APBRmetrics

Basketball-Reference.com's Ast% definition is

... an estimate of the percentage of teammate field goals a player assisted while he was on on the floor.

Chris Paul has the highest career playoff Ast% of anyone with at least 500 minutes: 53.2 (Stockton 47.8, Magic 42.2).
http://bkref.com/tiny/jVRBy

All 3 of his playoff appearances to date rank among the top 9 alltime single playoff Ast% (of at least 200 minutes), including the highest ever, in 2009 -- 60.2%.
In that first round elimination, he scored 27% of his team's points and assisted on just about 50% of the rest.

I don't know a practical way to search for it, but I'm guessing no one has topped 77% 'total scoring involvement', however defined.

In 1973, Nate Archibald was responsible for 30.8% of his team's total points and 43% his team's total assists. But Tiny played 80 games, not 82. He finished the season with 3681 minutes out of a possible 3956. Add the two percentages together and multiply by the ratio of possible minutes to MP and you get a percentage of 79.3%; not surprising since Tiny became the only player to lead the league in both scoring and assists that year

Mike G wrote:... he scored 27% of his team's points and assisted on just about 50% of the rest.
I don't know a practical way to search for it, but I'm guessing no one has topped 77% 'total scoring involvement', however defined.

This is a mess.
If he scored 27% of the points, assisting on 50% of the other 73% -- 36.5% -- then he was involved in 63.5% of the scoring.

The same analysis of Nate Archibald's 1973 season : scored 31% of all points; assisted (910/2593) 35% of the other 69% = 24% of all scoring; sums to 55% 'involvement'.

Whoops!

Now that I've bothered to look at how ast% is calculated [100 * AST / (((MP / (Tm MP / 5)) * Tm FG) - FG)], I have a question: Why use made FG in the calculation, rather than FGA? Using FGA would factor team pace and team FG% into a player's ranking; if a player gets a high percentage of his team's assists, but the team doesn't make a high percentage of it's FGA, maybe he isn't making his teammates better with his passing; I think that's relevant information.

Using [100 * AST / (((MP / (Tm MP / 5)) * Tm FGA) - FGA) as the formula, here are the top 5 seasons, followed by the actual ast%:

1990 JS: 28.9%, 57.4%
1991 JS: 28.1%, 57.5%
1995 JS: 26.6%, 52.6%
1992 JS: 26.5%, 53.7%
1988 JS: 26.1%, 54.8%

AYC wrote:... if a player gets a high percentage of his team's assists, but the team doesn't make a high percentage of it's FGA, maybe he isn't making his teammates better with his passing; I think that's relevant information...

With just boxscore totals, we can't know whether a passer is making his teammates shoot better, or good scorers are making it easy to get assists.

'Percent of team's assists' has never been what Ast% is.
The way b-r.com does it, there's no distinction between a passer with a low-efficiency offense or a high-powered offense. It's just the % of (others') made shots you assisted.

That's an interesting variation you've contrived. But I don't know what this quantity would be called.

John Stockton (JS) had a lower than average home/away disparity in his assists/TmFG. Chris Paul's has been high. Steve Nash's in Phx has been very low.

Well, the job of the point guard is to "run the offense". We assume that a player who gets a lot of assists is creating easy looks for his teammates; that means facilitating made FGs that otherwise might have been misses. So putting total FGA instead of FG in the denominator makes sense. Then if an offense is sub-par or mediocre, that will be reflected in the numbers

Nice job AYC. The idea of team FGA does make a lot more sense.

I don't know how to value assists. Without watching the game it's impossible to know how much credit you give to the passer or the scorer. How do you value perfect passes that lead to a missed shot or free throws?

The only advancement that makes sense to me is expected turnovers based on "possessions". Every pass leading to an FGA (incl fouled shots) and every shot created (un-AST'd FGA) is credited as a "possession". These can each be estimated for every player based on box score stats. Once you have these numbers you figure expected turnovers based on lg avgs. Then find "turnovers avoided" by subtracting turnovers from expected turnovers. Since we know the point value of a turnover avoided (or possession added) this seems logical.

AYC wrote:Whoops!

Now that I've bothered to look at how ast% is calculated [100 * AST / (((MP / (Tm MP / 5)) * Tm FG) - FG)], I have a question: Why use made FG in the calculation, rather than FGA? Using FGA would factor team pace and team FG% into a player's ranking; if a player gets a high percentage of his team's assists, but the team doesn't make a high percentage of it's FGA, maybe he isn't making his teammates better with his passing; I think that's relevant information.

