re: New advanced stats site - xRAPM
Could spend most of waking life studying all the stats out there for insights. If given the position and pay.
Wonder how advanced teams are in using AI to automate the process and make the findings more comprehensive, faster and cheaper. Bad trends allowed to continue can have several interpretations. Not knowing would be the worst. Not acting on it may or may not be reasonable.
Wonder how advanced teams are in using AI to automate the process and make the findings more comprehensive, faster and cheaper. Bad trends allowed to continue can have several interpretations. Not knowing would be the worst. Not acting on it may or may not be reasonable.