Help me create a modern formula for evaluating players.

getrichordie · Post #1 » by **getrichordie** » Sun Jun 4, 2017 7:23 pm

I personally think that stats such as PER/RPM isn't close to being accurate and only tells a portion of a player's story.

I'm wanting to start a project to create a modern formula for evaluating players including their floor-spacing impact and defensive impact all in one score.

Is there anything that's already close to this? If so, how can we tweak it to become more accurate?

Some ideas:

Weight 3P% a little more heavily to match the value 3-point shooting is adding in today's league.

Decide whether a turnover or a steal should be weighted the same and if not, how should they be weighted?

--

I'm just spitballin', but I'd love to hear your input and how I can move forward with this project. Thanks!

andyhop · Post #2 » by **andyhop** » Mon Jun 5, 2017 8:02 am

You are basically talking about some form of Adjusted +/- like RAPM or RPM or PT-PM or any of the members of that family

jgoldstein34 · Post #3 » by **jgoldstein34** » Thu Jun 8, 2017 7:14 pm

Seeing as you pretty much described RAPM and RPM, curious as to what your issue with those stats is? PER isn't a good stat for evaluating defense, but it's not bad for looking at a players scoring efficiency.

aaron_gray · Post #4 » by **aaron_gray** » Sat Jun 10, 2017 2:37 pm

win shares and vorp do just fine in terms of getting the value of a player down to a single number

Crawford_81 · Post #5 » by **Crawford_81** » Thu Jul 6, 2017 2:31 am

I am 100% down to help you. Just hit me up with anything.

feyki · Post #6 » by **feyki** » Thu Jul 6, 2017 11:58 pm

Steals also has to include forced turnovers. So, they're not equal.

I have my own formula, but too lazy to bring all the work and calculate all the history.

Generally, passing/playmaking gets underrating a lot. I would love to see a metric with ranking the running the point/feeding teammates highly.

Cosmic_Backlash · Post #7 » by **Cosmic_Backlash** » Sat Jul 8, 2017 6:40 pm

I think a lot of defensive impact metrics aren't good barometers of defensive skill and ability.

For example, Andre Drummond this past season had a 99 defense rating and a DBPM of +2.9. NBA stats have it that his opponents shot 4% better when he was defending. He also only defends a very average number of shots a game. Too many metrics reward things based on traditional box score stats, not how they are actually changing the game.

I'd love a defense metric that looks how they impact team rebounding percentage (rewards people that box out but don't get actual rebound), defended shots, defended shot differential, deflections, steals, blocks per/100 possessions. This would actually represent real defensive metrics.

Offensive metrics sorely overlook passing metrics, spacing ability, screen assists, etc.

HeatFanLifer · Post #8 » by **HeatFanLifer** » Sun Jul 9, 2017 12:24 am

Here's a rough draft (not my work) on a metric called "Spacing Rating." I think this is a good place to start.

http://fansided.com/2017/07/05/nylon-calculus-spacing-rating-timberwolves/

jinxed · Post #9 » by **jinxed** » Tue Jul 11, 2017 12:06 am

getrichordie wrote:I personally think that stats such as PER/RPM isn't close to being accurate and only tells a portion of a player's story.

I'm wanting to start a project to create a modern formula for evaluating players including their floor-spacing impact and defensive impact all in one score.

Is there anything that's already close to this? If so, how can we tweak it to become more accurate?

Some ideas:

Weight 3P% a little more heavily to match the value 3-point shooting is adding in today's league.

Decide whether a turnover or a steal should be weighted the same and if not, how should they be weighted?

--

I'm just spitballin', but I'd love to hear your input and how I can move forward with this project. Thanks!

RPM already accounts for spacing, defense, and every other thing that happens on a basketball court.

andyhop · Post #10 » by **andyhop** » Sun Jul 30, 2017 11:50 pm

Cosmic_Backlash wrote:I think a lot of defensive impact metrics aren't good barometers of defensive skill and ability.

For example, Andre Drummond this past season had a 99 defense rating and a DBPM of +2.9. NBA stats have it that his opponents shot 4% better when he was defending. He also only defends a very average number of shots a game. Too many metrics reward things based on traditional box score stats, not how they are actually changing the game.

I'd love a defense metric that looks how they impact team rebounding percentage (rewards people that box out but don't get actual rebound), defended shots, defended shot differential, deflections, steals, blocks per/100 possessions. This would actually represent real defensive metrics.

Offensive metrics sorely overlook passing metrics, spacing ability, screen assists, etc.

Its a shame Gotbuckets folded because they used to have four factor numbers produced using the APM method to produce this

sniper236 · Post #11 » by **sniper236** » Mon Aug 7, 2017 9:17 pm

I'm trying to create my own set of stats for the 2017 Finals (mainly for my own curiosity into the influence of KD compared to Curry, LeBron compared to Kyrie). But the defensive breakdowns are so tough to put any kind of objectivity into.

Even when looking over a season, how are you supposed to measure that without vastly generalised team defensive numbers?

Every way I've tried to look at it is flawed because there are too many variables to assign a rating to one player.

dhsilv2 · Post #12 » by **dhsilv2** » Sun Aug 13, 2017 8:58 pm

jinxed wrote:
getrichordie wrote:I personally think that stats such as PER/RPM isn't close to being accurate and only tells a portion of a player's story.

I'm wanting to start a project to create a modern formula for evaluating players including their floor-spacing impact and defensive impact all in one score.

Is there anything that's already close to this? If so, how can we tweak it to become more accurate?

Some ideas:

Weight 3P% a little more heavily to match the value 3-point shooting is adding in today's league.

Decide whether a turnover or a steal should be weighted the same and if not, how should they be weighted?

--

I'm just spitballin', but I'd love to hear your input and how I can move forward with this project. Thanks!

RPM already accounts for spacing, defense, and every other thing that happens on a basketball court.

It does in theory....if all values are maximized towards the goal of winning, then sure. If however a player has elite skills that aren't used in the system, then that is missed. It is however possible that we could identify these under under utilized realities. But then it wouldn't fit into a predictive win model and would make for a bad one stat for predictive power.