Rating the NFBC drafts

Dyv · Post by **Dyv** » Fri Apr 22, 2005 3:20 pm

Originally posted by ToddZ:

After driving home (amazing, I seem to do my best thinking at traffic lights) I think I have a way to present my case. Some of what Hughes just posted is valid about the manner I am approaching this.

Let me introduce four sets of player rankings:

1. A list of 345 player assigned value randomly so that the sum is equal to $3900 (15 x $260)

2. An end-of-the-season ranking based on actual stats

3. The median values discussed here

4. An list based upon the average values of your favorite 10 fantasy baseball prognosticators.

It is obvious that the winners will have the most correlation to the second list (actual year end values) and the worst correlation to the first list (random).

My point is I feel the 4th list is a better indicator than the third list. The median list will in all probability be based upon the prognosticator's list, but normal auction dynamics will introduce some fluctuations, most notably the stars and scrubs effect.

But will puffins list have some sort of positive correlation to winners? Of course it will because even with the flaws, ir is still a much better ordering of player's value than a randomly generated list (using an extreme example). Perhaps in my zeal to point out what I perceive to be flaws, I overlooked the positive correlation it will have.

Truth be told, I think this is a really neat exercise if thought of in the vein "let's see which is better, median values of NFBC auction drafts or an average of these 10 published value sets." Let's not presuppose anything and see what happens (which I suspect is the primary intent anyways). Todd, I understood your critique to be just that - a critique. I did not perceive you to be arguing Puffin's data or presentation was worthless, only flawed. The DEGREE of the flaw is what is truly at question here. I don't think anybody feels it's accurate projection material per se

It truly begs the question... is there any way to have an objective review of everyone's draft skill? Is there a better one that what Puffin has done? What would it be?

As to your method #4 above (averaging your favorite 10 prognosticators) - I'm not at all convinced there are 10 projections that are worth averaging together... maybe 4 or 5 total

Todd, what is the correlation Puffin has put together in your opinion? How many of the 60 league cash winners will come from his top 150 drafters?

Just curious - for entertainment purposes only

Dyv

Dyv · Post by **Dyv** » Fri Apr 22, 2005 3:28 pm

Originally posted by UFS:

What value do I get for my G.Zaun pick, laughed at by DYV

, but was the third from last pick of the entire draft?

Zaun just hit another RBI double. My analysis is awesome

lol, Greg Zaun?

John, he's almost as old as Ambrosius...

ToddZ · Post by **ToddZ** » Fri Apr 22, 2005 3:32 pm

It truly begs the question... is there any way to have an objective review of everyone's draft skill? Is there a better one that what Puffin has done? What would it be? A retro draft based on previous year's stats combines one's ability to assign value with one's ability to draft.

As to your method #4 above (averaging your favorite 10 prognosticators) - I'm not at all convinced there are 10 projections that are worth averaging together... maybe 4 or 5 total There's more than five. Truth be told, about 90% of a projection set is based upon the same foundation, it's that last 10% based on unique data analysis and gut hunches that differentiate the results.

Todd, what is the correlation Puffin has put together in your opinion? How many of the 60 league cash winners will come from his top 150 drafters?

Honestly, I can't put a number on the correlation. The part I stopped talking about was my belief that astute drafters are hurt by this analysis as they are being compared in general terms to the mainstream fantasy populace and you can't get credit for drafting players against the grain or putting together a strategy that lacks raw value but maximizes roto points (the key to this whole thing).

mulberry · Post by **mulberry** » Fri Apr 22, 2005 6:08 pm

Dyv, I noted the hq guy because I've read about him at the site. Apparently he is the appointed straight draft guru, thats all. I'm sure there are many others worth noting, he just caught my eye......Another point that hasn't been discussed when trying to correlate these values to projected money winners is this years' beltres, sheets, santanas, etc....so much of the value is affected over the season by over/under achievers and "the true key factor" in winning...INJURIES to highly drafted players. Someone can run numbers all day long, but when a Pujols or Beltran goes down, those draft ratings can be thrown out the window....Still enjoyed reading it though...it definately drew some passionate responses and convo

Jackstraw · Post by **Jackstraw** » Fri Apr 22, 2005 6:57 pm

Originally posted by Dyv:

Does anyone have the number crunching acumen to compare Puffins top 150 to the current top 150 and see how close it is RIGHT NOW ? Obviously this has little real value, but just curious... is there any trending already noticeable?

