Correcting Auction Calculation Article (1 Viewer)

[gheemony]

http://subscribers.footballguys.com/2006/0...e_johnson01.php

Nice work Forrest (and run, Forrest, run). Anyone relying on the auction values produced by DD or the VBD spreadsheet should read this article. I've known this problem for years and have my own spreadsheet that converts VBD to $$$. It's different than Forrest's system, and I have issues with Forrest's system, but at least he points out and fixes a major problem.

Now if I could just find more auction leagues to take advantage of my super-spreadsheet system. I am finding that draft leagues are getting pretty boring. I hope auction leagues are the wave of the future - I hope the eventually dominate like they do in Fantasy Baseball.

Again, great work Forrest.

*****************************************************

Gray edit: Just adding the spread sheet here. I had to zip it up, so hopefully that's not a problem for anyone.

New_Two_Line_VBD_vJ1_10team.zip

 

[Zasada]

The horrible auction values in the DD are the primary reason I have kept my own Excel workbook to manage my auction. I find it odd that they have stuck with such an obviously flawed algorithm for so long, given the level of effort in other areas.

But I applaud their willingness to let an "outsider" audit their methods and suggest an alternative. Kudos.

 

[CalBear]

One thing to check; by default, DD will modify auction values by its Strength of Schedule calculation. If you're not getting the results you expect, turn off SOS and see if that's better.

 

[doughboydeluxe]

Any of y'all who use a different way of computing auction values willing to share your methodology?

Maurile posted an article describing his way of computing auction values, and the spreadhseet Forrest uses will hopefully be posted soon.

 

[Volleydog]

Thanks for posting this as I hadn't gotten over to the articles page yet. Of course, I wish I'd read it BEFORE yesterday, when the 2nd of my 2 auctions completed.

I've noticed some problems with the VBD app over the past couple years, but had just manually adjusted totals to reflect a more (I thought) correct value structure. Hadn't taken the time to sit down & work through the problem, but it's nice to see someone else did.

Also, more kudos to the FBG Staff. It's a freelance article that exposes a flaw in one of their treasured "extras", yet they were more than willing to post it so their members could benefit. Truly a group that is involved for the greater good. Thanks, guys.

 

[ManODaHour]

The horrible auction values in the DD are the primary reason I have kept my own Excel workbook to manage my auction. I find it odd that they have stuck with such an obviously flawed algorithm for so long, given the level of effort in other areas.

Can you expand on this? How is your spreadsheet setup? I've been trying to find the right type of fit for my values on the DD to no avail. I've finally given up and posted my values in the Comments/Notes section. I wish there was some way of keeping tiers on the DD as well. I am pretty clueless to setting up Excel to calculate things for me, so I just make my own "cheatsheet" and keep manually updating - which really sucks! But that's what has worked best for me up to this point.

 

[Wooderson]

Great Article.

I too have always thought the auction value was a little strange in DD. I am going to have to read the article a few more times but I hope I can appy this to my draft.

Hopefully the spreadsheet that will be posted will help.

 

[Dickie Dunn]

I would love to take a look at any of your spreadsheets to see how they work.

As I've posted before, my auctions tend to have plenty of RB/WR value below the elite level players, so I would be interested to see how different calculations would affect this. Still got one more auction to go.

Thanks in advance.

 

[Wooderson]

Any of y'all who use a different way of computing auction values willing to share your methodology?Maurile posted an article describing his way of computing auction values, and the spreadhseet Forrest uses will hopefully be posted soon.

Do you happen to have a link to the article by Maurile?Thanks

 

[Big Sug]

this is the best article to come down the pike in a very long time.

it does a great job of defining and then properly articulating the meaning of value in an auction.

add keepers to the mix and then you have a really wacky situation. and also the most challenging and fun day of the year.

 

[Art_VanDalay]

I agree with everyone's comments. I looked for the spreadsheet he mentions at the end of the article but I don't see it yet. I assume it will be under tools. Perhaps someone can advise this thread when it is updated and I will keep an eye out for it. Have an auction draft this weekend so hoping it will be done soon.

 

[ManODaHour]

One thing to check; by default, DD will modify auction values by its Strength of Schedule calculation. If you're not getting the results you expect, turn off SOS and see if that's better.

Thanks for this heads-up. I will give it a try when I get home today! Anyone else have any other tips they use?

 

[secretid]

 

[Wooderson]

One thing to check; by default, DD will modify auction values by its Strength of Schedule calculation. If you're not getting the results you expect, turn off SOS and see if that's better.

