NFBC Random Slot Generator BLOWS

[Greg Ambrosius]

I guess I have a right to complain. Or do I?
I'll kds 1,2,8,7,6,5,4,3,9,10,11,12,13,14,15 and haven't received 1 or 2 yet in double digit drafts.

Since you and I both “change” our KDS prior to the lock, I wonder if we get shuffled to the back of the pack for purposes of pulling names and folks who go straight butter are first in line

Changing your KDS preferences has NOTHING to do with where you land. It's automated to run once the KDS deadline locks and spits out the names in seconds. Anyone in the Cutline sees that as the draft order email goes out seconds after the league is full email goes out. Switching preferences has nothing to do with where you land randomly in the draft order.

[Greg Ambrosius]

For historical context, some folks may remember the way we selected the high-dollar leagues during our one season at Fanball. We had hired a young secretary by the name of Kay Adams to work at Fanball at the time and we asked her to help with us on camera for this process. We bought an actual Bingo machine and placed 15 balls in there. We assigned owners' names to the numbered balls and Kay would roll them out and announce the names. Tom and I would then announce their top preference at that time and we'd post the draft order on the screen.

I think we did that in 2010 for baseball and football Super Leagues and maybe Ultimate leagues. It was as transparent as could be. Little did we know it would launch Kay Adams's career and vault her to Good Morning Football on the NFL Network!!

Nah, she did that on her own. She's fantastic and was fun to work with even back in 2010. It was a fun, transparent process, but I'm sure those owners whose Bingo balls came out near the end still weren't happy even with that process. At least in that case they could blame someone much prettier than me or Tom.

[TParsons]

We can't be more transparent than this...


There's no crying in baseball. Just good managing.

You still haven't been transparent on why your FAAB didn't work according to your rules last year, or what is being done to fix that this year.

Why is it that every time someone brings up potential issues with your back end you just brush it off, and we are being crybabies because we want to see improvements?

[Greg Ambrosius]

We can't be more transparent than this...


There's no crying in baseball. Just good managing.

You still haven't been transparent on why your FAAB didn't work according to your rules last year, or what is being done to fix that this year.

Why is it that every time someone brings up potential issues with your back end you just brush it off, and we are being crybabies because we want to see improvements?

I emailed you personally about this, but sure I'll gladly discuss here. And I'm not brushing this or the random selections off or anything else off. When issues are brought up, we take it right to our IT group and almost every time they address it when there's a problem. We want the improvements more than you do.

If you are asking about FAAB's issues during the season and the three times we had to delay FAAB for one hour, I have posted this already. The reason for the problem was due to us auto-saving every single bid you made in FAAB. This was the perfect move to avoid anyone from losing their bids by not saving. And it worked fine during the week, but when you have thousands of owners all making bids or bid changes at once it bogs down the system. Even doubling the server space wasn't enough. So for football we eliminated the auto-save feature and now you have to manually Save your bids before logging off, just like in the past. We had absolutely no issues with FAAB for football and we don't expect any for baseball.

If you're talking about the two manual FAAB bid changes we made in Week 27 -- which included us changing a player won for your team -- that too has been fixed. Again, no issues in football and we don't expect any issues in baseball as the programming has been updated. This occurred when one owner had many $1 bids for one pickup on a lower preference. Two owners brought this to our attention in Week 27, we changed to the right owners getting the players, Darik checked over all of that week's bids where there were no other insteances of this and we have updated the programming.

If there's something else you're inferring, let me know.

[Philippe27]

This thread is ridiculous. Take any random number generator online or with Excel, run it hundreds of times then take any specific group of 10 picks and you'll notice the same patterns that people posted here.

In fact when I ran it, I didn't get the first pick until my 38th draft.

I know that no number generator is truly random but who cares. As long as there's no advantage for a specific owner, it's random enough. If you don't trust Greg with that then why are you trusting him with hundreds or thousands of dollars every year.