Dyv [/QB]As weird as THIS may sound :rolleyes: ,

Correlations: Value, Batting Pts, Pitching Pts, Overall Pts

Correlation: Value

Batting: 0.101

p-value: 0.082

Pitching: 0.121

p-value: 0.036

Overall: 0.153

p-value: 0.008

Comparing Puffins' rankings to the current standings (I've included Batting and Pitching scored), there is very little correlation. If interested, the R-Sq value was only 2.4%.

BTW, hold on to your money. There are only 38 in his top 60 overall (I'm not sure about the top 3 from each league). Seeing the correlation so far, the bet would be pretty much a coin flip.

Jackstraw · Post by **Jackstraw** » Fri Apr 22, 2005 7:14 pm

I did the previous analysis on the whole lot of 300. When I reread the question and saw his top 150 compared to the overall 150, I reworked it. Here are the results:

Correlations: Value, Batting Pts, Pitching Pts, Overall Pts

Correlation: Value

Batting: -0.081

p-value: 0.324

Pitching: 0.084

p-value: 0.309

Overall: -0.007

p-value: 0.930

Far worse when just looking at the whole set.

mulberry · Post by **mulberry** » Fri Apr 22, 2005 7:31 pm

Since I'm not a real math genius, can you explain those numbers please ? I should have paid more attention in statistics class.

Jackstraw · Post by **Jackstraw** » Fri Apr 22, 2005 7:51 pm

Originally posted by mulberry:

Since I'm not a real math genius, can you explain those numbers please ? I should have paid more attention in statistics class. No problem... The top numbers are correlation coefficients. The closer the number is to 1 or -1 then the better the correlation is. When the coefficient is 0, there is no correlation at all. So you see, all of his values for coefficients are 0.153 or less, which shows very little correlation. As a matter of fact, from that analysis alone I would say that only the 0.153 shows some significance. But when you take that number and get the R-Sq value for the analysis there is only a 2.4% contribution to variance. That means that if you made an equation that related Puffins' data and the overall points scored, the actual data points would be all over the place when compared with the expected values. So, in short, when it comes to correlation coefficients you want the absolute value to be near 1 for good correlation and you want the R-Sq value to be near 100% to show that the variation is directly related to what you are measuring. That leaves us with the nasty little p-value... The p-value represents the probability that the analysis fits within some accepted confidence level. In my case, I'm shooting for 95% confidence that what I am saying is correct. For this analysis, I want the the p-values to be less than 0.05 to show that I am within that confidence level. From the numbers, I will say with 95% confidence that there is a correlation value of 0.153 between Puffins' rankings and the overall points YTD. In other words, I am positive that there is very little relationship between his rankings and how anyone is doing and, to go out on a limb, will do through out the season.

nydownunder · Post by **nydownunder** » Sat Apr 23, 2005 1:22 am

Originally posted by mulberry:

Funny I just looked at the same thing. How can the top 2 overall finishers last year rank 248th and 208th coming out of the draft and all the league winners be that far down as well. Plus the straight draft guru for hq is 245th? One would think these guys come prepared to the draft and knew what they were doing enough to make this data flawed. Interesting read though, Puffins. I suspect these guys are much more capable of finding value in the middle rounds. These players wouldn't project highly, thus the lower scoring. But also keep in mind that these players are more risky and I suspect last year's winners are more likely to take chances as they probably assume they were correct last year. I would guess half of them were smart and the other half lucky.

nydownunder · Post by **nydownunder** » Sat Apr 23, 2005 1:26 am

Originally posted by Dyv:

quote:Originally posted by mulberry:

Funny I just looked at the same thing. How can the top 2 overall finishers last year rank 248th and 208th coming out of the draft and all the league winners be that far down as well. Plus the straight draft guru for hq is 245th? One would think these guys come prepared to the draft and knew what they were doing enough to make this data flawed. Interesting read though, Puffins. What makes you think the straight draft guru from HQ is any good? I have no idea who this guy is or what his credentials are, just curious why he's given special note?

Does anyone have the number crunching acumen to compare Puffins top 150 to the current top 150 and see how close it is RIGHT NOW ? Obviously this has little real value, but just curious... is there any trending already noticeable?