Using [100 * AST / (((MP / (Tm MP / 5)) * Tm FGA) - FGA) as the formula, here are the top 5 seasons, followed by the actual ast%:

1990 JS: 28.9%, 57.4%
1991 JS: 28.1%, 57.5%
1995 JS: 26.6%, 52.6%
1992 JS: 26.5%, 53.7%
1988 JS: 26.1%, 54.8%

I like the idea - BUT, you'd have to look at it season to season. Maybe adjust your A% by league average for that season. I'd also maybe take out that specific players shot attempts and misses, since one can't assist (or attempt to assist) their own shot attempt - or maybe not, one taking the shot maybe shows they aren't finding the open guy. Up to you.

The Stockton seasons you showed above were some of the better FG% seasons historically (not best mind you, but well above average) - before NBA players started shooting tons of threes, and after the old school low %s of the 50s through the 70s. Without any adjustments - old school PG assist ratios (or % or whatever you'd call it) during tougher assist times (ala Cousy) would be downright pitiful - even more so than using convential assist %.

I've been doing tons of historical stuff recently - trying to be another Mike G I guess - and everything for me starts with a normalized season baseline to work every season off of - since the game statistically has changed so much through the years.

That being said - quick glance at my recent (unposted as of yet) work - I have a handles/passing subset of my ratings (ast & to with league, pace, team adjustments all inherent within) that could give my statistical perspective on regular season Chris Paul. 2008 Chris Paul is #3 over the last 32 years (since 1979-80) in the BH rating - just barely behind 98 & 99 Mark Jackson. '90, '88, '91, '95, & '92 Stockton seasons come in #4 & #7-10, respectively. 2011 Paul is #12, 2009 Paul is #17, 2010 Paul is #26.

Paul has been #1 in the NBA in BH the last 4 years, in '07 at 21 years old he was #2 to Nash, in '06 at 20 he was #5 behind Brevin Knight, Billups, Kidd, & Nash.

By age groups the last 32 years 25 year old Paul ('11) was #2 to Stockton ('88), 24 yo Paul was #2 to Bogues ('89), 23 & 22 yo Paul were easy #1s, 21 year old Paul was #2 to Sessions ('08 - in only 450 minutes though), and 20 yo Paul was #2 to Sergio Rodriguez ('07 - in only 862 minutes).

Chris Paul is pretty good.

Statman, a player's own FG (or FGA in my preferred version) are already taken out of ast%. Also, while FG percentages were higher in the '80s, the pace was also faster, which means more FGA in the denominator; so a pace adjustment is already built in.

PS I have found 42 seasons with an AST/FGA% (not sure if that works as a name) of 20% or better; Lebron in '10 just missed. There might be some seasons I have missed. CP3 does fall on this list relative to his position on the all-time Ast% list, but Steve Nash fares better than anybody besides Stockton:

1990 JStockton 0.2887
1991 JStockton 0.2805
1995 JStockton 0.2658
1992 JStockton 0.2653
1988 JStockton 0.2609
1989 JStockton 0.2554
2010 SNash 0.2484
1994 JStockton 0.2473
2011 SNash 0.2471
1993 JStockton 0.2438
1989 EJohnson 0.2429
1987 EJohnson 0.2428
2007 SNash 0.2424
2009 CPaul 0.2404
2008 CPaul 0.2392
1991 EJohnson 0.2392
2008 SNash 0.2361
1986 EJohnson 0.2353
1996 JStockton 0.2342
1985 IThomas 0.2330
1984 J.Lucas 0.2324
2005 SNash 0.2322
1988 EJohnson 0.2309
2011 RRondo 0.2301
1985 EJohnson 0.2301
2009 D.Williams 0.2274
1990 EJohnson 0.2241
1984 EJohnson 0.2230
1979 KPorter 0.2228
2002 A.Miller 0.2224
1990 KJohnson 0.2202
2005 B.Knight 0.2136
2009 SNash 0.2135
2011 CPaul 0.2095
2006 SNash 0.2083
1984 NNixon 0.2079
2004 JKidd 0.2056
1984 IThomas 0.2019
2002 JStockton 0.2013
1992 KJohnson 0.2010
1990 M.Bogues 0.2006
1989 KJohnson 0.2005
2010 LJames 0.1995

AYC wrote:Statman, a player's own FG (or FGA in my preferred version) are already taken out of ast%. Also, while FG percentages were higher in the '80s, the pace was also faster, which means more FGA in the denominator; so a pace adjustment is already built in.

