Jump to content
NBC Sports Edge Forums

Sleepers and players to avoid 2017-2018


Recommended Posts

15 hours ago, v1n5anity said:

I agree it's not a perfect measurement. However, let's look at it practically. Gobert's 2.6 blocks per game are very useful, but he won't single handily win you the category. If a player scored 72-91 points per game, there's no way you lose that category. For example, if you have 10 active players you need to get about 150 points per game combined (10 players * 15 ppg). If 1 player scored 91 points per game, the remaining 9 players only need to average about 6.5 points per game, which is basically impossible not to meet since almost every single relevant fantasy player scores more than that. And if that player averaged 72 points per game, then the remaining players need to score 8.7 ppg - also a figure basically impossible not to achieve.
 

As for the median number you calculated, that may be correct, however as I mentioned before I only play roto (8-cat) and in roto you don't really look at median or average of all players, what you really want to aim for is to rank 3rd in every category - this will essentially set you up to win the league. Zack Rewis breaks it down perfectly here: http://thefantasyfix.com/fantasy-basketball/2015-16-fantasy-basketball-rotisserie-draft-strategy/. The number he posted here if you want to place 3rd in blocks is 0.74 blocks per game. So if we have Gobert averaging 2.6 blocks per game, the other 9 active players need to still average 0.53 blocks per game, which isn't super hard to do, but no where near as a lock as the 6.5 ppg or 8.7 ppg example above. That's why it's inaccurate to say 2.6 blocks is equivalent to 72-91 ppg.

 

There's a saying in statistics circles: All models are wrong, but some are useful. By that I mean that your offering an alternative approach does not disprove the usefulness of mine, and if yours works for you, I've got no reason to argue against it. Still, using the numbers in that article, if you are looking for 0.73 bpg, then Gobert is getting in enough blocks to cover more than 3.5 players, more than a third of your target. That would be the equivalent of around 52 ppg or 4260 for the season, or 1800 more than Westbrook actually tallied. The rest of your team would then only need to average ~10.6 ppg for you to hit your target. Sound more credible?

 

I think that the specific numbers don't matter as much, when the point I was making was about z-scores and the problems they have grappling with basketball statistics. What happens in the calculation is something like this: they calculate the mean, which you stated was about 15 for points and 0.63 for block. Then they find the standard deviation, let's say 5 points and 0.6 blocks (and I don't actually have these numbers handy, but these are close enough to make the point). Then they check how many standard deviations each player is above or below the mean in each category, and it is the sum of these standard deviations that comprises the player's z-score. But notice what happened with points and blocks there. A player scoring 30 ppg would be 3 standard deviations above the mean. A player swatting 2.43 bpg would also be 3 stdevs above, leading to statements like yours that Gobert's blocking is worth about the same as Westbrook's scoring. However, 30 ppg is only twice the average, while 2.43 bpg is 3.85 times the average. Not every standard deviation is equal--which, in this example, is a consequence of the heavy right skew in blocks--and the appearance that they are equal is only one of z-scores' distortions.

 

I find myself wondering how easy it is to actually achieve the target numbers from the article in a fantasy draft. Is it normal for your to be able to reach them, across the board? I'm used to a deeper league (13-15 players per team), so they look kind of out of reach to me.

 

9 hours ago, pdids911 said:

 

AD over giannis? its Kd, giannis, then maaaybe AD at 3 for me. 

 

I rank AD second last year, both per game and for the season entire, with Antetokounmpo sitting at fourth in both rankings. With AD's . . . vulnerability, an argument could be made for either at second in the draft, but I try to avoid players the season after their big breakout because they tend to be overhyped, defenses adjust, injured teammates come back, etc. (See under Bazemore, Kent.)

Link to post
Share on other sites
  • Replies 579
  • Created
  • Last Reply

Top Posters In This Topic

1 hour ago, Meadowlark said:

 

There's a saying in statistics circles: All models are wrong, but some are useful. By that I mean that your offering an alternative approach does not disprove the usefulness of mine, and if yours works for you, I've got no reason to argue against it. Still, using the numbers in that article, if you are looking for 0.73 bpg, then Gobert is getting in enough blocks to cover more than 3.5 players, more than a third of your target. That would be the equivalent of around 52 ppg or 4260 for the season, or 1800 more than Westbrook actually tallied. The rest of your team would then only need to average ~10.6 ppg for you to hit your target. Sound more credible?

