The Race To The Bottom - theBigLip Tank Machine

[theBigLip]

As optimistic as I like to be and hope the Pistons make the playoffs, I don't think that's going to happen. So I like looking at what that will look like for us. In past years midway through the season I've usually done a tank thread to see where we will end up in the lottery (or at least how many ping pong balls we will get). With Victor as the prize, I said WTF, and decided I'm going to track things the entire season.

I know there are other trackers such as Tankathon that are pretty good. But it's a black box and we don't know the logic. So I created a model that has played out pretty well, at least in relative terms. So here is my logic:
- teams are either Great, Good, or Bad
- Great teams beat Good and Bad, Good teams beat Bad ones
- If Great teams play each other, the home team wins. Same for Good and Bad teams playing their own.
- I don't do it now, but I'll look into adding some back-to-back logic. Those second games in back-to-backs are a little tougher.

So to make the model work, the teams need to be categorized. There are some questionable ones like Washington and Portland. But if they need to be changed, the good thing the model (and my spreadsheet) update automatically. Let me know what you think, but here is what I have to start:
Atlanta Good
Boston Great
Brooklyn Great
Charlotte Bad
Chicago Good
Cleveland Good
Dallas Great
Denver Great
Detroit Bad
Golden State Great
Houston Bad
Indiana Bad
LA Clippers Great
LA Lakers Good
Memphis Great
Miami Great
Milwaukee Great
Minnesota Good
New Orleans Good
New York Good
Oklahoma City Bad
Orlando Bad
Philadelphia Great
Phoenix Great
Portland Good
Sacramento Good
San Antonio Bad
Toronto Good
Utah Bad
Washington Good

So before we even play a game, it seems some teams have a slightly easier or harder schedule. Here are the predictions:
Team W L
Houston 12 70
Oklahoma City 12 70
San Antonio 11 71
Utah 11 71
Orlando 11 71
Detroit 11 71
Charlotte 11 71
Indiana 10 72

I would think some of these will be going up, but this will be a serious tank race. I'll update weekly. Regardless of where we land, we still need to get lucky in the lottery, but the more ping pong balls the better. Let the race to the bottom begin!

[Billl]

Some offense intended, but this is stupid. Your "model" has 8 teams finishing with 12 or less wins? The worst record last year was 20-62.

[theBigLip]

Some offense intended, but this is stupid. Your "model" has 8 teams finishing with 12 or less wins? The worst record last year was 20-62.

I know. I've never done it this early. Upsets will happen over the course of the year. By the time those are embedded into the records, it becomes more accurate. And it is also relative in nature, so if every team is good for 5-8 upsets a year, then all these win totals just bump up. One thing it does do pretty well is capture strength of schedule. If one team finishes the year with a lot of road games, that will reflect in the predicted standings.

So if this is so stupid, can you come up with something better? Or a logical way to capture the "upsets"?

[kellmellus50]

Playoffs for the Pistons, I'm drinking the kool-aid

[bstein14]

Brooklyn goes from great to good.
Dallas goes from great to good.
Miami goes from great to good.
One of Phoenix, Denver or Memphis goes from great to good. Not sure which quite yet.

Each conference should have at most 3 "great" teams IMO. YOu're projecting all of those teams to have great win/loss records.

[Billl]

Some offense intended, but this is stupid. Your "model" has 8 teams finishing with 12 or less wins? The worst record last year was 20-62.

I know. I've never done it this early. Upsets will happen over the course of the year. By the time those are embedded into the records, it becomes more accurate. And it is also relative in nature, so if every team is good for 5-8 upsets a year, then all these win totals just bump up. One thing it does do pretty well is capture strength of schedule. If one team finishes the year with a lot of road games, that will reflect in the predicted standings.

So if this is so stupid, can you come up with something better? Or a logical way to capture the "upsets"?

Yes, but in real life I'm a data scientist. With zero effort, I could predict that every team has the exact same record as last year and it would be way more accurate than your model.

[Manocad]

Some offense intended, but this is stupid. Your "model" has 8 teams finishing with 12 or less wins? The worst record last year was 20-62.

I know. I've never done it this early. Upsets will happen over the course of the year. By the time those are embedded into the records, it becomes more accurate. And it is also relative in nature, so if every team is good for 5-8 upsets a year, then all these win totals just bump up. One thing it does do pretty well is capture strength of schedule. If one team finishes the year with a lot of road games, that will reflect in the predicted standings.

So if this is so stupid, can you come up with something better? Or a logical way to capture the "upsets"?

Yes, but in real life I'm a data scientist. With zero effort, I could predict that every team has the exact same record as last year and it would be way more accurate than your model.

I always pictured you as a state trooper or maybe an IRS auditor.

Kidding

[theBigLip]

Some offense intended, but this is stupid. Your "model" has 8 teams finishing with 12 or less wins? The worst record last year was 20-62.

