Daily Thunder.com

Look at this latest facebook post by the Sonicsgate movie page about Jeff Green and then read the comments:

http://www.facebook.com/#!/sonicsgate?v=wall&s...

speaking of twitter,

@thunderchants is coordinating crowd chants for the 2010-2011 season. Thunder faithful: Follow and spread the word. Let's organize the power of the Thunderdome to change the game!

@DailyThunder, this was started yesterday and has nearly 200 followers already. How is this not a bolt? For shame.

nick :@KevHey, could you give me a breakdown of your defensive system? I’ve always enjoyed the analysis, but never know exactly what you numbers mean when you give them. It’d be cool to see your spreadsheets too if you don’t mind, I know justin requested that. enick89@yahoo.com if you want to email me.

I will email it to you when I get it done - right now, here is a teaser . . .

http://www.goldenstateofmind.com/2010/3/16/1375989...

@nick
That got a but rambling because I responded based off what I thought you meant, re-read it and then decided I wasn't sure, and then kept going. Haha, sorry about that.

@AC
Just to make sure I'm following your point, you don't like the expected points thing because you think it was basically even out to points produced by actual assists over time? I guess that may be true, but I think it does account for how good of a passes someone is, especially if you compare the two numbers. If your expected points produced is lower than the actual number of points produced by your assists, it could suggest not doing a good job of hitting people in rhythm or in the right spot. Plus, I would prefer to deal with potential assists as opposed to actual assists because of how subjectively assists are kept. If the data is playing with every pass instead of just those the the scorekeeper decided was an assist, I think you get a better sense of things.
Maybe I misunderstood you though, not 100% I answered what you were referring to. Re-reading what you wrote, I feel like the expected points thing that I threw out has a lot of value to your scenarios. The point of it is that it adjusts for passing it out to Ron Artest and therefore does hurt your rating if you continually set up bad shooters instead of good ones.

@Kev
Hey, could you give me a breakdown of your defensive system? I've always enjoyed the analysis, but never know exactly what you numbers mean when you give them. It'd be cool to see your spreadsheets too if you don't mind, I know justin requested that. enick89@yahoo.com if you want to email me.

@nick
Nick, adjusting for what a player's expected shot is seems redundant and tons of work for not much payoff. Sure in small sample sizes (ie. individual games) the shooting of the player you are passing to matters, but lets say over the course of a season, durant makes 20 kickouts to an open Thabo for 3, these will average out to roughly Thabo's season average for open 3s, and if not, the difference between expected assists vs actual assists may be a good metric of how well the player passes to set up the shot.

The expected assists may be interesting, but over the long run player C is going to shoot just as well from player A's passes than from player B's passses, unless player A's passes are easier to catch and shoot, or player A's passes hit player C in stride more than player B. Basically, its a pain that someone will miss an open look on a player's "potential assist" but in all odds it will be made up for by that same player making a tougher shot later. And honestly, if someone keeps passing it to say a Ron Artest, who keeps missing open 3's, it shouldn't help a player's passer rating but hurt it.

Making the effective assist ratio a percentage of the total points scored while the player is in the game also helps to correct for the quality of a player's teammates. If someone is playing alongside KD, his team will be scoring a ton more points, so having an assist isn't weighted as strongly as someone playing alongside Thabo.

I've thought about it more and feel strongly about the quality of my metric, especially the one with hockey assists and usage rate to account for passes setting up assists and the amount of time a player handles the ball.

Mastrmatt :Question for all you stat freaks: Does the enjoyment you guys obviously get out of the stats generated, or potentially generated, from a game give you a bigger thrill than the game itself?
Myself not being a stat guy, I would go insane if I had to sit there with a clicker or spreadsheet and keep track of things. I like getting involved in the emotion of the game. Can you guys do both?
I’m not being critical, just wondering. Different strokes for different folks.

it helps me understand the game better - when I do the breakdowns, I don't have to listen to our announcers say "Player X is playing great defense" - and they are basing it off one steal. I think when you have more knowledge, you feel better . . .