[TParsons]

I emailed you personally about this, but sure I'll gladly discuss here. And I'm not brushing this or the random selections off or anything else off. When issues are brought up, we take it right to our IT group and almost every time they address it when there's a problem. We want the improvements more than you do.

If you are asking about FAAB's issues during the season and the three times we had to delay FAAB for one hour, I have posted this already. The reason for the problem was due to us auto-saving every single bid you made in FAAB. This was the perfect move to avoid anyone from losing their bids by not saving. And it worked fine during the week, but when you have thousands of owners all making bids or bid changes at once it bogs down the system. Even doubling the server space wasn't enough. So for football we eliminated the auto-save feature and now you have to manually Save your bids before logging off, just like in the past. We had absolutely no issues with FAAB for football and we don't expect any for baseball.

If you're talking about the two manual FAAB bid changes we made in Week 27 -- which included us changing a player won for your team -- that too has been fixed. Again, no issues in football and we don't expect any issues in baseball as the programming has been updated. This occurred when one owner had many $1 bids for one pickup on a lower preference. Two owners brought this to our attention in Week 27, we changed to the right owners getting the players, Darik checked over all of that week's bids where there were no other insteances of this and we have updated the programming.

If there's something else you're inferring, let me know.

You did email me, but you ignored my follow-up questions many times. The issue is that your explanation doesn't make sense. It shouldn't matter how many bids there are or how long the bid string is. The logic should never change from the first winning bid to the last. It should work in a loop until all bids have been executed. This tells me that the logic was wrong all along. Is that true?

The problem is that the players had to police it themselves. If I hadn't caught mine, then the player would have been awarded to the first place team, potentially impacting the overall standings as well. If the logic was wrong all the way up to the final week, do you know with 100% certainty that there weren't other bids prior to the final week that fit the same bid scenario that weren't caught by the players?

[Gekko]

Greg - Can you ask IT if the process of choosing owners is truly random for KDS selection? If their answer is “yes”, can they describe how they validated their claim.

I don’t have any expertise in this area; however my results appear skewed a bit, and KJ (who I respect) also expresses some longer term concerns. Has an independent 3rd party validated the process as random? If so, could the results be published?

Sure, I'll ask for it all. As you can see on your League Home page, we are trying to be as transparent as possible with the draft selection process. We are showing who was randomly picked first through 15th and what draft spot they received. What I want next is the preferred draft spot they received so that I can anonymously list all of the KDS preferences people are posting and what the average draft spot preference was for each league. I have gotten requests from some writers who want this KDS information so that they can show where our top players want to draft from.

But it's also important to show the average draft spot preference received. I did this for football a few years ago when so many owners wanted to move to the back of the draft order and in most cases it was 3.0 or lower. If 12 owners on average get their third preference that's a good system. And hopefully that's the case here. While you aren't randomly getting picked early, you have received your 3rd preference, 9th, 10th, 10th, 10th and 11th. Not ideal, but not the worst.

We'll see if we can get this extra data, but I guarantee the random optimizer doesn't know your name from any other name. They are all mixed together and spit out 1 to 15 with no care of if you're a Hall of Famer or not. But sure, we'll look again and make sure we are doing it right and if we can add even more transparency with the draft spot preference received. Unfortunately, in each draft someone gets pulled first and someone gets pulled last. That's just the fact. It happens no matter what format you use.

Thanks for looking into this Greg. I look forward to seeing what you find out.

And to be crystal clear...overall, the NFBC is a fantastic event and the premiere high stakes fantasy baseball championship. It has evolved a lot since 2004. The enhancements are pretty amazing and a lot of them are a direct result of Greg’s willingness to listen to his customers.

I wouldn’t play anywhere else. Thanks again.

[chiznad21]

For viewing and conversation here are my first 15 DC's draft positions. The first 9 leagues I intentionally left the KDS alone, here are the results.

13,6,7,5,12,5,5,1,7

The following 6 leagues I played with the KDS a bit, here are the results.