 

I think that the specific numbers don't matter as much, when the point I was making was about z-scores and the problems they have grappling with basketball statistics. What happens in the calculation is something like this: they calculate the mean, which you stated was about 15 for points and 0.63 for block. Then they find the standard deviation, let's say 5 points and 0.6 blocks (and I don't actually have these numbers handy, but these are close enough to make the point). Then they check how many standard deviations each player is above or below the mean in each category, and it is the sum of these standard deviations that comprises the player's z-score. But notice what happened with points and blocks there. A player scoring 30 ppg would be 3 standard deviations above the mean. A player swatting 2.43 bpg would also be 3 stdevs above, leading to statements like yours that Gobert's blocking is worth about the same as Westbrook's scoring. However, 30 ppg is only twice the average, while 2.43 bpg is 3.85 times the average. Not every standard deviation is equal--which, in this example, is a consequence of the heavy right skew in blocks--and the appearance that they are equal is only one of z-scores' distortions.

 

This is a great post illustrating the problem with using z-scores; especially with weighting them all equally across categories to generate a ranking.

To check this out I made histograms of the 6 primary counting stats (apologies to TO) from the BBM top 168 players (9-cat) in 2016-2017, in order to eyeball the distributions. As you mention, none of them are looking remotely normal, and they are all positively skewed to varying degrees.  This has me curious if there would be a better way for a place like BBM to adjust their ranking system to account for these statistics (standard deviation, skewness, etc.) Any stats geeks out there with ideas?

0oXRSKT.png

 

Link to post
Share on other sites
8 minutes ago, jay14bay said:

 

This is a great post illustrating the problem with using z-scores; especially with weighting them all equally across categories to generate a ranking.

To check this out I made histograms of the 6 primary counting stats (apologies to TO) from the BBM top 168 players (9-cat) in 2016-2017, in order to eyeball the distributions. As you mention, none of them are looking remotely normal, and they are all positively skewed to varying degrees.  This has me curious if there would be a better way for a place like BBM to adjust their ranking system to account for these statistics (standard deviation, skewness, etc.) Any stats geeks out there with ideas?

0oXRSKT.png

 

I know a way to do it conceptually but I'm not good enough at math to write a program for it.  The idea is to focus not only on the median and standard deviations from it, but also on polarity and scarcity.  Polarity is defined by how much difference there is between the worst players and the best players and whether the cat is polarized or not.  Is it a cat in which most players either produce heavily in that stat or not much at all (e.g. blocks/assists) or is it like points and rebounds where everyone produces a little?  I would assign more value to big producers in polarized cats because one or two such producers would have you dominating that cat.  Scarcity is defined by how many players exceed the median, if few players exceed the median then producers in scarce cats also have more value.

Link to post
Share on other sites
2 hours ago, Meadowlark said:

There's a saying in statistics circles: All models are wrong, but some are useful. By that I mean that your offering an alternative approach does not disprove the usefulness of mine, and if yours works for you, I've got no reason to argue against it. Still, using the numbers in that article, if you are looking for 0.73 bpg, then Gobert is getting in enough blocks to cover more than 3.5 players, more than a third of your target. That would be the equivalent of around 52 ppg or 4260 for the season, or 1800 more than Westbrook actually tallied. The rest of your team would then only need to average ~10.6 ppg for you to hit your target. Sound more credible?

 

I think that the specific numbers don't matter as much, when the point I was making was about z-scores and the problems they have grappling with basketball statistics. What happens in the calculation is something like this: they calculate the mean, which you stated was about 15 for points and 0.63 for block. Then they find the standard deviation, let's say 5 points and 0.6 blocks (and I don't actually have these numbers handy, but these are close enough to make the point). Then they check how many standard deviations each player is above or below the mean in each category, and it is the sum of these standard deviations that comprises the player's z-score. But notice what happened with points and blocks there. A player scoring 30 ppg would be 3 standard deviations above the mean. A player swatting 2.43 bpg would also be 3 stdevs above, leading to statements like yours that Gobert's blocking is worth about the same as Westbrook's scoring. However, 30 ppg is only twice the average, while 2.43 bpg is 3.85 times the average. Not every standard deviation is equal--which, in this example, is a consequence of the heavy right skew in blocks--and the appearance that they are equal is only one of z-scores' distortions.