Once again - you are hurting PGs from the more current era - say after 1995 - because teams shoot a ton more threes - making normal FG% LOWER despite TS% being higher. You are giving more credit to players assisting in an era of higher FG% because they shot alot less threes, but didn't score any more per possession.

PG's from earlier eras (70s & earlier) will rate MUCH lower than the 1980ish PGs - since league FG% was MUCH lower back then (style of play, more liberal defensive rules, less elite shooters, maybe even harder rims). There is absolutely no way Stockton would have put up those type of %'s in the 60s or 00s. That's all I'm trying to say.

It doesn't seem to me that current point guards are hurt all that much. Look at that list; after Stockton--whom we should all expect to be on top --the best performers are Nash and CP3, along with Magic. Btw, Magic team's were the most accurate from the field in history, yet his best season's aren't in the top 10; remember, the faster pace then puts Stock and Magic at a disadvantage. And how difficult is it to assist a 3-pointer?

I've recently been making a bunch of adjustments to assist rate. These aren't per 100 teammate FG or FGA, but rather per 36 minutes adjusted by several factors.

If a player gets more assists at home than away -- as a % of teammate FG -- I reduce his home Ast/FG to his away Ast/FG and give him the equivalent total assists.

This reduces by quite a bit the league overall assist total. But it's countered by another adjustment: I call an assisted 3-pointer 1.5 assists. This is estimated by teammate 3fg/2fg ratio.

Adjustments are made for estimated minutes with and against starters, and team + opponent points.

So if everyone were put on a standard team -- league avg 3fg/2fg ratio, scoring and allowing 100 PPG, home scorekeepers no more liberal than away -- and everyone playing 2/3 of his minutes vs starting players, we could have adjustments to total assists in 2010-11.

Ranked by 'equivalent assists added' -- that is, players whose assists (or the effects thereof) are understated by pure Ast/36 min.Note lots of players from good defensive teams, with teammates who make plenty of 3's. And all starters.
These teams' scorekeepers also tend not to inflate their home assists.
Billups was +8 in NY, Felton +1 in Den, Hedo zero in Phx.

Those whose raw Ast/36 most overstate their assist totals:Mostly players from weak teams or non-starters, mostly younger, and basically all PG's.

The league adjustment totals -406, among all players with an assist. The league Ast/36 adjusts downward from 3.20 to 3.18 .

"Chris Paul has the highest career playoff Ast% of anyone with at least 500 minutes: 53.2 (Stockton 47.8, Magic 42.2)."
http://bkref.com/tiny/jVRBy

That is impressive but with Stockton at #2 without a title, I wondered how well top Ast% guys did on winning championships.

Among the top 10 on career Ast% there was one multiple title winner (Magic Johnson) and a one-time title winner (Rondo). Of there 6 title seasons they were over 40% Ast% 3 times, over 30% once, under 30% twice.


That seemed kinda light on titles by this top 10, so I investigated further. Here are the totals for descending groups of ten. I went fairly fast, checking some cases I wasn't sure of but not absolutely everyone, so I might have missed a title winner here or there, but probably not many.

Ast% rank Multiple One title
1 - 10 1 1
11 - 20 1 2
21 - 30 1 1
31 - 40 0 1
41 - 50 2 2

1 - 50 5 7

51 - 60 1 1
61 - 70 1 0
71 - 80 0 2
81 - 90 0 0
91 - 100 0 1

51 - 100 2 3


The top 50 were more than twice as likely to be title winners but the rate was still a bit less than 1 in 4. The second 50 won a title in 1 in 10 cases.

The 41-50th ranked group with Ast%s between 28.3 and 29.5% did best in terms of winning at least one title.


How does that compare to points per game leaders?
(using same low 500 minute cutoff)


Pts/ g rank Multiple One title
1 - 10 2 2
11 - 20 1 3
21 - 30 4 1
31 - 40 3 3
41 - 50 1 1

1 - 50 11 10

51 - 60 2 2
61 - 70 2 1
71 - 80 0 5
81 - 90 4 0
91 - 100 2 0

51 - 100 8 8

The top 50 on points per game were almost twice as likely to win at least one title as the top 50 on Ast%. More than 40% did. The second 50 were more than 3 times as likely to win at least one title as the second 50 on Ast%.

