APBR-DraftExpress 2015 NBA Draft Project

[vjl110]

I'd say the minimum should be:

1. Make DV very clear. What is your model trying to predict?

2. What is the key thing that likely distinguishes your model from the others.

I don't plan on giving the actual coefficient weights, or going into how I handle some of the more subtle steps in the process.

[ampersand5]

If the explanation of the model in some or many cases is limited to mentioning most or all the variables included but not the exact formula with all the individual weights then there won't be enough shared information to really analyze all and improve the methodologies as much as with that information (unless someone or some stat package is really good at reverse engineering). It is a voluntary choice to reveal but do the participants want to push each other on at least a certain minimum level of openness or possibly total openness?

Is the agreement though to just share rankings? It was mentioned that different ratings formats could muddy things. They also could help more if kept on same scale or revealed in their distinctions. Perhaps more could be revealed optionally offsite by the authors at their own locations or in future contests?



Amp, in the write up of the contest, perhaps ask that the reference is to "members of the APBRmetrics bulletin board" and not loosely to "APBR", as that is a pre-existing and distinct association. I dunno if anyone would actually mind or if many care about the distinction but why not be precise.

Since some or many of the participants may be open to work with a team or major media, maybe the submissions- and the writeup- should include a sentence or two about the author's background, at their option. I assume that was likely; but if desired, why not specifically ask for it ahead of time instead of at last minute.

Is there going to be an official, planned scoring of the methods after 1/2 season, 1, 2, 3 of more years? At DX, here or elsewhere? By BPM, RPM, a blend? I hope a composite avg. ranking is calculated and tracked along with the individual ones. (If a composite is included, the article could mention where the suggestion came from, unless it was already planned.)

Anyone want to do a podcast (or more) to support the article? Any chance Grantland basketball show invites a couple to tape something about it or appear?

Any chance a sponsor- team, major media or other- could offer a prize in future?

What's on DE is not there to supplant APBR. I hope this will serve as a launching pad for further discussion.

You raise a good point that I completely forgot to bring up before - how should everyone be rated. If its possible to have everyone using the same units, that would obviously be ideal, but I understand that this likely is not possible. Can some of the people doing their models now speak to the feasibility of this?

At the very least, I will make a combined/consensus ranking based on how each player is ranked (rather than rated) across the board.

Yup, the write up will refer to the APBR board, as hopefully this can increase awareness of this community.

There will forsure be a short bio on each individual with a link to further information on them.

[DSMok1]

Excellent work!

Everyone should be ranked, not rated. If all of the systems are using the same set of players, simply ranking them is both adequate and the easiest to explain to the average reader.

The descriptions of the models for the Draft Express articles should be quite simple and straightforward, just a couple of paragraphs that explain how the model might be different from other models, and what its strengths and weaknesses are. There should also be a link in the descriptions to some write up of more details about the model and that is where any hard math details should go.

[vjl110]

I am trying to decide whether I should use my EWP model or a combination of the EWP and HUM model (just the average of the two). I personally weight the combined score higher than any individual model, but it might seem a little weird to throw a combined model into the mix.

Any thoughts on which approach would be preferable?

[DSMok1]

I am trying to decide whether I should use my EWP model or a combination of the EWP and HUM model (just the average of the two). I personally weight the combined score higher than any individual model, but it might seem a little weird to throw a combined model into the mix.

Any thoughts on which approach would be preferable?

It should be whatever you, the modeler, prefers as the best solution for your model.

[ampersand5]

I am trying to decide whether I should use my EWP model or a combination of the EWP and HUM model (just the average of the two). I personally weight the combined score higher than any individual model, but it might seem a little weird to throw a combined model into the mix.

Any thoughts on which approach would be preferable?

I understand your concerns and ultimately, if you have a preference for one of the choices, then pursue that.

However, I think for the purpose of this, you are correct and it's probably better to have one independent model rather than a blend.

[Crow]

The goal is the best performing model, right? Blend of numbers and eyes seems like a good approach to me, though still has to be proven. Could be less likely to miss badly. Might be more sellable / acceptable to folks not comfortable with a purely mathematical approach. It is a philosophic, strategic choice. You'll be able to score both whichever you chose though and I hope you will report on which was better.

[vjl110]

I'm putting together my list and had a couple comments on the player-set.

George Lucas is an international prospect, so he probably shouldn't be on there.

Also... I completely agree that the focus should be on likely draftees, but it would be fun to see some of the "deep-sleepers" that pop out of different models. Not sure how best to structure that though. Maybe a top-N likely undrafteds list or something.

[Statman]

I'm putting together my list and had a couple comments on the player-set.

George Lucas is an international prospect, so he probably shouldn't be on there.

Also... I completely agree that the focus should be on likely draftees, but it would be fun to see some of the "deep-sleepers" that pop out of different models. Not sure how best to structure that though. Maybe a top-N likely undrafteds list or something.

If I'm involved - I'll include the guys that "break" the above list. Somebody not listed won't project as a "top" prospect at all, but someone might break the top 20 overall in my model. We'll see.

[vjl110]

"If I'm involved - I'll include the guys that "break" the above list. Somebody not listed won't project as a "top" prospect at all, but someone might break the top 20 overall in my model. We'll see."

I am saying we should find a way to note those guys, but in the interest of standardizing it should be outside of ranking the set list.

[nrestifo]

When we're ready to submit our rankings/ratings and description, who should we send this to?

[ampersand5]

I'm putting together my list and had a couple comments on the player-set.

George Lucas is an international prospect, so he probably shouldn't be on there.

Also... I completely agree that the focus should be on likely draftees, but it would be fun to see some of the "deep-sleepers" that pop out of different models. Not sure how best to structure that though. Maybe a top-N likely undrafteds list or something.

my apologies about George Lucas; most definitely a mistake on my end.

In regards to your second point - please let me know about any players that come up. First of all, its possible that everyone is ranking certain unlisted players that high and they should be included on the list. Moreover, even if its independent, its still worth likely including in your write up.

[ampersand5]

When we're ready to submit our rankings/ratings and description, who should we send this to?

It would be ideal if everyone would post their rankings on here, but otherwise, everything should be submitted to me so I can organize everything.

[vjl110]

Top guys by my numbers who aren't listed...

Seth Tuttle
Briante Weber (awkward b/c of injury)
Ousmane Drame
Royce O'Neale
Rayvonte Rice
Kevin Pangos
Damion Lee

Tuttle is the only guy who the numbers argue should be considered in the 1st round, but the others look like good to decent 2nd round selections. I'll post mine soon. I just need to figure out where I have a name inconsistency with TaShawn Thomas.

[nrestifo]

I was about to mention George Lucas, as I'm looking at my stuff now. Should we replace him with Seth Tuttle?