Dyv [/QUOTE]I've got the average player projections using 4-6 publications, but were talking about a lot of work. I can tell you one thing though, the leader, who is raqted highly in Puffins analysis will not finish in the top 50-75. His team projects far too low in many offensive categories, especially AVG. Not even a few player breakouts would siginificantly improve those stats.

nydownunder · Post by **nydownunder** » Sat Apr 23, 2005 1:32 am

Originally posted by Jackstraw:

quote:Originally posted by mulberry:

Since I'm not a real math genius, can you explain those numbers please ? I should have paid more attention in statistics class. No problem... The top numbers are correlation coefficients. The closer the number is to 1 or -1 then the better the correlation is. When the coefficient is 0, there is no correlation at all. So you see, all of his values for coefficients are 0.153 or less, which shows very little correlation. As a matter of fact, from that analysis alone I would say that only the 0.153 shows some significance. But when you take that number and get the R-Sq value for the analysis there is only a 2.4% contribution to variance. That means that if you made an equation that related Puffins' data and the overall points scored, the actual data points would be all over the place when compared with the expected values. So, in short, when it comes to correlation coefficients you want the absolute value to be near 1 for good correlation and you want the R-Sq value to be near 100% to show that the variation is directly related to what you are measuring. That leaves us with the nasty little p-value... The p-value represents the probability that the analysis fits within some accepted confidence level. In my case, I'm shooting for 95% confidence that what I am saying is correct. For this analysis, I want the the p-values to be less than 0.05 to show that I am within that confidence level. From the numbers, I will say with 95% confidence that there is a correlation value of 0.153 between Puffins' rankings and the overall points YTD. In other words, I am positive that there is very little relationship between his rankings and how anyone is doing and, to go out on a limb, will do through out the season. [/QUOTE]I think the best coorelation analysis, which would be a lot of work, is use the average projections of several publications, stack the stats, and comapre points to last year's standings (adjusted for 300 maangers this year).....and then adjust for player injuries (ie out of a manager's control and assumes the player would obtain their projection). You would adjust by using a % of stats and include a replacement (this number could be standard condiering what little is out there). I would bet the correlation could get up in the .80 range.

ToddZ · Post by **ToddZ** » Sat Apr 23, 2005 3:38 am

I suspect the correlation will improve a bit as the year moves on, but no where near that of an aggregate list of projections (which itself won't be as good as some may think).

I can compile the industry average list and assign values customized to the NFBC league values if we want to revisit this at season's end.

At the end of each month, I will be calculating year to date values as well, so month to month correlation on value earned thus far can be run.

Give me a holler and I can supply those numbers.

bjoak · Post by **bjoak** » Sat Apr 23, 2005 5:49 am

You know, it's sad that you guys know math so well and yet you don't understand that a sample size of three weeks is worthless. As proof, after three weeks the Yankees are tied with the Devil Rays for last place. I wish that would keep up as much as the next guy, but I have common sense.

Also, rankings, aggregate or no, of national publications are worthless just the same. These are the same guys who have Sean casey ranked in the top 75. As I've pointed out before the season even started, that's just silly--especially when Mark Kotsay was ranked some 10 or 15 rounds later, though he had similar numbers last year and before...and if you want to look at the teeny-tiny sample size for those two players go right ahead.

Now, it's not like one has Kotsay in the top 75 and another has Casey in the top 75. No, they're all equally subjective and worthless, and if you want to spend your time making an aggregate of worthlessness go right ahead.

I admit that there are places where you can find objective math analysis to determine projections, but based on the fact that you're going with an aggregate of espn, yahoo, yawn, you don't know what they are, and I'm certainly not going to tell you.

I can't believe I'm losing to you guys.

[ April 23, 2005, 12:00 PM: Message edited by: bjoak ]

ToddZ · Post by **ToddZ** » Sat Apr 23, 2005 6:00 am

Also, rankings, aggregate or no, of national publications are worthless just the same. These are the same guys who have Sean casey ranked in the top 75. As I've pointed out before the season even started, that's just silly--especially when Mark Kotsay was ranked some 10 or 15 rounds later, though he had similar numbers last year and before...and if you want to look at the teeny-tiny sample size for those two players go right ahead. Kotsay and Casey have posted similar numbers the past 2 seasons?

Really?

2003

Kotsay .266 7HR 38RBI 64R 6SB

Casey .314 15HR 80RBI 71R 4 SB

2004

Kotsay .314 15HR 63RBI 78R 8 SB

Casey .324 24HR 99RBI 101R 2SB

It's not even close, and I like Kotsay.