The first 50 on points per game had a modest sized lead over the second 50 on points per game.

The sweetest spot was from 21 - 40 with guys averaging 21.9 to 24. 3 points, but most of the scorer subgroups throughout the top 100 did pretty well.

Are big time points per game players overrated or not? It is not a question with a simple yes or no answer. This data should be part of the discussion.

It might be interesting to check the frequency of title winners when a team has at least one player in the top 50 or 100 on both stats. I am not sure if I feel like matching them up and compiling that data for a long time period right now but I'll look backwards from last season a modest distance.

The 2010-11 Mavs had the 9th best career points per game scorer and the 14th best career Ast % guy.

The 2007-8 Celtics had the 44th, 69th and 75th best career scorers and the 10th best Ast% guy.

The 2005-6 Heat had the 11th and 21st best career points per game scorers and the 43rd best career Ast % guy.

The Spurs title winners had the 33rd and 84th best career points per game scorers and the 69th best career Ast % guy.

The 2003-4 Pistons had the 52nd best career points per game scorers and the 45th best career Ast % guy.

So 7 of the last 9 title winners had guys in the top 100 on both. 5 had one in the top 50 on both.


Who else last season had guys in the career top 100s on both ?

Besides Dallas, Miami, Boston, Oklahoma City, Chicago (Rose providing both), San Antonio, Utah / NJ (with Deron providing both), Detroit, Orlando, Portland and New Orleans (Paul providing both) met these criteria.


Who had guys in the top 20 on both during last season?

Just Dallas, Miami, Oklahoma City, Chicago (Rose providing both), Orlando and Portland. All 4 of the conference finalists made this cut.

Only Oklahoma City and Chicago had one in the top 10 in each. Dallas was pretty close with a 14th on Ast% from Kidd. Miami got a 18th from James on Ast%. Orlando was outside the top 10 on both (one barely). Portland was was outside the top 10 on both by modest amounts.

Among other teams fairly close to twin top 20s, New York didn't get enough Ast% from Billups to make it. Bryant hasn't gotten (or permitted) that level of Ast% beside him and hasn't provided it himself either. The Clippers were close with Williams-Griffin. The Raptors were only short by a bit. The Nets have 2 guys flirting with top 20 scoring along the 3rd best Ast%. Houston, Golden State and Indiana have one top 20 scorer but lag moderately on Ast%. Maybe someday Rubio will provide that counterpart for Love. Ridnour wasn't bad but was a bit outside the top 30 on Ast%. Randolph barely made the top 20 on points per game but Conley was in the 30s on Ast%. Utah. Washington, San Antonio and Charlotte had the top 20 guy on Ast% but were off by a modest amount or more on the scorer. Atlanta missed the top 30 on both (one barely, one badly). Not good. Denver missed badly on the top scorer criteria.

Which team were in both "sweetest spots" according to recent history (as cited above) on these criteria? Houston. Dallas was close. The Lakers, Indiana, Minnesota, Golden State Memphis and the Knicks weren't too far off either.

Having the combination of top guys on these 2 criteria seems like a good way to roll. But it doesn't always work, as it didn't for Stockton-Malone, a lot due to Jordan-Pippen.

Among the title winners in the last 9 years meeting both criteria it always involved 2 different players. Maybe a bad sign for the title hopes of Rose and Paul who didn't have a 2nd guy meeting the other criteria but instead did both or nearly did both themselves. The Lakers with Kobe are the only recent exception without both criteria met and without 2 different players involved.


Which teams will give high priority to getting dual current top 20s or improving to top 10s or dual career leaders (with the experience factor that brings to the playoffs)? Or even just improving on their previous shortfalls from these standards? Which teams will drift down some from dual high ranks? Player movement and changes in how plays are run and shot are distributed well affect these rankings. How conscious and concerned are teams to changes on these criteria?

How much does achieving these ranks actually impact winning titles compared to missing one or both? I don't know for sure. The recent data suggests it might be important but a longer study would help determine if this recent trend of usually meeting both criteria is typical over the longer term or not.

What is the sweetest range for total scoring involvement for the top player or the top 2 or 3 on a team? I briefly looked at something similar to that when I checked number of players on teams with usage over 20% recently but more analysis could be done for a longer time period. Is it historically better to exceed or be below the sweetest spot or sweetest range for your top player? That should be known and used to help guide team construction & management.