It's simpler than that, isn't it? At the start of a quarter, teams aren't in the bonus. Reserves are in by the time they get in the bonus in the first and third quarters, and I'm willing to bet teams play their full starting lineup more often in these quarters than in the second and fourth quarters, when the situation is reversed.
Sheesh, I can't believe I didn't think about that. It makes perfect sense, of course, and also explains the ramp up in FTM/FGA (and therefore efficiency) between 10 and the peak of 7 starters.
Mike G wrote:
So, the players in the game at the beginning of the possession are those responsible for fouls, and I'd also think they are the ones responsible for FTA. Yet your dilemma is that sometimes it's the possession-starters that determine a possession (foul+FT), sometimes the possession-enders (foul/no-FT). That certainly skewers things.
Exactly - you can have a situation where there's a rebound in the backcourt, a foul after or during the rebound, subs, and then a continuation of the possession (an inbound and a play). Or you could have a play where the shot is blocked out of bounds, subs come in, and then the possession continues. I'm not quite sure how to handle it. With the last numbers I put out, everything in the possession is just attributed to the lineup that ended the possession. But obviously that's problematic in the block example, and would be in the foul example as well.
Mike G wrote:
The turnover breakdown is quite weird:
Start - TO%
5 v 0 - 22.1%
5 v 1 - 20.5%
5 v 2 - 15.3%
5 v 3 - 16.2%
5 v 4 - 15.7%
5 v 5 - 15.5%
What do you find weird about this? The only thing that seems really out of place is the dip at 5v2. Other than that it seems to trend upwards, consistent with my analysis in the original post: the subs that come in are better (or at least neutral) defensively and worse offensively, thus turning it over more on their end and forcing more or the same on the other end.
Mike G wrote:
I would love to play with some others of these. Maybe if you repost, just the totals? I think I can copy/paste these things if the lines don't wrap. Placing a spreadsheet somewhere to download would also work. I'm also wondering about steals and blocks.
Sorry, I forgot that my resolution is generally bigger than most - I saw that the tables didn't wrap on my monitor and assumed they were OK. I'll try and link up to a Google Spreadsheet, it's easier for me anyway.
BadgerCane wrote:
I'd guess that those last minutes, with no starters for anyone, take place at the wrong end of a blowout where nobody in the building even cares what happens at that point. Also, I would guess that teams with worse rotations and generally worse players would be the types of teams to experiment with lineups that do not involve starters. So, this dip in production could be a reflection of that.
This is a good point, but from observation I think that you're wrong to assume all or most of the low-starter time is when the game is in hand. 23% of all possessions are played with 4 or less combined starters in the game, meaning an average of 11 minutes per game. From my earlier study, we can see that even the best teams have relative blowouts for 10 min per game - and most of the time the coaches don't feel secure in that blowout for a long time after that. So 11 minutes per game, league-wide, is an awful lot to attribute to blowouts.
I'll see if I can look into it, though, just to check all the factors. You very well could be right that it's having a large influence on the numbers.
Last edited by Ben F. on Wed Jan 23, 2008 12:04 pm; edited 1 time in total
Tied up in this data may also be another phenomenon -- that players tend to do better when their minutes come in larger chunks. When you get down to four or fewer starters on the court, there's almost guaranteed to be a couple guys out there who are playing just a handful of minutes that night, and the data shows it's really hard to be effective coming in cold and coming out before you're warm.
Tied up in this data may also be another phenomenon -- that players tend to do better when their minutes come in larger chunks. When you get down to four or fewer starters on the court, there's almost guaranteed to be a couple guys out there who are playing just a handful of minutes that night, and the data shows it's really hard to be effective coming in cold and coming out before you're warm.
Out of curiosity, how would you study this? I would think that looking at the production of low minutes players would be incredibly hard because it's hard to separate out the different causal factors. If a player gets low minutes and performs badly, is that because he's not as good or because he's cold? And if he gets more minutes and plays better, is that because he's having a good night, or is there a mismatch or a suitable style of play for him so the coach leaves him in? In other words, aren't there a lot of confounding factors that could get in the way of studying how players play in short minutes versus long minutes?