Thanks for this heads-up. I will give it a try when I get home today! Anyone else have any other tips they use?

Where do you turn this off?

 

[CalBear]

One thing to check; by default, DD will modify auction values by its Strength of Schedule calculation. If you're not getting the results you expect, turn off SOS and see if that's better.

Thanks for this heads-up. I will give it a try when I get home today! Anyone else have any other tips they use?

Where do you turn this off?

Go to the Game by Game window, and turn off "Use strength of schedule."(It is not very intuitive that this window affects VBD/auction values, but I don't think there's anywhere in Setup to change it).

 

[Wooderson]

One thing to check; by default, DD will modify auction values by its Strength of Schedule calculation. If you're not getting the results you expect, turn off SOS and see if that's better.

Thanks for this heads-up. I will give it a try when I get home today! Anyone else have any other tips they use?

Where do you turn this off?

Go to the Game by Game window, and turn off "Use strength of schedule."(It is not very intuitive that this window affects VBD/auction values, but I don't think there's anywhere in Setup to change it).

Thanks makes a little difference I will have to see which one I agree with more.

 

[ManODaHour]

Hopefully the spreadsheet that will be posted will help.

ATTENTION MODS: we need to get our hands on Forrest's spreadsheet (by this Sunday - HINT, HINT). Any way you can pimp this?!

 

[Z-Dog]

Any of y'all who use a different way of computing auction values willing to share your methodology?Maurile posted an article describing his way of computing auction values, and the spreadhseet Forrest uses will hopefully be posted soon.

I actually use a multi-line fit, but to use my method you need decent historical data on your auctions.First step is to figure out the percentage of total cap dollars spent by position. This can vary heavily based upon your scoring and rules, but a good palce to start is about 45% of cap on RBs, 30% on WRs, 20% QBs, and 5% TEs (I exclude Ks and Ds, and subtract the estimated number of dollars that they cost). I then look within the positions for the premium player division line. 85/15 is an estimation, buit it will be different by position.Once I've got all that data, I can choose my lines. I usually go with last starter at the position for the first line, and best player who can be bought for $1 as my second line. I compute the VBD points for the non-premium group at each position and generate a cost per point of VBD of non-premium (what I call backup VBD, vs. the otehr group,whcih is starter VBD). Once I've got that, I generate prices for this lower group.I do the same for the premium groups at each position as wel, but with the following key change: rather than assign only $1 to the price generated by multiplying VBD points x cost/point (which is what you do for non-premium, so taht even the worst player in the non-premium costs $1 minimum), I add the dollar amount of the highest-priced non-premium player at that position, and subtract $1. Basically, what I'm doing here is this: I'm saying that any premium player must cost at least as much as the non-premium player, PLUS the cost for the premium player's extra production. The extra production ahs a higher cost per point, and I simply add that to the baseline of the best non-premium player at his position.I have found this method to be awesome for actually using at auctions. It accurately predicts the real shape of my league auction, does not overvalue QBs, and with some cleverness, you can make it give you dynamic auction pricing (though this IS tricky). My only caveat is that this is not a predictive methodology. It doesn't tell you what any player is really worth. It just tells you what he's worth relative to the amount of money being spent at his position. If you set the RBs to 60% of cap, or if you set premium players to 70% of cap instead of 85%, you will get very different results. That's why this method works best fi you have historical data, or play in a relatively standard league, so you can use the rules of thumb above.

 

[ManODaHour]

Any of y'all who use a different way of computing auction values willing to share your methodology?Maurile posted an article describing his way of computing auction values, and the spreadhseet Forrest uses will hopefully be posted soon.