I know. I've never done it this early. Upsets will happen over the course of the year. By the time those are embedded into the records, it becomes more accurate. And it is also relative in nature, so if every team is good for 5-8 upsets a year, then all these win totals just bump up. One thing it does do pretty well is capture strength of schedule. If one team finishes the year with a lot of road games, that will reflect in the predicted standings.

So if this is so stupid, can you come up with something better? Or a logical way to capture the "upsets"?

Yes, but in real life I'm a data scientist. With zero effort, I could predict that every team has the exact same record as last year and it would be way more accurate than your model.

I’m in tech as well and if you want to wager your lazy ass model vs mine, let’s bet. Also one of my initial disclaimers that it will show relative predicted records, which is really the important thing. Do you care if Houston gets 13 wins, or that they finish w the worst record?

[theBigLip]

Brooklyn goes from great to good.
Dallas goes from great to good.
Miami goes from great to good.
One of Phoenix, Denver or Memphis goes from great to good. Not sure which quite yet.

Each conference should have at most 3 "great" teams IMO. YOu're projecting all of those teams to have great win/loss records.

Dallas and Miami are tough teams to categorize. Probably somewhere in between.

I also think Portland and/or Washington will eventually realize they are not making the playoffs and start to tank, so they should be “bad”.

[ElectricMayhem]

Brooklyn is the most obvious great to good candidate. They have no depth, awful defense outside of Simmons, and apathetic/untrustworthy stars.

[Snakebites]

There are 6 teams in NBA history that have failed to get 13 wins in an 82 game season, and we can go ahead and add the 7-59 Bobcats in the shortened season that had a winning percentage low enough to qualify- that makes 7 teams in NBA history that have done that badly . You have a greater number than that doing it in one single season.

[Manocad]

I want to bet on something.

[theBigLip]

There are 6 teams in NBA history that have failed to get 13 wins in an 82 game season, and we can go ahead and add the 7-59 Bobcats in the shortened season that had a winning percentage low enough to qualify- that makes 7 teams in NBA history that have done that badly . You have a greater number than that doing it in one single season.

I agree with what you are saying but…

Each bad team is not going to be favored in any of the games they are modeled to lose. And each team is going to win 5-15 of them. How can you model/predict them? If you have a great answer, I’m taking it to Vegas

So in the meantime, the relative ranking counts. Utah somehow beat Denver last night. That’s going to bump them up against the other tankers. And as the season progresses, the predicted win totals will get more realistic.

[theBigLip]

I want to bet on something.

I’m down

[theBigLip]

So maybe I should have waited until there are a few months left in the season to roll this out like I have in the past. I was curious what it would look like, but I guess the result shouldn’t be surprising. It currently reflects the games a team will be favored more so than their final record. That will change as the season progresses.

[topsearch92]

So maybe I should have waited until there are a few months left in the season to roll this out like I have in the past. I was curious what it would look like, but I guess the result shouldn’t be surprising. It currently reflects the games a team will be favored more so than their final record. That will change as the season progresses.

I’ve always enjoyed this annual prediction post. I’m sure with the Victor tankathon your numbers may end up accurate for some hopeful teams.

[Manocad]

So maybe I should have waited until there are a few months left in the season to roll this out like I have in the past. I was curious what it would look like, but I guess the result shouldn’t be surprising. It currently reflects the games a team will be favored more so than their final record. That will change as the season progresses.

I see both sides. I get the relative ranking and a model that doesn't account for upsets. But I also got the idea that there won't be 8 teams with less than 20 wins when 20 wins was the worst last season.

Now, does that mean you bump all 11-12 wins teams by 9-10 wins to shift all the records to 20-22 wins instead but with the same rankings? I don't know.

That all being said, for sure I'm more down with predicting (and betting on) relative finishes rather than number of wins.

[Snakebites]

There are 6 teams in NBA history that have failed to get 13 wins in an 82 game season, and we can go ahead and add the 7-59 Bobcats in the shortened season that had a winning percentage low enough to qualify- that makes 7 teams in NBA history that have done that badly . You have a greater number than that doing it in one single season.

I agree with what you are saying but…

Each bad team is not going to be favored in any of the games they are modeled to lose. And each team is going to win 5-15 of them. How can you model/predict them? If you have a great answer, I’m taking it to Vegas

So in the meantime, the relative ranking counts. Utah somehow beat Denver last night. That’s going to bump them up against the other tankers. And as the season progresses, the predicted win totals will get more realistic.

Yeah, I get it. It is a fun exercise later in the season when strength of schedule actually starts differing more significantly between different teams.

But here it kind of gives off the feeling of attempting to use a simple formula to solve what is a rather complex issue.

[breezypeezy]

Im not feeling tbe Brooklyn as "Great".
Based on what?
I would take Minnesota (ranked Good) over the Nets in a straight up contest.

[Piston Pete]

Thinking we finish the season with the 8th-worst record. Not high lottery, no playoffs.