Joined: 14 Jan 2005 Posts: 1712 Location: Delphi, Indiana
Posted: Wed Jan 23, 2008 12:19 pm Post subject:
Ben F. wrote:
Mike G wrote:
The turnover breakdown is quite weird: ..
What do you find weird about this? ... the subs that come in are better (or at least neutral) defensively and worse offensively, thus turning it over more on their end and forcing more or the same on the other end.
.
Ohh, my bad. I was thinking the subset "5v0" referred to the productions of the 5 starters (vs 5 non-starters). But you've lumped together the (5 vs 0) and the (0 vs 5) subsets. And we really don't see from this how the 5-0 squad does vs the 0-5 unit; nor vice-versa.
So, I guess if you're really going to upload a spreadsheet, then it's more universally useful to give us the 5-4 and the 4-5, etc. Just twice as many lines. Looking forward to it. _________________ 40% of all statistics are wrong.
Ohh, my bad. I was thinking the subset "5v0" referred to the productions of the 5 starters (vs 5 non-starters). But you've lumped together the (5 vs 0) and the (0 vs 5) subsets. And we really don't see from this how the 5-0 squad does vs the 0-5 unit; nor vice-versa.
Sorry for not being clear. The 5v0 means the production of both lineups when one side has 5 starters and the other side has 0. So it includes both the 5 starter lineup and the 0 starter one.
I guess for the spreadsheet I'll break down how each side performed, like Mountain and Chicago were asking for. Any idea what I should do about the split possessions? Is it a problem if it all gets grouped in the possession ending lineup? By my count, 4,607 possessions end up "split" in this way, which is about 4.3% of the total. Would you rather see the totals sooner, or have me work on trying to split possessions in the middle?
The spreadsheet is now online. When possessions are split by subs, I simply grouped them in the possession ending lineup. Again, this represents only 4.3% of all possessions, so I don't expect it to skew the data much at all.
Here's the really interesting addition: so much for the "bonus" theory. Unless I'm doing something wrong, it seems that fouls are down almost as much as FTs in the 5v5 lineup. The 5v5 lineup gets to the line at a rate that's 60% of league average (0.14 FTM/FGA compared to a 0.23 average) but only draws fouls at a rate that's 71% of league average (0.17 Fouls/Poss compared to a league average of 0.24). So there's still something else going on here.
Thanks for sharing the dataset. It could be a very fruitful database especially drilling down further using other parameters. I'd ask if anyone has ever prepared such before but I assume most of those who might have wouldn't say.
I see that generally the offensive efficiency of the two sides of a starter imbalance are fairly close until you get to 4 or 5 on 1 or 5 on 0. I didnt expect that but maybe I should have at league level given different team qualities and player quality and situation and performances on a given night. Good to see & learn. Most of the extremes are small sample sizes though so not sure a lot should be made of their behavior.
Time & situation data could be interesting: this data for bulk of game vs garbage or clutch time.
Or with starter PG vs not. Or #1 star or not.
If you overlaid player quality maybe you could find some stuff about how many good players or how good a leader needs to be on the floor to "hold the fort"- or score.
Maybe Brandon Roy could be an interesting case study.
You could look at the patterns at team level. Which coaches get more starter advantage minutes (and big advantages) and what do they get out of it? Who faces more starter disadvantage minutes and are they cool with it or getting taken by the opposing coach? I'd hope that something like that could come out even for a sample team. You could even take it down to play against important rivals. Teams might be interested in this data if they don't have it. Basketballvalue.com could potentially add a layer for 5 man lineups to show performance against 0-5 starters. It could be an intermediate step between the raw rollup data and true player quality adjusted data.
In the copy I downloaded I am going to add per possession values where appropriate to help analysis of those columns.
5 x 5 represents 20.9% of all possessions or about 10 minutes a game.