 

I find myself wondering how easy it is to actually achieve the target numbers from the article in a fantasy draft. Is it normal for your to be able to reach them, across the board? I'm used to a deeper league (13-15 players per team), so they look kind of out of reach to me.

 

52 ppg is definitely more credible than 72-91 ppg IMO. I would love to see some model that properly measures blocks against other categories - as I said before I think they are the most scarce category. They have a very wide range out outcomes: ranging from less than one-seventh of the target value (0.1 bpg) to more than 4 times (2.8+ bpg).

 

As for hitting the target numbers, I play in leagues with 16 players per team, 12 actives. I'm not sure you will come out the draft with your 16 players averaging those numbers, so I typically just look at my top 12 players to hit the target numbers. Most likely you'll have a couple duds so some free agent will come along that will boost your averages for your bench. They are definitely achievable by the end of the year as I was able to do it in one of my leagues last season.

Link to post
Share on other sites

Some great discussion here guys. I think this stuff is not talked about enough. I definitely agree that Z-Scores aren't perfect - I talked about it before that certain categories are more important in earlier rounds than others - but it's the best I've found so far. Certain categories are more scarce than others and Z-Scores don't really take that into account. Would be great to see if someone could come up with a better measurement of value that takes into account statistical scarcity.

Edited by v1n5anity
Link to post
Share on other sites
1 hour ago, jay14bay said:

 

This is a great post illustrating the problem with using z-scores; especially with weighting them all equally across categories to generate a ranking.

To check this out I made histograms of the 6 primary counting stats (apologies to TO) from the BBM top 168 players (9-cat) in 2016-2017, in order to eyeball the distributions. As you mention, none of them are looking remotely normal, and they are all positively skewed to varying degrees.  This has me curious if there would be a better way for a place like BBM to adjust their ranking system to account for these statistics (standard deviation, skewness, etc.) Any stats geeks out there with ideas?

0oXRSKT.png

 

 

Whoa, you know it's a party when someone busts out the histograms!

 

Interesting visual observation is that 3pm has a staggering amount of absolute 0 contributers compared to any other CAT. 

Link to post
Share on other sites
1 hour ago, StifleTower2 said:

I know a way to do it conceptually but I'm not good enough at math to write a program for it.  The idea is to focus not only on the median and standard deviations from it, but also on polarity and scarcity.  Polarity is defined by how much difference there is between the worst players and the best players and whether the cat is polarized or not.  Is it a cat in which most players either produce heavily in that stat or not much at all (e.g. blocks/assists) or is it like points and rebounds where everyone produces a little?  I would assign more value to big producers in polarized cats because one or two such producers would have you dominating that cat.  Scarcity is defined by how many players exceed the median, if few players exceed the median then producers in scarce cats also have more value.

 

I have a way of doing my own rankings in a spreadsheet. I have had vague hopes of monetizing it, but at this point I might be happy just to get a site or a program that does it for me and saves me an hour or two each week. The basic gist is that it takes all the categories including TOs and percentages and converts them to point equivalents by comparing the individual's production in a category against the median (a variation of a trick I learned playing World of Warcraft, of all things, where you sometimes have to compare similar pieces of equipment to determine which stats will improve your character the most). I have never given a thought to polarity or scarcity, though--I kind of assumed that dividing the player's production by the median would handle that well enough. I will have to look into it.

 

I would be interested to hear how you think they should be handled and what they should change in the rankings, but I disagree with the idea of valuing high-production-in-category players more highly in order to "dominat[e] that cat." If the next-highest team has x, all you need is x + 1. Continuing to value high production in that cat even as you are approaching or exceeding that threshold will offer diminishing returns and devaluation of the pick/player resource. For example, suppose your target is 12 blocks per game (arbitrary number) totaled across your team and you are sitting at 11.5. If you have two good options with nearly identical overall production, except that one offers 1.5 bpg and the other provides only 0.6, the latter is your better option, the better value, because he is getting you over the threshold in blocks while producing more in other categories as well.