This is in front of the TV - at the game?? No way! I would be too caught up in it to track stats - I can DVR the game and do the stats the next day . . .

@Sammy
Really? I was under the impression it was out there somewhere. I'm not the one to ask about that though, so perhaps not. I do know there are numbers by shot location and I do know there are at least some numbers of contested vs uncontested shots, as I've seen them referenced. I just don't know where to find them.

@Keith
The weighting is definitely the fun part. But what you want to do with the data obviously drives what data you take. I guess you could just go all out and track every pass- who threw it, who they threw it to, and what situation the receiver was in. I guess if we're talking about willingness to pass, the situation that the passer is in matters too. Obviously that's a heck of a task, but as you said, once you get all that data you can have all kinds of fun with it.

@Mastrmatt
I love watching basketball, and I love stats. I kinda fall in line with Keith, if there's a great game going I'd rather just watch it to be honest. But charting something (like when I used to take stats) didn't really take away from my enjoyment, it was just a little more stressful trying not to miss anything live. Being able to watch it back and do that would remove that negative for me. The other thing about keeping stats as you go is that, at least for me, I always felt like I had a better sense of what was going on. Even the simple box score stats. Whatever you think you see, you actually have some data to back it up. Which I enjoyed.

@nick
I thought about this, but I don't know where you find contested vs uncontested numbers by shot location.

Yeah the watching of the games is the longest part, but it should be possible to watch games faster than real time, especially if we have an easy grading/tabulation method. Im pretty sure I can edit the games to have no commercials and play at a FF speed. What would be really nice is a software backend where there is a list of player names and buttons that tabulate each stat, but I am not a programmer so I don’t think I could make that happen.

@Mastrmatt
For me it goes both ways. When I'm watching a game, unless I'm specifically charting something, I'm just into the game. After the game I like to comb through the box score and see what else it can tell me that I missed while watching.

Being big on stats really only becomes an issue when I can't find a game on TV or streaming. If all I can follow is an updating box score, the statistical stuff really is where I find enjoyment. And yes, sometimes the stats are as satisfying as the game for me. But then, it's only with really average games. If it is a great game in the box score, I'd rather have seen it live. If it is a crappy game live, I probably would have preferred just to follow the box score.

@Mastrmatt

Believe it or not this stuff is actually fairly tame compared to some of the ABPR guys...

I don't have trouble enjoying a game while I'm tracking it in some way. It's a little different, though, and it doesn't lend itself well to watching with others.

Question for all you stat freaks: Does the enjoyment you guys obviously get out of the stats generated, or potentially generated, from a game give you a bigger thrill than the game itself?

Myself not being a stat guy, I would go insane if I had to sit there with a clicker or spreadsheet and keep track of things. I like getting involved in the emotion of the game. Can you guys do both?

I'm not being critical, just wondering. Different strokes for different folks.

@justin
Awesome

I'm gonna chart game three of the Laker series this weekend and see what we can come up with.

@nick
I think figuring out how to weight things will be the fun part. Obviously if you have the data (the hard/long part), you can always change your formulas whenever you want. I think it would be good to have lots of different weights and compare how that effects the final score for different players. There's the simple measure: is a player passing without being forced, and are his passes going somewhere? There's more detail: weighted situations for whether a player passes to a better shot or just because, weighted measure of how good a pass it was based on open-ness of teammate and ability to wrangle the pass. Intense detail: weighted measure of every pass made, how often, whether teammates made themselves open, whether player looked for open teammates; weighted measure our how easy pass was to happen, how open player was when passed to, whether player passed to was worth being passed to.

It would be a lot of fun once we have the raw data.