4 X 4 or greater totals 58.3%.
3 X 3 or greater totals 76% of time.
Last edited by Mountain on Thu Jan 24, 2008 5:25 pm; edited 1 time in total
Somebody asked how to study the performance of players in games they get very few minutes -- there are two ways to do this -- one is to compare individual players based on their "big minute" games vs. their "small minute" games; the issue is that they may just be playing in matchups that are most advantageous, thus biasing the result.
The other, more robust way is to look at players who suddenly see a big increase in playing time due to external factors -- i.e. injuries or trades.
You could look at the patterns at team level. Which coaches get more starter advantage minutes (and big advantages) and what do they get out of it? Who faces more starter disadvantage minutes and are they cool with it or getting taken by the opposing coach? I'd hope that something like that could come out even for a sample team. You could even take it down to play against important rivals. Teams might be interested in this data if they don't have it.
Bettors might be interested in this data for their over/under issues (joined with x pace vs. y pace data), which is the most this study can be useful.
Any thoughts on the new development that it doesn't seem to be simply the lack of being in the bonus suppressing FTM/FGA but rather a lack of fouls?
I want to make sure I'm understanding the spreadsheet first:
When I look at 4x0 and see 0, those are stats for the team with no starters vs. 4, right? That would mean that the team with no starters only commits fouls .17 times per possession...
At the low foul end of the spectrum, I see that a team of no starters tends to not foul against anyone, regardless of whether they're playing against 0 to 5 starters. Three possible explanations:
1-The zero starter lineup is presumably pretty fresh, so they're not committing stupid reach fouls. They're able to move their feet more. A lot of fouls are the result of poor defensive position and fatigue.
2-The zero starter squad sees limited minutes, so they're more likely to be conservative in just about everything they do. As mentioned, it's difficult to enter a game and "play".
3-Zero starters against a full squad is likely to happen in situations where the zero starter team is well ahead. You don't want to foul in these situations, you want to keep the clock moving.
Every other lineup with <= .22 fouls/poss involves the team that presumably would either be playing from well ahead or has the more rested lineup, ie, the team with more bench guys on the court:
1 vs. 5
4 vs. 5
1 vs. 4
2 vs. 5
The other lineup is 5 vs. 5, which tends to happen at the beginning of each half, when players are well-rested.
Somebody asked how to study the performance of players in games they get very few minutes -- there are two ways to do this -- one is to compare individual players based on their "big minute" games vs. their "small minute" games; the issue is that they may just be playing in matchups that are most advantageous, thus biasing the result.
The other, more robust way is to look at players who suddenly see a big increase in playing time due to external factors -- i.e. injuries or trades.
It's difficult to isolate three or four possible reasons for the performance of low minute players:
-it could just be tough for anyone to come off the bench when everyone else is in the flow of the game. You might be able to test this for everyone by comparing their performance off the bench the first 3-4 minutes vs. the remaining minutes of their rotation.
-do bench players perform better when you increase their minutes in a given game? If so, is the 8 min a night guy getting increased minutes because they're playing well in a particular game? Could it be due to favorable matchups?
-how do bench players perform when their role is suddenly increased due to team injuries?
I suspect a lot of players do find it difficult the first few minutes of entering the game in the first item. For the last one, it could be the case that bench guys don't have first team repetition. They're not incorporated into plays when they're getting 6 minutes a night, because this is an afterthought from a team stategy standpoint. When your 9th man suddenly becomes your first shooter off the bench due to injuries, a team may need to start running plays for the guy to get involved in the offense.
Maybe there's an obvious answer to this that I'm missing, but it really does seem as if fouls are way down at the beginning of each half - meaning the reduced efficiency of starters is not just a bonus effect. Take a look at this graph. It plots the number of fouls occurring based on how many minutes have elapsed in the game (so any fouls occurring within the first minute are labeled 0 on the x-axis).
(This data through January 16th.)