3,7,10,10,1,11

[Yah Mule]

After two seasons of not getting the first pick in baseball or football, I'm on a 1, 3, 9, 1 run this year.

[Greg Ambrosius]

We have talked with IT and we will add a third column under League Home to show which preference everyone in your league received. So we will now show your random selection (again, someone has to be randomly picked last), your draft spot and now everyone's preference received. We will also have access to all KDS preferences in one spot and can analyze the average KDS preference received for each league. You can actually do that as well from this new column we are putting under League Home.

Good luck to all. As they say in Hunger Games, "Let the odds be with you."

[KJ Duke]

This thread is ridiculous. Take any random number generator online or with Excel, run it hundreds of times then take any specific group of 10 picks and you'll notice the same patterns that people posted here.

In fact when I ran it, I didn't get the first pick until my 38th draft.

I know that no number generator is truly random but who cares. As long as there's no advantage for a specific owner, it's random enough. If you don't trust Greg with that then why are you trusting him with hundreds or thousands of dollars every year.

You're 100% wrong here. If it's not random it IS creating both advantage and disadvantage for certain owners. I trust Greg, I don't trust the code that assigns these selections as there is strong anecdotal evidence over many years and numerous players.

That said, anecdotal evidence isn't enough but it should be more than enough (along with actual knowledge that code-generated random is not random - see the excellent piece Gekko posted from MIT) to commission a low-cost data study to see if over the hundreds of thousands of data points of NFBC and NFFC KDS selections the results are truly random and/or HOW biased they are.

Likewise, the argument that a high KDS is no advantage doesn't hold up. If that was the case no one should have a problem with assigning picks in order of number of years played, or money won, or alphabetically. And you know any of those methods would cause an uproar. This is no different, there is a bias in favor of some players and against others, we just don't know what is driving it.

Or an even easier solution without the data study ... admit that the code could be problematic and get out in front of the situation and go back to ping pong balls on youtube which we all loved (with or without Kay).

[Philippe27]

I disagree, it's not because something isn't 100% random that it's creating an advantage for certain owners.

There's a difference between being able to find a pattern to a random number generator or to a poker site. Say you're able to figure out the pattern on a poker site that cards are dealt in a certain order an that in x hand, after a 7 of spades will come the 10 of hearts. You've figure out the pattern and can gain an advantage from that.

However, at the beginning of the game if everyone is dealt a card based on that same pattern which will decide who deals first (or gets to choose whether her wants to deal (draft) first or 2nd or 3rd). It doesn't give anyone an advantage of knowing the pattern if you can't take advantage of it.

[KJ Duke]

I disagree, it's not because something isn't 100% random that it's creating an advantage for certain owners.

There's a difference between being able to find a pattern to a random number generator or to a poker site. Say you're able to figure out the pattern on a poker site that cards are dealt in a certain order an that in x hand, after a 7 of spades will come the 10 of hearts. You've figure out the pattern and can gain an advantage from that.

However, at the beginning of the game if everyone is dealt a card based on that same pattern which will decide who deals first (or gets to choose whether her wants to deal (draft) first or 2nd or 3rd). It doesn't give anyone an advantage of knowing the pattern if you can't take advantage of it.

The trouble is this pattern seems to carry forward for certain players each year. So perhaps it is based on a static identifier such as your name or player ID. I'm not complaining for myself here, but for a system that seems to have bias.

Using your poker analogy, suppose the poker site dealt the button upon starting a new game based on the third number in your online player ID; so user #571000 would always have an edge over user #579000. They may not know it or how to manipulate it, but a consistent edge exists for Player A over Player B to the extent that having the button is an edge, or in the case of the NFBC to the extent that having a high KDS is an edge.

Would you say it's fair that one guy has gotten, let's say, the #1 KDS selection in 20% of his leagues over 10 years (more than 3 times expected random) while another player has rec'd 0-2% over the same period? And more importantly, that this was not the result of a random outlier 4 standard deviation event but because of bias in the code (even if we and the game operator doesn't know what that bias is) and that bias is likely to continue in the future.