I actually use a multi-line fit, but to use my method you need decent historical data on your auctions.First step is to figure out the percentage of total cap dollars spent by position. This can vary heavily based upon your scoring and rules, but a good palce to start is about 45% of cap on RBs, 30% on WRs, 20% QBs, and 5% TEs (I exclude Ks and Ds, and subtract the estimated number of dollars that they cost). I then look within the positions for the premium player division line. 85/15 is an estimation, buit it will be different by position.Once I've got all that data, I can choose my lines. I usually go with last starter at the position for the first line, and best player who can be bought for $1 as my second line. I compute the VBD points for the non-premium group at each position and generate a cost per point of VBD of non-premium (what I call backup VBD, vs. the otehr group,whcih is starter VBD). Once I've got that, I generate prices for this lower group.I do the same for the premium groups at each position as wel, but with the following key change: rather than assign only $1 to the price generated by multiplying VBD points x cost/point (which is what you do for non-premium, so taht even the worst player in the non-premium costs $1 minimum), I add the dollar amount of the highest-priced non-premium player at that position, and subtract $1. Basically, what I'm doing here is this: I'm saying that any premium player must cost at least as much as the non-premium player, PLUS the cost for the premium player's extra production. The extra production ahs a higher cost per point, and I simply add that to the baseline of the best non-premium player at his position.I have found this method to be awesome for actually using at auctions. It accurately predicts the real shape of my league auction, does not overvalue QBs, and with some cleverness, you can make it give you dynamic auction pricing (though this IS tricky). My only caveat is that this is not a predictive methodology. It doesn't tell you what any player is really worth. It just tells you what he's worth relative to the amount of money being spent at his position. If you set the RBs to 60% of cap, or if you set premium players to 70% of cap instead of 85%, you will get very different results. That's why this method works best fi you have historical data, or play in a relatively standard league, so you can use the rules of thumb above.

This sounds very similar to what Forrest does. Is there any way you can post a link to your model?

 

[inca911]

Thanks to all for the kind words on the article. I hope you find it useful.

I just sent a file to Clayton to host on the site. Please note that this is not an automated spreadsheet, so anyone who wants to use it will need some instructions on how to make adjustments. Please post your questions in this thread, and I'll put together a primer on how to use the spreadsheet.

 

[ManODaHour]

Thanks to all for the kind words on the article. I hope you find it useful.I just sent a file to Clayton to host on the site. Please note that this is not an automated spreadsheet, so anyone who wants to use it will need some instructions on how to make adjustments. Please post your questions in this thread, and I'll put together a primer on how to use the spreadsheet.

Thanks, dude! I'll be watching for it. Hopefully it's up and ready by Sunday.

 

[kal-el]

I thought it was a great article, as well. Very well thought-out and presented.

I do want to raise a conceptual question, though. The methodology takes projected fantasy points and uses comparisons to a baseline to determine auction values. This would be perfect if we were looking backward at what actually happened. In reality, however, we don't have a high degree of confidence as to which players will actually be above the baseline. I think this is justification as to why the top players' $ values should be at least somewhat lower than their projected points-over-baseline would suggest. You can't put too much money on one player. You have to be able to draft a number of players below the baseline for more than $1, as well, to give you additional possibilites for breakout seasons.

Does my point make sense?

 

[Wooderson]

Thanks to all for the kind words on the article. I hope you find it useful.I just sent a file to Clayton to host on the site. Please note that this is not an automated spreadsheet, so anyone who wants to use it will need some instructions on how to make adjustments. Please post your questions in this thread, and I'll put together a primer on how to use the spreadsheet.

Thanks, dude! I'll be watching for it. Hopefully it's up and ready by Sunday.

Even better if it's ready by Today. Draft tomorrow so I need to do last minute prep work tonight.

 

[Zasada]

The horrible auction values in the DD are the primary reason I have kept my own Excel workbook to manage my auction. I find it odd that they have stuck with such an obviously flawed algorithm for so long, given the level of effort in other areas.

Can you expand on this? How is your spreadsheet setup? I've been trying to find the right type of fit for my values on the DD to no avail. I've finally given up and posted my values in the Comments/Notes section. I wish there was some way of keeping tiers on the DD as well. I am pretty clueless to setting up Excel to calculate things for me, so I just make my own "cheatsheet" and keep manually updating - which really sucks! But that's what has worked best for me up to this point.