 

Not that you actually know while you are drafting what your opponents will average. I suppose the ideal would be a program that updates throughout the draft and adjusts rankings based on team selections and the players remaining in the pool. But that would basically just be cheating.

 

Speaking of getting over the threshold, I was going to say to the guys who were talking about gathering up roto data to figure out targets for finishing third in every category, instead of finding the averages for third-place finishers, you should find the maximal fourth place in each category. That's the guy you're trying to beat.

 

And finally, on the subject of z-scores, I exchanged a few emails with Ryan Knaus about them a couple of years ago, likely after one of his columns in which he proclaimed that so-and-so was six standard deviations above average in assists or blocks or whatever--for anyone who doesn't know, it should be virtually impossible in a population as small as the NBA for a player to be that far above or below the average, if the data were normally distributed. I hinted that his observation about that player should have raised a statistical red flag, and that maybe z-scores were being misapplied to fantasy sports. It ended up with him saying that he knew of no alternative, but he would contact a friend of a friend who was a stats professor to see if he had any ideas. I expressed enthusiasm but never heard anything more about it; I suspect that the prof felt that it wasn't worth his time. It is heartening to see how quickly you guys have begun looking at the problem and searching for solutions.

Link to post
Share on other sites
10 hours ago, Gorgeous George said:

Glenn Robinson Ill could be a sneaky pick in deeper leagues. 

Yeah I am eyeing that right now. However, Bojan Bogdanovic really turned it up in Washington as a 3 pt specialist and is a real threat to eating Glen's mins.

Link to post
Share on other sites

9 cat re-draft H2H...

 

I am seeing a lot or Harden/Westy not even cracking the top 5...in some cases ranked like 8 or 9...

 

For the case of Harden....Is this mainly because of the addition of CP3?

 

For Westy, if you punt 1 or 2 cat in h2h....how can you still draft him so low?

Link to post
Share on other sites
1 hour ago, crocp said:

9 cat re-draft H2H...

 

I am seeing a lot or Harden/Westy not even cracking the top 5...in some cases ranked like 8 or 9...

 

For the case of Harden....Is this mainly because of the addition of CP3?

 

For Westy, if you punt 1 or 2 cat in h2h....how can you still draft him so low?

 

In 9-cat roto, it's easier to see why those two are generally dropping.  Although either of them are liable to improve upon their flaws (TO, FG%) given their new situations.

 

In H2H I think it's more about preference.  Any of the top 8 or so guys could go in almost any order depending on how you want to build your team.

Link to post
Share on other sites
3 minutes ago, Gorgeous George said:

I guess Derrick Favors would count as a sleeper.

 

Either sleeper or DND. 

 

I drafted him end of the 6th round in the mock draft. Probably could have let him drop another round or 2 but I think he'll get himself healthy this year and bounce back. It's a contract year for him and he's still only 26.

Link to post
Share on other sites

I think he's a sleeper for sure and would have drafted him if it weren't for him going way earlier than expected.  His career highs are 16.4 PPG/8/7 RPG/1.7 BPG/1.2 SPG and he has been doing close to that consistently for 3 years prior to last year.  His poor performance last season was due solely to injury and he performed very well in the playoffs when given the opportunity.  With Gobert in tow he may be expected to play further away from the rim, which could result in lower rebounds and blocks.  Alternatively, the Jazz have few scorers, and he might be there best midrange scorer.  Therefore, his points could increase.  A line of 17 PPG/7 RPG/1.5 BPG/1 SPG could be quite achievable.  

Link to post
Share on other sites

Utah's pretty interesting this year I think. I'm willing to bet that they are a non-playoff team. They ranked last in possessions per game last season, but now have Rubio to potentially increase the pace. As for Favors, if he's not 100%, his knee could really snowball into a problem again over the season if the pace picks up.

I would think Favors' minutes would decrease as the season progresses to develop their younger players, but looking at the depth charts there really is nobody right now to take the 4-spot. 

Stifletowers' 17-7 line compares closest to Aldridge's stats from last year that put him at #68. I'm not targeting him, but I'd have no problems taking him after around 75, but I would really need to be confident that the rest of my players aren't going to miss time (I play H2H). Definitely can't have him as a part of my starting 5 though.