In such as case, wouldn't it be better to assign slots based by ping pong ball operated by someone we trust?

[DOUGHBOYS]

I've found all of this highly amusing.
Trying to Sabermetricly analyze the randomizer.
Maybe if we put in different numbers and improve our launch angle, we'll get a number one or number two pick.
If punching in my KDS faster, it'll improve my velocity and the randomizer will like me better.
Love it.

[Deadheadz]

You're 100% wrong here. If it's not random it IS creating both advantage and disadvantage for certain owners. I trust Greg, I don't trust the code that assigns these selections as there is strong anecdotal evidence over many years and numerous players.

That said, anecdotal evidence isn't enough but it should be more than enough (along with actual knowledge that code-generated random is not random - see the excellent piece Gekko posted from MIT) to commission a low-cost data study to see if over the hundreds of thousands of data points of NFBC and NFFC KDS selections the results are truly random and/or HOW biased they are.

Likewise, the argument that a high KDS is no advantage doesn't hold up. If that was the case no one should have a problem with assigning picks in order of number of years played, or money won, or alphabetically. And you know any of those methods would cause an uproar. This is no different, there is a bias in favor of some players and against others, we just don't know what is driving it.

Or an even easier solution without the data study ... admit that the code could be problematic and get out in front of the situation and go back to ping pong balls on youtube which we all loved (with or without Kay).

Funny, we play a game centered around a sport where industry analysts constantly warn us against reading too much into a “small sample size” yet some very skilled roto players are here saying the results of dozens/hundreds of KDS pulls hints at an unfair playing field.

In the realm of outcomes there are over 130 billion combinations of 1-15 and the random number generator spits out one for each league. More than 122 billion of the outcomes will not start with the number 1.

After NFBC hits the first billion leagues we can get a good idea of how fair the RNG/KDS is but after one or two thousand leagues it’s still...SMALL SAMPLE SIZE.

[Greg Ambrosius]

KJ, we've only used the SportsHub random software for one year. If you want to complain about the STATS one from 2011-17 I'll understand that, but we aren't using the same software. And the Fanball one was for 2010, and before that I believe we used random.org for a lot of the draft spots. But if you are looking for data over 10 years, it's definitely multiple software platforms that we used for the random draft selections.

[KJ Duke]

You're 100% wrong here. If it's not random it IS creating both advantage and disadvantage for certain owners. I trust Greg, I don't trust the code that assigns these selections as there is strong anecdotal evidence over many years and numerous players.

That said, anecdotal evidence isn't enough but it should be more than enough (along with actual knowledge that code-generated random is not random - see the excellent piece Gekko posted from MIT) to commission a low-cost data study to see if over the hundreds of thousands of data points of NFBC and NFFC KDS selections the results are truly random and/or HOW biased they are.

Likewise, the argument that a high KDS is no advantage doesn't hold up. If that was the case no one should have a problem with assigning picks in order of number of years played, or money won, or alphabetically. And you know any of those methods would cause an uproar. This is no different, there is a bias in favor of some players and against others, we just don't know what is driving it.

Or an even easier solution without the data study ... admit that the code could be problematic and get out in front of the situation and go back to ping pong balls on youtube which we all loved (with or without Kay).

Funny, we play a game centered around a sport where industry analysts constantly warn us against reading too much into a “small sample size” yet some very skilled roto players are here saying the results of dozens/hundreds of KDS pulls hints at an unfair playing field.

In the realm of outcomes there are over 130 billion combinations of 1-15 and the random number generator spits out one for each league. More than 122 billion of the outcomes will not start with the number 1.

After NFBC hits the first billion leagues we can get a good idea of how fair the RNG/KDS is but after one or two thousand leagues it’s still...SMALL SAMPLE SIZE.

And as you are having a laugh, I will politely suggest you're in over your statistical head here in that sample sizes need not even be into the thousands or billions to make probabilistic determinations on a data set.