@nick
I guess you could do this for every pass to answer Keith's criticism about passes that don't lead to shots, but it gets kinda crazy I think

@AC
Exactly, I was just playing with that in my head. I think I would want a sort of expected points created by passing per possession using personalized data based on who the pass went to and what situation they were in, then adjust for usage. So it would go something like:

Let's say you pass to an open Thabo for 3. Using shot data that I'm pretty sure is out there, you know that Thabo shoots 40% on open threes (I'm making up numbers obviously). Therefore, the expected points created by that pass are 3 * 0.4 = 1.2,regardless of whether or not he makes the shot. Or you pass to an Durant for a contested 18 footer, and you know from shot data he shoots 35% on contested 18 footers. Then the expected points created by the pass are 2 * 0.35 = 0.7. Throw that all together and adjust for usage and I think you have something pretty interesting. I think a guy named Tom Haberstroh tried something a little like this on Hardwood Paroxysm awhile ago, but it was with actual assists as opposed to any potential assist. I guess you could try to throw hockey assists in there as well, but I think that gets awful complicated.
Obviously this would be really time intensive, I wish this could just be my job.
I guess this isn't tracking every pass the way Kev does, but I think its a little less subjective and data-driven.

@AC
I think they are more informative than assists alone, but part of the issue we are trying to circumvent is whether an assist, in and of itself, is an accurate measure. Guys who make the first of more than one pass that ends up in an open shot shouldn't be penalized just because the first guy didn't shoot. Further, the way assists are counted is entirely subjective. Only about 50% of passes that lead to makes also count as assists. The idea of measuring passing is to measure how willing and well someone passes, regardless of whether a player makes the shot (which again is on teammates, not the pass).

I dont' know if those formulas make sense but they kind of do to me and I think they would be more informative than just assists. And if those formulas or something like them already exist I swear I didn't know.

Nick's point is very good, I think. When playing basketball you have to make decisions based on who is on your team. Do you feed Green with good position in the block or swing it to an open Thabo? These types of decisions are hard, but basically either way a good pass will make it easier to score and a bad pass will make it harder. Also the luck that Justin mentioned (layup missed vs. tipped pass to Thabo) would even out over time. Most layups go in, and most tipped passes result in turnovers.

I think maybe an effective AST% would be best. Taking into account usage, points scored while on the floor, and points accounted for by an assist.

So lets say eAST% =(assisted points=free throws, 2s, and 3s from passes player A made)/ (total points scored while on floor)

And then adjusted eAST% = ((assisted points)/(total points))/(usage rate)

the adjustment would be made to take into account good passers who don't have the ball in their hands often.

Hockey assists (assists of assists) could also be included as another type of adjustment

passer%= (assisted points + 1/2 *(hockey assisted points)/(total points scored while on the floor)

Probably way too late on this discussion (damn class), but I love the idea of more advanced assist stats. I used to keep score for my little brother's games and we kept track of potential assists, which is a very basic start to what you guys are talking about. I agree about finding some sort of weighted assist stat based on the value of the shot (open 3/layup versus long 2), but I do think there is more to it than that. Its probably more than anybody can deal with though. For example, in my ideal assist stat that would take up all my time, I wouldn't weight a pass to an open Thabo for 3 as much as I would weight a pass to an open Harden or Durant. Heck, even an open Thabo versus an guarded Durant is only arguably better. I think I'd get into a black hole if I really tried to mess with all that, determining the exact value of a pass would be a hell of a project. But it would be so cool if you could.
I hope people actually see this post, haha. Here's another vote for a forum Royce.

@kev

I don't think, in most cases, it's highly subjective who is open and who is not when making a play. Usually it's cut and dry...

Keith :@kevThe point is that the pass is being measured, not score. A guy who misses a wide open three doesn’t have more points than a guy who hit a tough jumper over a double team. But, a guy who finds teammates open at the 3 pt line more often than he forces the ball to a guy being double teamed is a better passer.
This stat, in a vacuum, would measure whether a player is a willing passer, whether he is making good passes (teammates in position to score), and with a little more fine tweaking could also (theoretically) measure how well any team moves the ball on offense.