Link to post
Share on other sites
Quote

Timofey Mozgov scored 20 points with five rebounds in a friendly against Hungary on Friday.

He didn't take any 3-pointers for the Russian national team. Mozgov is hopeful he can add that to his game, but it won't be easy. While he was just 7-of-40 from deep in his career, he was 13-of-34 (38.2 percent) on his shots from 20-24 feet over the last three season. He's a terrible fit for the NBA's reigning fastest team, so he's not a great fantasy target.

 

Even though he won't be a top target  (obviously), I don't think it's out of the question for him to average late DEN/ early CLE numbers: 9-10 pts on 50+% fg, 70-80%ft, 7 reb, a block, little over a TO. Once again not, great, but you could do worse in the last round of your draft. 

 

Link to post
Share on other sites
On ‎13‎/‎08‎/‎2017 at 3:31 AM, PuzzBeterson said:

 

Even though he won't be a top target  (obviously), I don't think it's out of the question for him to average late DEN/ early CLE numbers: 9-10 pts on 50+% fg, 70-80%ft, 7 reb, a block, little over a TO. Once again not, great, but you could do worse in the last round of your draft. 

 

As long as you know you will have to replace him with someone else mid season, you should be able to manage drafting this guy.

Link to post
Share on other sites
Quote

Timofey Mozgov: Kenny Atkinson said I’ll get a lot of minutes

“I met with the coach and discussed some aspects of the upcoming season,” said the 7’1” center acquired from the Lakers. “From the conversation, it became clear that I would spend a lot of time on the court, and the team will play fast basketball.“

 

https://www.netsdaily.com/2017/7/10/15949486/timofey-mozgov-kenny-atkinson-said-ill-get-a-lot-of-minutes

 

I kinda feel like the first and second part are somewhat mutually exclusive, but hey. 

 

 

Link to post
Share on other sites

There are definitely other upside young bigs I'd be targeting late (Bobby Portis, Thon Maker, Skal Labiessere, maybe Sabonis) but Mozgov has already proved himself. 10/7/1blk should not be left on the waiverwire if he can do it. 

Edited by s-kayos
Link to post
Share on other sites
13 hours ago, s-kayos said:

There are definitely other upside young bigs I'd be targeting late (Bobby Portis, Thon Maker, Skal Labiessere, maybe Sabonis) but Mozgov has already proved himself. 10/7/1blk should not be left on the waiverwire if he can do it. 

I'm really liking Thon Maker right now, basically just because how much Jason Kidd hates using Monroe and Henson. Kidd also seems to be making no progress trading the 2.

 

I'm a little uneasy about Bobby Portis though. He really doesn't seem to be disrupting shots at the rim so he has to earn his playing time with outside shooting and rebs. I feel like the rookie Lauri will get decent minutes cause he can be a potential stretch 5 and limit chances Portis can be a C in a small lineup.

Link to post
Share on other sites

Mozgov, IMHO, is not worth a draft spot in a 12 team league. Streamer/14 team maaaybe. No upside there. I'd rather churn his spot than hold him. They took him on as a salary dump only. It's full on youth movement time so Allen should get plenty of minutes as well. They play at a high pace and love 3's. They will probably play a lot of small ball with Booker/Acy as well. Wouldn't be shocked if other players play some 5 as well (RHJ?). Don't waste your time.

 

Love Thon this year in my keeper league. Lottery ticket. Probably going to be low end production at first but very high ceiling. Definitely some Anthony Randolph type potential as a disappointment so I wont break the bank. 

Link to post
Share on other sites

haha! nice one lusamsam. anyways, the issue for me with mozgov is the minutes. its been 3 seasons ago, where he was big part of the rotation, or even starting, getting 25min where he avg along the lines of 10pts 7reb 1.2 blk. since then, the league has played a lot faster thus the small lineups where for me, part of the reason he has seen his minutes decrease. and the fact that hes now 31 yrs old. no issue here, as you can get him in the last round. but personally, i would rather get a young high upside or in prime player coming of a down season in a discount as a flier with my last picks. again, if u need a big and playing it safe, i like mozgov as well. but IMO, with the nets playing style, those numbers from 3 seasons ago might be his upside if he gets 25min.     

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    No registered users viewing this page.


×
×
  • Create New...