LA Bird wrote:Doctor MJ wrote:First thing I want to say is that I hope I'm not coming across too negative. I think you clearly have the tools to do awesome things even if I'm not particularly excited about this specific lens.
We're cool. New methodologies should be able to stand up to scrutiny anyway.
Glad to hear it LA Bird!
LA Bird wrote:Re: "using points per possession to directly compare teams playing at different paces...swimmers different events". No it doesn't, because basketball players are competing in the same event.
A more apt comparison here would be to compare freestyle, to backstroke, to butterfly, etc, because what we're actually talking about here are basketball teams choosing strategy. The difference between the sports is largely that swimming still awards medals for sub-optimal strokes whereas in basketball, if you choosing not attack as optimally as your opponents, you just get smoked.
Comparing the different strokes is more like comparing basketball to netball. The focus is on distance vs time, not stroke vs time. Minimizing time given a certain distance is the same idea as maximizing points given a certain number of possessions. Teams can choose the pace they play at just as swimmers can choose the distance they compete in. But whatever number of possession or distance they choose, their opponent will also be around the same level. You won't have a guy in the 50m freestyle competing against a 100m freestyle. And you won't have a NBA team getting 80 possessions in a game while the opponent gets 120. Any per possession offensive efficiency difference across total possessions is very small on a game by game level.
Also, an optimal basketball strategy includes both offense and defense, not just offense only. There is no correlation between MOV and possessions played so on average, nobody is smoking their opponent just by playing fast.
Okay we need to make ontological distinctions here when you're saying stuff like "Teams can choose the pace they play just as swimmers can choose the distance they compete in". Yes there's volition involved, but one of these is a choice that is made long before the event in question, and the other choice is made during the game when looking to attack the other team.
I'll put it another way:
When you choose a swimming distance, you choose your competitors.
In the NBA, your goal is to score more points per possession than the team that's in front of you, and any choices you made within the context of beating that team falls within the realm of strategy. I think anything that looks to normalize away the effects of strategy when evaluating play is something to be careful with.
Re: Netball. I'm no expert on netball but to me it largely looks like basketball without dribbling. Thought it worth noting that you can play basketball using this strategy and that this was once the dominant way the best professional teams looked to play in the half court. Only reason we don't see it any more is that it's been supplanted by more effective strategy.
Netball then represents essentially a rule-handicapped version of basketball that forces inferior strategy a bit like soccer is a sport for humans where the rule is a "don't use the parts of the body that make humans special" handicap.
Not making any particular argument with that observation - I actually love soccer - just thought it was interesting.
LA Bird wrote:I'm not saying that you're hard-coding a penalty for playing fast, but if you're not suggesting that we should normalize ORtg by Pace, then yes, I'm confused by your intentions here.
Re: "If league ORtg goes up, that doesn't mean everyone's more efficient." It literally does mean that though.
If every steel factory starts using a better technique at some point on the production line, then they all get more efficient. I understand you're trying to say "relative to each other", but the word efficiency used here precisely because it's talking about how much bang you're getting for your buck in any given possession.
Re: "A strategy which only improves raw efficiency but not relative efficiency is not an efficient strategy at all." I would completely disagree with your characterization because it implies that when everyone starts using the superior strategy it stops being a superior strategy compared to what came before.
Consider the Fosbury Flop in the high jump. By virtue of it being a glaringly superior strategy that rendered all other strategies forever stupid in comparison, relatively quickly all the major jumpers adopted the new technique. By your approach and semantic labeling you'd suggest that this means that the Fosbury Flop was efficient only briefly at the moment when most jumpers hadn't adopted it yet, and as more jumpers adopted it, it lost its efficiency.
When the participants in the sport running in the direction of a particular strategy is improving all of their results, you coming in and saying "But you're not being efficient because everyone else is doing the same thing" just doesn't make sense.
You misquoted me. Everyone is more efficient =/= everyone is efficient. The first statement compares every team in a current situation against themselves in an earlier situation. This is absolute efficiency. It is possible for every team to be more efficient in absolute terms than they were before. The second statement compares every team against each other in the same current situation. This is relative efficiency. By definition, it is impossible for every team in a year to be relatively efficient, regardless of absolute efficiency. Some teams will be relatively efficient and some teams will be relatively inefficient. Relative ORtg itself is a relative efficiency stat.
The Fosbury flop is an improvement in
absolute efficiency over previous techniques. It also provided a
relative efficiency advantage for those few athletes who were the first to adopt it. Once everybody learned to use the technique and they are all on a level playing field, the Fosbury flop no longer gave a
relative efficiency advantage to anyone, not even Fosbury himself. But the technique itself still holds an
absolute efficiency advantage over previous methods. There is a key difference between absolute and relative efficiency. For basketball, both teams always play around the same possessions each game. They are on a level playing field already in terms of possessions so there is little relative efficiency advantage to be gained through any Fosbury flop like strategy when it comes to exploiting efficiency differences over different total possessions.