ToddZ · Post by **ToddZ** » Sat Apr 23, 2005 6:03 am

I admit that there are places where you can find objective math analysis to determine projections, but based on the fact that you're going with an aggregate of espn, yahoo, yawn, you don't know what they are, and I'm certainly not going to tell you. Sorry, I missed where anyone suggested using an aggregate of ESPN, yahoo, etc.

bjoak · Post by **bjoak** » Sat Apr 23, 2005 6:26 am

Uh, huh, 2004 was almost even if you count in the stolen bases. 2003 was bad for Kotsay as he had back problems all year. Also, your '03 stats for Casey are blatantly...um...wrong as he hit .291 with 14 HRs. Not terribly impressive. Then, if we have the good sense to look at 3 year samples, we find 2002:

Casey: .261 6HRs 2 SBs

Kotsay: .292 17 HRs 11 SBs

Not even close, you say?

Casey has had his health problems too, but that in itself is a concern as Kotsay is fully healthy.

Sorry, I missed where anyone suggested using an aggregate of ESPN, yahoo, etc.

I give you that, but there aren't enough places that focus on objective math analysis to create a strong aggregate so I'm assuming. If you're using fantasy sports mags and half-assed websites, that's not any better. At all.

nydownunder · Post by **nydownunder** » Sat Apr 23, 2005 7:36 am

Originally posted by bjoak:

You know, it's sad that you guys know math so well and yet you don't understand that a sample size of three weeks is worthless.

Also, rankings, aggregate or no, of national publications are worthless just the same.

I can't believe I'm losing to you guys. Ok, so you have proved you not only don't know anything about Math, but you also failed miserably in reading comprehension. No one is using 3 weeks worth of data to grant somone the champion already...and No one is saying this was an exact science...but what you can take from this is if it were a perfect world (ie players either performed to their projections, or auction value) this is how everyone would perform. No one here agrees that it is a perfect world - no one! And the rest of this topic was about how this analysis could be improved to provide more predictive power (ie correlation). If it can be proved with say 90-95% confidence, then it would be an extreemly valuable tool in winning this thing.

An aggreagte of publications are a better guide than any one's person's analysis - KEY WORD: Guide! I said Guide! Got it? Good!

I'm not surprised at all!

ToddZ · Post by **ToddZ** » Sat Apr 23, 2005 7:51 am

Sorry about 2003, I inadvertently put in 2004 BA and HR.

Uh, huh, 2004 was almost even if you count in the stolen bases Almost even? SB are valuable but 6 of them by no means even come close to equaling 9HR, 36 RBI and 23 Euns, not to mention 10 points of BA.

In a mixed league format, last season Casey earned $21 and Kotsay earned $13. Personally, I expect that gap to be narrowed this season, but with Casey still holding an advantage. That said, $8 is a ton of value--it wasn't even close.

Then, if we have the good sense to look at 3 year samples, we find 2002:

Casey: .261 6HRs 2 SBs

Kotsay: .292 17 HRs 11 SBs

Not even close, you say?

I said 2003 and 2004 were not even close. Nice try straw man.

Casey is over his main injury issue, his shoulder. But history does suggest he is more of a health risk than Kotsay.

The only way you even have a remote case is looking at 3-year averages from '02 to '04, which is in fact close to even.

Casey .294-14-73-2-76

Kotsay .292-13-54-8-74

So then the question becomes how much weight should be put on 2002? A lot of people like to use weighted averages so 2002 would contribute the least giving Casey the edge.

I like using 3-year averages as a base, but you can't just use them blindly. It is clear 2002 was the aberration, not the norm. In 2005, Casey should be projected to produce better stats than Kotsay.

I give you that, but there aren't enough places that focus on objective math analysis to create a strong aggregate so I'm assuming. If you're using fantasy sports mags and half-assed websites, that's not any better. At all. You assume wrong. There are plenty of sources for projections done by people who understand the science behind the process.

Spyhunter · Post by **Spyhunter** » Sun Apr 24, 2005 5:46 am

Hi, just wanted to say thanks for the analysis - It reinforced my own belief that we got great value.

Sadly, this value has yet to manifest itself with actual performance, but perhaps, with warming weather will come better output....

I have been away from the message boards due to work but will try and catch up today.

Thanks again for the good work

AND BTW: I TOTALLY agree with your assessment that the higher the draft order pick (i.e. highest pick being the #1 pick), the better the team. I have noted this for years and still get into stupid arguments with people about this....