It's really a striking trend. In the first minute only around 200 fouls were called, and then there's a slow upward trend to get to around what seems to be the normal number of 500. You can see that right before the end of a quarter the number jumps up a little (except for in the 4th, where it jumps up a lot) because of intentional fouls. And at the beginning of each quarter it drops - but the numbers go much lower to start each half than to start each quarter.
Is this as simple as the players being fresh, thus moving their feet and fouling less? Is it that starters are more reticent to foul because they want to stay in the game so when the bench comes in there are more fouls (but then why the drops at the beginning of quarters where it's not guaranteed at all that the starters are playing)? Maybe the refs don't call as many fouls to start the quarters to let players get into the flow?
It's interesting because I've never heard about this anywhere, really. The idea that fouls are much lower to start each half and even each quarter has been completely overlooked as far as I can tell. You'd think this would have been noticeable before, and that it would affect coaching decisions as well - maybe it would argue for using a player who can draw fouls at a high rate off the bench, once fouls are back up and you'd get the added effect of the bonus as well.
-----
Further investigation:
An obvious explanation would be that teams tend to play slower in the opening minutes of quarters, which would of course reduce fouls. So I looked at possessions per minute, and you see that borne out in the data - large dips at the quarters:
You can see that teams maximize possessions per minute at the end of quarters (taking quick shots at the buzzer), and especially at the end of games (intentional fouls), but that there's also a corresponding dip to start all quarters.
But I don't think this fully explains it. For one, it seems like it jumps back up to normal after the first minute of each quarter, which fouls don't. Secondly, the dips are exactly the same at each quarter, whereas we saw in the foul chart that they differed between the start of a half and the start of a quarter.
So let's combine the two charts now, and have a chart representing fouls per possession:
The issue still seems very much alive. You can see the same trend noticed before - very low to start the game, dips at the start of quarters but much more at the start of a half. An interesting note about it is that it slowly increases throughout the game, perhaps lending credence to the "fatigue" theory.
Joined: 30 Dec 2004 Posts: 583 Location: Near Philadelphia, PA
Posted: Sun Jan 27, 2008 8:05 pm Post subject:
Ben -
Very good stuff.
Let me ask about offensive rebounding. Your table had eFG, turnovers, and foul rates. What about OR%? I assume it's a small difference, but want to check.
If it is foul shots, the next thing is to start looking into the who and how. Are there particular teams with this being stronger than other teams? Is there an obvious reason why those teams have the pattern? What about a home/road breakdown?
I'm not sure how important this all is, but the fact that it is interesting probably means it is useful somehow... _________________ Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
Joined: 14 Jan 2005 Posts: 1712 Location: Delphi, Indiana
Posted: Mon Jan 28, 2008 7:04 am Post subject:
Great stuff, Ben.
Chicago76 wrote:
...
When I look at 4x0 and see 0, those are stats for the team with no starters vs. 4, right? That would mean that the team with no starters only commits fouls .17 times per possession...
At the low foul end of the spectrum, I see that a team of no starters tends to not foul against anyone, regardless of whether they're playing against 0 to 5 starters. Three possible explanations:
...
Well, none of those seem very likely, so I'll cut to the chase. I suspect Ben F. has put 'fouls drawn' in the 'PF' column, since his study is (or started out as) 'offensive efficiency'.
Non-starters (subs) foul at least 50% more than starters do. It's not feasible that starters foul the heck out of every sub that comes in, but take it easy on other starters. The PF numbers make sense only when they are applied to the offense, i.e., the 'team' designated.
(I assumed these aren't just fouls committed by the opposing defense, but all fouls, including offensive?)
And so, we may as well see the whole range of statistics. As Dean says, OReb; also, DReb and Ast. (I will be putting PF where I normally find them: as fouls committed.)
As subs enter the game, fouls rise. No surprise and no contradiction. _________________ 40% of all statistics are wrong.
All times are GMT - 5 Hours Goto page Previous1, 2, 3Next
Page 2 of 3
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