My workbook is fairly complicated. Uses array formulae with nested LARGE functions to be able to auto-rank players. I will admit that neither my projection methodology nor my $ allocation is terribly sophisticated, but it works for me. My projections are simply last year's per-game totals, multiplied by the number of games forecast to be played this year, with a SOS modifier and a personal "gut" modifier. Definitely the weakest link in my preparation, but I just don't like doing any more detailed projections than that.My VBD calcs are basically straight off custom baselines based somewhat on what players I think will be available on the WW plus a holistic evalutation of year-to-year reliability in scoring.The part you probably care about is my auction algorithm. It is a straight-rate prorate method of allocating $. I take the total VBD available in the league, the total $ in the league, and allocate $ based on points. So, if all players are forecast to generate 4,000 VBD in aggregate, and we have 12 teams with a $100 salary cap, then I need to prorate $1200 across 4,000 points. In essence, I'm saying that one VBD point is worth $0.30. So if Shaun Alexander has a VBD of 207, then he's worth $62.10 to me. In an auction, you're "buying" VBD points for your team. At the end of the auction you want to have the best VBD total of anyone in the league, so you should be shopping for bargains. Except in special cases, I always stop my bidding around 80% of a player's worth. That way, at the end of the auction I have purchased $120 worth of VBD points for only $100.I also have a "flex" value calculation that will re-price players through the auction based on the total $ remaining and the total VBD remaining. While I still want to only pay based on the base amount (and not the flex), the flex amount gives me an idea of what players are likely to go for as the auction progresses. In theory, as owners over/under-pay for players, that creates a shortage/surplus of cash for future player purchases and market prices could shift. Something to be aware of when bidding...I'll send anyone my workbook that wants it. But it's big (500K zipped) and takes about 5 seconds to calc on a high-end PC. Not sure how long it will take on a lower-end PC. Just shoot me a PM. And it probably won't be terribly useful for anyone else because it's set up for my league's scoring. But if you just want to get an idea of how an excel geek sets up his FF workbook, this might do it. I got a pretty big thrill when I figured out that you could use nested LARGE() functions inside an array formula to have a workbook custom-sort players after making adjustments to their projections. This kind of multi-conditional SUMIF() ranking actually helped me with some of my work stuff! I know, I need to get a life.

 

[inca911]

I thought it was a great article, as well. Very well thought-out and presented.I do want to raise a conceptual question, though. The methodology takes projected fantasy points and uses comparisons to a baseline to determine auction values. This would be perfect if we were looking backward at what actually happened. In reality, however, we don't have a high degree of confidence as to which players will actually be above the baseline. I think this is justification as to why the top players' $ values should be at least somewhat lower than their projected points-over-baseline would suggest. You can't put too much money on one player. You have to be able to draft a number of players below the baseline for more than $1, as well, to give you additional possibilites for breakout seasons.Does my point make sense?

Your specific projections for each player should already account for the level of confidence you have in the player. If you've already bumped LJ's TDs down to the same level as SA's, then you don't want to do it again when you're running the math of the auction calculations. You only want to account for uncertainty once in your method. Similarly, possibilities for breakout seasons should also be factored into that player's projections.I just emailed a copy of the spreadsheet to everyone who made a request. If you would like a copy and did not receive one, please email me. I plan to post a brief primer on using the spreadsheet later today.

 

[Endangered Monk]

PM sent inca. This is great stuff.

 

[Clayton Gray]

Bumping to add the spreadsheet to the first post.

 

[ManODaHour]

Thanks to all for the kind words on the article. I hope you find it useful.I just sent a file to Clayton to host on the site. Please note that this is not an automated spreadsheet, so anyone who wants to use it will need some instructions on how to make adjustments. Please post your questions in this thread, and I'll put together a primer on how to use the spreadsheet.

Forrest...You RAWK!

 

[inca911]

Please note that the spreadsheet is available for download in the first post in this thread.

I need to add the large caveat that this spreadsheet is not a finished product, and manual adjustments are necessary to make the spreadsheet work. I've used this for a few years, so I'm accustomed to making tweaks fairly quickly. The spreadsheet demonstrates an application of the concepts, but there is much automation still required to make this a tool for the general public. I'm open to helping people with the spreadsheet, and will post a brief primer on the site hopefully tomorrow. Best of luck!

 

[Faust]

 

[Jeff Pasquino]

 

[videoguy505]

I've never tried an auction draft but have been reading up on them anyway for the eventual date it happens. Plus I just love FF theory. But, I'm a rookie so I may have a dumb question--why are premium players under a different baseline? Isn't it still possible to draft two similar teams for two different amounts, if someone drafted the top "non-premium" players at every position, theoretically couldn't they amass a lot of points cheaper? Like I said, I'm a rookie, so I'm not quite understanding why premium players are worth a higher cost-per-point price.

 

[inca911]

I hope to update tomorrow with the primer for the spreadsheet. The holiday weekend has been very busy. I hope this has been helpful to some of you this year. Getting a fully automated tool is the next challenge. Please email or post with any comments or suggestions for improvement. Thanks!

BTW, I did an experiment and changed Larry Johnson's TDs to 99 total TDs (811 fantasy points) in the T1 version of the VBD application. Larry's auction cost did not change after hitting "recalculate".

 

[inca911]

I'll be working on this more during this regular season. Rather than do a primer for the current working version of the spreadsheet, I'm going to spend time working on the default variables for a version of the spreadsheet for next year. Thanks to all for your comments, and I'm interested in other ideas or points of view.