[Yah Mule]

I've found all of this highly amusing.
Trying to Sabermetricly analyze the randomizer.
Maybe if we put in different numbers and improve our launch angle, we'll get a number one or number two pick.
If punching in my KDS faster, it'll improve my velocity and the randomizer will like me better.
Love it.

Juiced randomizer, Dan.

[KJ Duke]

KJ, we've only used the SportsHub random software for one year. If you want to complain about the STATS one from 2011-17 I'll understand that, but we aren't using the same software. And the Fanball one was for 2010, and before that I believe we used random.org for a lot of the draft spots. But if you are looking for data over 10 years, it's definitely multiple software platforms that we used for the random draft selections.

But much of the base code has remained the same from the old sites, has it not?

Regardless, even if from 2018 forward would not have the same bias as 2011-2017 it would remain a bias, i.e., perhaps conferring an edge/barrier to different players than in the past. Still a problem. And even if the code has changed, it's possible that the same software "module" is embedded into the new site.

Likewise, as I've said in the past, it's not just an NFBC/NFFC issue. I've seen it on FFPC and other sites I've played in the past. It's a random assignment problem based on inherently "too" simple code.

[Greg Ambrosius]

KJ, we've only used the SportsHub random software for one year. If you want to complain about the STATS one from 2011-17 I'll understand that, but we aren't using the same software. And the Fanball one was for 2010, and before that I believe we used random.org for a lot of the draft spots. But if you are looking for data over 10 years, it's definitely multiple software platforms that we used for the random draft selections.

But much of the base code has remained the same from the old sites, has it not?

Regardless, even if from 2018 forward would not have the same bias as 2011-2017 it would remain a bias, i.e., perhaps conferring an edge to different players in the past. Still a problem.

Likewise, as I've said in the past, it's not just an NFBC/NFFC issue. I've seen it on FFPC and other sites I've played in the past. It's a random assignment problem based on inherently "too" simple code.

Absolutely not on the code. Completely new code on everything, including the random draft spot code.

Go to random.org and pop in 10 names. You can shuffle them any number of times. It looks pretty simple, but hopefully not "too" simple.

[KJ Duke]

https://www.random.org/randomness/

This link offers a good explanation on true random vs psuedo-random (not really random).

Random.org claims to generate true random.

So is this embedded into the NFBC code now, and has it been used in the past, or does the NFBC use a PRNG (pseudo random number generator)?

Maybe we can get to the bottom of this.

[Baseball Furies]

As someone who virtually never gets a number one pick in literally dozens of drafts that I do every year whether they be the DC's or Live Events, I'll just say this. First of all, anyone who knows me knows that I often like the back-to-back picks. As a result, I will mostly set my KDS to start 1,15 etc. from there mixing it up a bit with the preferences towards the front of the draft. Do I get a lot of 15 hole picks? Yes. But do I proportionately get a lot of #1's? Not even close. Could probably count them all on one hand in my 12 years in the NFBC. KJ made a great suggestion about the randomizing process, but probably the best thing that he said I have been pushing for years, and it's time to innovate and go outside of the box on this.

Several years ago, when I came up with the "Iron Balls" KDS idea for the MTM Ultimate when draft spots were drawn randomly from a bucket using baseballs with each respective league participant's name on them, it became a huge hit and one of the highlights of the draft that everyone looks forward to. Not only does this ensure that it will be randomized without the use of a computer program, it makes for some great anticipation, interaction, excitement, and drama doing this live right before the draft. It also adds an element of strategy as well in deciding what choice to make on who you want to draft behind, in front of, next to, etc. It puts all the competitors on a level playing field with the only advantage if any being how well people did their "homework" during the pre-season drafts leading up to the big money live events. I think it's time to make the change to this for the live high stakes events at least as KJ was saying. There's always going to be push back for various reasons when you try to change something from what's accepted as the norm and comfortable, and I can't begin to list all the changes that have been made to these contests over the years since I started participating in them, but in the end we all adapted, adjusted, and came to love the new normal we have today. So rather than poo-poo this contest change out of hand, this is yet another idea that should seriously be taken under consideration for next season in determining drafts spots for the live high stakes events, Random Slot Generator be damned. So can we at least get a survey going on this?