Spy aka Chris Throop

[ April 24, 2005, 12:21 PM: Message edited by: Spyhunter ]

bjoak · Post by **bjoak** » Sun Apr 24, 2005 7:11 am

In my case, I'm shooting for 95% confidence that what I am saying is correct. For this analysis, I want the the p-values to be less than 0.05 to show that I am within that confidence level. From the numbers, I will say with 95% confidence that there is a correlation value of 0.153 between Puffins' rankings and the overall points YTD. In other words, I am positive that there is very little relationship between his rankings and how anyone is doing and, to go out on a limb, will do through out the season. Ok, so you have proved you not only don't know anything about Math, but you also failed miserably in reading comprehension. No one is using 3 weeks worth of data to grant somone the champion already...and No one is saying this was an exact science My reading comprehension is fine. Here's a second chance to check yours.

bjoak · Post by **bjoak** » Sun Apr 24, 2005 7:18 am

Todd, you seem to agree with me. I mean I do think Kotsay is a better player but not a waaay better player and I don't expect everyone to agree with that. It's a matter of opinion.

My argument was merely that it's ridiculous that Kotsay goes in maybe round 16 while Casey goes in maybe 4. Even with the normal projections Casey gets, he's a silly sub-8 pick. If they were both going in round 11 or even 2 or 3 rounds apart, I'd have not made the comment. They did, however, go 10 or 12 rounds apart and that just doesn't figure.

[ April 24, 2005, 01:22 PM: Message edited by: bjoak ]

ToddZ · Post by **ToddZ** » Sun Apr 24, 2005 7:42 am

For the record, I think Casey is decidedly better than Kotsay, but I like Kotsay more than the norm. It wouldn't surprise me if Kotsay follows a similar career path as Steve Finley, with maybe not quite as much power, especially if he remains in Oakland.

I see Casey as someone who will produce at a 3rd round level, is draftable in the 4th and you get a profit in the 5th. Someone in my league read this analysis before the draft and snagged him in the 3rd--so be it.

Kotsay falls in drafts mainly because people overcompensate for scarcity early on, letting a ton of outfielders fall. By rights he is likely a 12th round pick or so but often, there are SOOOOO many OFers still left on the board that he is bypassed. I did a pair of NFBC style drafts before the real thing--using NFBC style rules with only a $100 entry fee and Kotsay went 12th in one, 18th in the other. It has to do with draft dynamics more than value.

I did an NFBC NL only draft and an NFBC AL only draft. My bid value for Casey was about $28 and was about $15 for Kotsay.

Chest Rockwell · Post by **Chest Rockwell** » Sun Apr 24, 2005 3:38 pm

I read this entire thread: 3 things jumped out at me.

1) NY Down Under is one of the brightest guys on this board. That has nothing to do with him giving me a pretty favorable draft grade.

2) This league is the classic moneyball argument, number crunchers, vs scouting types. In my opinion the winners over the years will come from people who consistently can balance the 2.

3) Debating with this Zola cat is a waste of time, he does not see it as a debate but as a chance to educate you, after you tell me you value Casey as top 60 and that Patterson is worth an outlandish bid I stopped listening...

Cherokee Nation · Post by **Cherokee Nation** » Sun Apr 24, 2005 8:09 pm

That data is weak and off base

The baseline data is invalid. It's not based on anything worthy of a unit of measurement, thus invalidating the entire presentation, relegating it to what it in fact is, a bunch of meaningless numbers

I'm in Chicago 2 and I drafted in Vegas last year ( in a league sprinkled with top teams ). This years Chicago 2 draft was much tougher than last years Vegas....without a question

Am I right Kings ? We drafted in both

Pedro Martinez is one of the worst picks at the end of the 1st round ? I knew he'd be back this year, and pitch well in pitcher friendly home park....and he'd have a 4-0 record if the Mets had a bullpen, but it was a bad pick because your data says so ?

Mark Loretta was a bad 6th rounder ? Because I knew he was going to add 30 steals to his already solid game and your data didn't, makes it a bad pick ? Come on man

And you give big marks to people who took Barry Bonds in round 6 ? What a joke, he can't take roids anymore, he's 55 years old, he's hurt and walks don't mean **** . He was a lousy pick in the top 10. What's he done ? When he's back ? And it was already there ( roids, health ) to be seen before the draft.

Love to play against the data guru who put these bad, invalid statistics together, love to have one patsy in the league

JohnZ · Post by **JohnZ** » Sun Apr 24, 2005 8:48 pm

Worst pick #33

33 William Hoffman Brian Roberts 6 6 NY Lg 7 18.2 10.0 -8.2

Just something I saw that cracked me up.