I know - thanks, Justin cleared it up for me . . . I can see what you guys are saying . . . you are still making subjective decisions as to who's open on a given play and who isn't - the other way (even though the passes aren't isolated) . . .

Sammy :@Bryan

I dunno. The bulk of the work is in actually taking the time to watch the games, not in the act of tabulation. It makes more sense to me that if you’re gonna go through the games to tab this stuff up, might as well make the data as detailed as you can.

I don't think the tabulation would really distract that much from actually watching the game.

My concern with what you guys are debating, is that it would require multiple people to make it feasible. It would also be very subjective.

@kev
The point is that the pass is being measured, not score. A guy who misses a wide open three doesn't have more points than a guy who hit a tough jumper over a double team. But, a guy who finds teammates open at the 3 pt line more often than he forces the ball to a guy being double teamed is a better passer.

This stat, in a vacuum, would measure whether a player is a willing passer, whether he is making good passes (teammates in position to score), and with a little more fine tweaking could also (theoretically) measure how well any team moves the ball on offense.

Sammy :@BryanI dunno. The bulk of the work is in actually taking the time to watch the games, not in the act of tabulation. It makes more sense to me that if you’re gonna go through the games to tab this stuff up, might as well make the data as detailed as you can.

I concur . . .

justin :@kev
My concern would be finding the value of the pass. The pass has the same value on the same shot whether the player makes it or not. Finishing the play should have nothing to do with establishing the value of the pass.
If Kevin Durant makes a brilliant wrap around pass to Krstic who blows an open layup, that pass has more value, to me, than if Kevin Durant makes a cross court ill-advised pass that gets deflected, but it’s picked up by Thabo who makes a long two pointer. Durant gets the assist (in most cases) for the bad pass, but the first one basically vanishes from the record books. Even though it was the better play.
I want to isolate the pass.

I see your point - that's why I would have to think about it before formulating a point system . . .

@Bryan
I dunno. The bulk of the work is in actually taking the time to watch the games, not in the act of tabulation. It makes more sense to me that if you're gonna go through the games to tab this stuff up, might as well make the data as detailed as you can.

@justin
This is actually related to a disagreement kev and I have over his defensive scoring system. His defensive scores penalize a defender for a well-contested shot if the shot goes in, and I don't think it should. It's a difference in philosophy, I guess.

I see the value of all the nuanced bits of info, but you have to start someplace; you have to find a way to improve the quality of the data without adding too much in the way of subjective scoring.

@Sammy
I know... but it's really that simple. Like playing an old school video game, just put four fingers on the keyboard and designate a finger for each event type. You could do it while sleeping and eating cheetos and drinking a beer all at the same time.....

....then after the game, just count them up and work up the spreadsheets.... I do stats, this is tooooo simple.

@kev

My concern would be finding the value of the pass. The pass has the same value on the same shot whether the player makes it or not. Finishing the play should have nothing to do with establishing the value of the pass.

If Kevin Durant makes a brilliant wrap around pass to Krstic who blows an open layup, that pass has more value, to me, than if Kevin Durant makes a cross court ill-advised pass that gets deflected, but it's picked up by Thabo who makes a long two pointer. Durant gets the assist (in most cases) for the bad pass, but the first one basically vanishes from the record books. Even though it was the better play.

I want to isolate the pass.

@Bryan
lol @ word document.

I'm chiming in waaaay late on the assists thing, but it seems to me that you could glean a lot from some more basic counts.

If I were tracking a game, I would count Total Passes, Assist Attempts (passes that lead to a shot attempt), Actual Assists and Passes Leading to Turnovers. You could track this easily during a game using a word document and punching in a corresponding letter for each type of event and counting them up at the end.