Did I misquote you? What you say after that statement seems to say that you just used ambiguous language before and thus your point got confused. I like your use of absolute vs relative efficiency here. Effective disambiguation.
You're losing me again at the end though.
The Fosbury Flop exists in a sport that is highly analogous to the possession-based nature of basketball.
The advantage of the Fosbury Flop is that it will allow you to score more effectively in one attempt.
The advantage of offensive basketball strategy is that it will allow you to score more effectively in one possession.
LA Bird wrote:It's not a separate topic though. Consider:
If basketball were a sport where offense and defense were completely independent of each other, then it wouldn't make any sense at all to just a team's offense by anything other than ORtg, right?
I'm pointing right to the core thesis of why you started doing this statistical analysis in the first place, so it really is on topic even though it focuses on something that was not where you were directing the brunt of your energy in the OP.
I still don't know what you are trying to get at here. Yes, we know offense and defense are two sides of the same coin in basketball. But the core thesis of this analysis is on how league average scoring should be defined, not on balancing offense-defense tradeoffs. Once the league average line is fitted, how I evaluated offenses and defenses relative to the league average is the exact same process as relative ORtg/DRtg. So what exactly is the problem there?
If I can use a new strategy on offense that improves ORtg (while also happening to affect the Pace in some way), and this new technique has no effect on defense, then why would we attach some kind of effective cost to this technique?
LA Bird wrote:The existing stat tells us how many points you score per possession.
Can you explain your stat so succinctly without using any words more esoteric than what I just used? I cannot, but you go ahead if you can.
Points scored relative to the league based on possessions played. It's the residuals in a y = bx + c model where y is points and x is possessions. This is standard approach for modeling the relationship between two variables. I would actually turn the question back to you and ask how you explain the logic behind relative ORtg. What is the reason to take the average of all the individual slopes (y/x) to model x and y? I don't know of any other time when this approach is taken instead of just fitting a linear model.
I'd note that your explanation was more complicated than mine, as it couldn't help but be.
Re: "instead of just fitting a linear model". When you try to make the existing standard look like it's more work than your model you paint a problematic picture because it's just not the reality. You're the one trying to add an extra adjustment into the mix.
Re: "standard approach for modeling the relationship between two variables". Uh huh, and why do those techniques exist? To try to help us to identify correlations which for which we can posit causal theory for.
What I'm saying is that when you make the adjustment you do based on the correlation here, you're essentially assuming that Pace is acting as Cause while ORtg is the Effect. And I'm saying I object to this assumption.
LA Bird wrote:What I'm looking to say about your stat is that I don't see a justified causal explanation for why a given team's ORtg should be looked upon as effectively inflated simply because they are playing at a pace that correlates with higher ORtg in general.
To put it one other way, in my conversation with Odinn we reached a disconnect based on the fact that I was assuming that he was thinking that Pace could be effectively attributed as a Cause of higher ORtg, but in reality he wasn't concerned about Causality at all.
What is the justified causal explanation for why a given team's ORtg should be looked upon as inflated simply because they are playing in a league that correlates with higher ORtg in general? Are we concerned about establishing causality in this scenario first when we are currently using relative ORtg?
Hmm. WelI, in general, I am reluctant to talk about era-offense being "inflated" simply because ORtg's are higher when I believe the shift to be do to strategic and skill-based improvement, as I think it paints the wrong picture.
I'm all for the use of league-average-adjusted rORtg & rDRtg for a variety of purposes - most notably getting your bearings within the context of understanding a particular season - but I would want to avoid directly attaching the concept of "inflation" here, because people tend to think "inflation" means counterfeit in some way.
And to put it in concrete: I don't think the GOAT offense is necessarily the one with the best ORtg or best rORtg. I like both stats, but I think the answer is more complicated than using one or the other, or frankly some new stat that tries to directly use both together.
I feel like I do need to end by making clear however that I see your pace adjustment as different from a league-average-ORtg adjustment, and I'm trying to think of something to say here to boil it down. I'll go back to swimming:
Let's say that we get some wave quantification detector that measures the amount of energy a swimmer gives to the water around him as he swims.
Let's say that they find that this energy-dissipation has outstanding correlation with how fast you swim, to the point where you can tell who won the race simply by looking at the energy dissipation data.
That means means you do all the things with those two variables - time and energy dissipation - that you could with any other two variables in statistics, and that also means you can normalize time performance with respect to energy of dissipation.
Now, the question: What would you try to use your new relative time score for?
I'm not saying it has no use, but what I am saying is that it would never make sense to argue that the person with the relative time score was a better swimmer than the guy who actually won the race.
Why? Most simply: Because the race is the thing. Your job is to go as fast as you can, however you can do it, within the parameters of the rules.
But also: Because the energy dissipated by the swimmer was done as a by product of him trying to swim as fast as he could. It's thus an Effect created by the Cause we're looking to study, and thus we don't want to normalize it away.