And one last thing: Correct me if I'm wrong, but the NFL does this all the time when it comes to making rules changes to the game, and this is supposedly a PROFESSIONAL SPORT with tens of billions at stake every year! (about $25 billion annually according to Forbes). So if they can do it, so can the NFBC. Thanks.

[Gates]

here we go (AGAIN)...

so far, seven "random" pulls out of the hat this year.

when i was selected out of the hat in each of the 7 leagues: Choice Order: 4, 9, 11, 12, 13, 14, 14

really?

hopefully my "luck" starts to improve for the remainder of the draft season; however this is beginning to get mind-numbing how "unlucky" the NFBC random slot generator has been (at least to me)

I've done two cutlines.

I drew #4 in the first one.
Second one? #4.

Same non-random generator as every year. For any spot you get, you're more likely to get that spot again in the same year. I've seen for 10+ years.

Manual picks should be happening for high-value leagues like the Main and higher.

If you don't believe me hire a statistics grad student intern and have him run a study on KDS preference allocation doled out by the system. I am highly confident the system will be shown as statistically not random.

I would say two leagues is a small sample size for conclusions like this. Here is what Gaetan Lavoie has gotten in the NFBC Cutline Championship so far as his random selections:

3, 4, 6, 9, 8, 9, 4, 10, 3, 6.

Should he be complaining that in 10 tries he didn't get a single 1 or 2? Or does he complain that only 4 of 10 were in the top half? Or does this show that randomly he landed all over the board?

I get it. Folks want us to check the randomizer and sure we will AGAIN. It's random and it doesn't care what your name is. It just spits out 10 names, 12 names, 15 names and then runs the KDS to give you an even better spot. But we'll keep analyzing the data and being as transparent as possible. That's what we'll do.

Well well well, after exposing my *unluckyness* after 10 drafts, on the very next draft I get to pick 1st. Coïncidence? Hmmm. Thanks Mark for flagging this obvious injustice. LOL

Gaëtan

[Deadheadz]

FWIW

Just found out the draft order for my 8th entry.

So far:
15,8,9,4,10,1,6,2

The RNG seems pretty random to me.

[Gekko]

Since we now have access to 2018 teams, I've validated all of my league "choice orders" for 2018 + 2019 (to date). here are the results...

Total of 32 leagues...
Choice Order: 2
Choice Order: 3
Choice Order: 3
Choice Order: 4
Choice Order: 4
Choice Order: 4
---------------------------
Choice Order: 6
Choice Order: 6
Choice Order: 7
Choice Order: 7
Choice Order: 7
Choice Order: 7
Choice Order: 8
Choice Order: 8
Choice Order: 8
Choice Order: 9
Choice Order: 9
Choice Order: 10
Choice Order: 10
---------------------------
Choice Order: 11
Choice Order: 11
Choice Order: 12
Choice Order: 12
Choice Order: 12
Choice Order: 13
Choice Order: 13
Choice Order: 14
Choice Order: 14
Choice Order: 14
Choice Order: 14
Choice Order: 15
Choice Order: 15

TOTAL # OF 'CHOICE ORDERS' IN EARLY THIRD: 6 OF 32 (18.8%)
TOTAL # OF 'CHOICE ORDERS' IN MIDDLE THIRD: 13 OF 32 (40.6%)
TOTAL # OF 'CHOICE ORDERS' IN LAST THIRD: 13 OF 32 (40.6%)

Summary ---> NFBC Random Slot Generator RANDOMLY HATES me. Not a single #1 Choice Order spanning 2 years AND 32 leagues. And some owners get multiple #1's in the same year. Makes me wonder how successful I could be if I actually benefited from "randomness" as much as my competitors