Obviously there is a lot more info that could be tracked (as discussed by everyone above), but these bits would tell you how efficient the passing is without forcing you to make snap judgment calls about how to score each individual event. It's very raw and basic data, but even if you just tracked totals for the entire team, you could gain a better understanding from game to game (combined with other stats like shot distribution or shot selection), what things seems to be changing from game to game - and why.

Once you grew accustomed to this system, you could add things on, such as specific types of passes or shot attempts, or whatever.

justin :I always hate those assists.. like say Russ passes to KD on the baseline and he’s guarded by Artest and KD shoots over him. Those would be weighted much less than, say, a Russ pass to a wide open James Harden who bricks a three. I think that’s a perfect example of where a pass that doesn’t result in an assist is of higher value than a pass that lead to the assist.

still, that's like saying a guy that is wide open (and bricks a three) should get more points than a guy that makes a contested ten footer . . . short term you have a point, but in the long run the guy that shoots wide open threes will score more points . . .

f5alcon :@kevyeah, it would be really nice to have these stats for everybody, maybe some team can put you on payroll, or espn. Really to get all these stats it would probably take a whole group of guys to get all the games done.

I wish . . . LOL . . .

I wonder if Brian Windhorst ends up leaving The Plain Dealer if LeBron leaves Cleveland. Windhorst has carved out a niche as a fantastic beat reporter who's built a career on LeBron; it'd be really weird to see them separate.

I meant on one of our playoff games that I have recorded...

@Sammy

Right. I think I'll try that out on one of the playoff games this weekend and see what I come up with.

I always hate those assists.. like say Russ passes to KD on the baseline and he's guarded by Artest and KD shoots over him. Those would be weighted much less than, say, a Russ pass to a wide open James Harden who bricks a three. I think that's a perfect example of where a pass that doesn't result in an assist is of higher value than a pass that lead to the assist.

I think the thing to do would be to record the information – the pass, the shot it led to, and whether that shot was made or not – then figure out the weights and everything after you have that raw data.

@justin
No, I think that's reasonable. But then what do you do if the pass results in a made contested long two? Do you not reward the passer at all in that instance?

I think, instead of traditional assist, there should be a 'weighted' passing statistic. Where, say, a pass to a wide open three point shooter is worth the same whether the shooter misses or makes the shot. Because if you think about it, the pass has the same value at the time the shooter catches it. Or is that too much?

@Sammy

Exactly...

Related: I find passing turnovers to be a really crappily recorded stat. They are almost always rewarded to the passer when often times the turnover is the fault of the intended recipient.

@justin
If I were to create my ideal assist stat it would include:
* passes that lead to foul shots
* differentiate between shot location and difficulty (assists that lead to contested long twos would carry less weight than assists that lead to open layups under the basket and open threes)
* would reward players for finding open shots for teammates even if the teammate misses
* would differentiate between assists on fast breaks and assists in the half-court
To get an accurate read on all of these, it'd probably necessary to create a few separate stats.

@kev

I'm going to play around with this and see what I can come up with. At the very least, it'll be illuminating to see how many good passes are made that aren't finished.

something I wrote about assists earlier this year...

I find it hard to believe that with all the statistical analysis out there, there is no better way to evaluate point guards is just by assists. Therefore, I believe one statistic that would be tremendously helpful in judging point guard play would be the "potential assist." The potential assist would be defined as any pass that led directly to an open shot. In other words, even if the player missed the open shot created for him, the point guard would be credited with a "potential assist." The potential assist statistic could lead to other statistics as well, including one that we will call "assist completion percentage." This statistic would measure the percentage of assist opportunities that a point guards teammates are converting for him. For example, if the point guard had 20 potential assists in a game and 9 of them were converted, the point guard would have a 45% assist completion. If we had this stat at our disposal we would know if Westbrook is really two assists better than last year, or if maybe he is creating the same number of potential assist and his teammates are knocking a higher percentage of them down.