Could this be possible? (1 Viewer)

[gianmarco]

I apologize in advance if some of this stuff doesn't make sense. After reading a couple of threads, this idea kind of came across and I haven't exactly completely thought it through. So, this is going to be more along the lines of free thinking.

So, this place is amazing for analysis about FF with numbers and statistics and evidence for what we do and how best to approach putting a team together. Things like VBD can help us look back at the previous year and assess how much of an impact a player had and then determine how to use that information in the future. And while that information is incredibly useful, the one thing missing from a lot of what we do is the end result and how it all fits together.

Ultimately, the bottomline in fantasy football is winning championships. We've seen polls around here like "Did the team with Brady (or LT or Moss) win your championship last year? Or, "So how well did the team with Peyton Manning do?" when discussing the advantages of drafting a QB in the 1st round.

So, what I was thinking was trying to find a way to determine what the best "strategy" for a draft is. Of course, there will never be a "right" answer, but we may be able to use %'s to help us out. So, this is what I had in mind.

This is a project that would take a lot of work and a lot of time to compile, analyze, and complete. Essentially, it would involve taking real teams with real results and submitting them as data. As large as the FBG community is, this is a unique environment that something like this could actually be done if a few people had the time, willingness, and resources to do it.

Essentially, we'd have as many people as possible submit how they drafted and at the end of the year submit their final results. This could be done a couple of ways, both analyzing the order of positions taken and also analyzing the impact certain players had (i.e., what were the results of teams that draft Randy Moss in 2007).

Of course, there are going to be barriers and inclusion/exclusion criteria in something like this. First of all, this really only works with redrafts. It would also have to be a standard scoring/starting rosters with only a few possibilities in order to make analysis worthwhile. And of course, there are going to be lots of confounding variables that simply can't be controlled for such as draft expertise, H2H luck, talent of opposing owners.

Still, what we might eventually see is that teams that take a QB in the 1st round have a much larger or a much smaller % of championships or playoff visits. We may see that taking WR's early is the way to go. We may see that going for a TE early is the way to go.

So, 1 year of data simply wouldn't be enough. It also wouldn't be enough to just have ~100 leagues or so. This is something that would require a large # of members submitting data and some people to compile it all and keep track of it. It would then involve submitting end of year results and doing this over a long period of time as a prospective study. However, with the # of people here and the # of leagues that most play in, I think it IS possible.

What do you guys think? Crazy, dumb idea that has no chance or something that might be possible to do? Is it even worth it or is it going to tell us a lot of the same things we already know? Curious what others think about this......flame away.....

 

[doowain]

You wanna get married?

ETA: I think it's a great idea. I'd actually been thinking about something very similar to this in the last couple weeks. I wasn't quite sure how to go about it and the more I thought about it, the more I realized it would be a huge undertaking in order to get valuable results. Seeing as how I'm great at coming up with ideas, but horrible at putting them in motion.....I brushed it aside.

I think this would be very valuable if done right and I'd be both excited and honored to be a part of it if my help is needed.

 

[gianmarco]

Btw, I'm not sure I made this clear above. This is an example of what would be necessary.

We would set up a "default" roster, scoring, etc. criteria for a league. Everything else would be excluded. So, for example, we would only do 12 team leagues that start 1 QB, 2 RB, 3 WR, 1 TE, 1 K, 1 Def with standard scoring and PPR. Every person that is in a redraft league that fits those criteria would submit their positional draft order. So, for example, if I were in a redraft league that qualifies, I submit my data as RB/WR/WR/TE/RB/QB/RB/WR/QB/WR (say we limit it to first 10 rounds). Then, at the end of the year, I give my final position.

If we get enough of these, we then put all of them together and can see which draft orders had the best results. Which had the most championships. Which had the most playoff berths. Which did the worst.

We essentially all do this on our own without realizing it when we analyze our end of year results and then apply them to our future drafts.. Those that have played FF for a few years look back and figure out what works best for them and what doesn't. What this would do is greatly increase the sample size. So, if instead of the teams that you get to see in your leagues, we could have 1000 teams to get #'s from and do this over the course of a few years. By doing this, we may see some surprising results or we may see that it really doesn't matter and the order which you draft in doesn't really matter in the end.

 

[shakeybarn]

Standardize the process and methodology. But allow for various scoring/league types. Maybe minimize"

12 and 16 team formats

Scoring for TDs (4 vs 6)

Long TD bonus vs no TD bonus

2 pt PPR vs none or small

and so forth.

Problem is , I see most of the data after-the-fact... and not intel... if you tie in, say a VBD or ADP... what their projection is beforehand... THEN YOU'VE got something.

I've plotted value (beauty of an auction league) vs actual outcome to find the actual efficient frontier of $1/fantasy point. Basically, trying to find the greatest value return trends.... and there is a sweet spot, where you get, for example, $30-40 value for $10-20 cost... and that's the art on the second tier of your team.

 

[doowain]

Rather than having everyone submit results and such, I wonder if there is some way to get ahold of MFL and ask them for permission to do a study of the leagues. I'm not sure how feasible this is, but all of the leagues on there are open for the public to view I believe. Getting a listing of all of the redraft leagues would be ideal.

So, if we could set up a listing of different situational criteria (drafted QB early, drafted two WRs in the first two rounds, etc) and go through league by league, you could put together a nice database of teams that fit each situational mold. At the end of the year, go back to those leagues and find out where each team finished. This is a simplistic solution, but I hope you see where I'm headed with it.

 

[gianmarco]

Sure, doesn't have to just be one league type. Thing is, the more league types you have, the more difficult it could be to keep track of. Definitely something that could be expanded in the future. I think if there were something like 5 different "standard" or "common" league settings, then you could get enough information to analyze the data. And yes, this is data after the fact, but it's actual real data that you accumulate and then analyze. Prospective studies like this are actually some of the best ways to get significant results.

Also, after quickly thinking about this, we could actually get some pretty large data sets since you wouldn't just have to submit info on your team. For every redraft a person is in that qualifies, they could submit a draft order for all 12 teams (or however many) and their final positions. Thus, even if only 100 people in only 1 league participated, we would have 1200 data entries. That would be easy to achieve, I think. Getting that increased 10 fold should be quite doable. Then, run this for 3-4 yrs, and there may be substantial data to work with.

 

[doowain]

if you tie in, say a VBD or ADP... what their projection is beforehand... THEN YOU'VE got something

I don't know if this really works with the idea of mapping the "best strategy". Of course the thing you typically want to do is draft based on value, but with a "draft strategy" you are determining based up starting requirements/scoring system/etc. what is the best course of action to take in a draft before it starts. Do I go RB/WR to start or take Brady in the 1st or....? That's what I think this study would accomplish. What is the success rate of certain draft strategies and from that you could maybe determine the best way(s) to attack a draft.

 

[gianmarco]

if you tie in, say a VBD or ADP... what their projection is beforehand... THEN YOU'VE got something

I don't know if this really works with the idea of mapping the "best strategy". Of course the thing you typically want to do is draft based on value, but with a "draft strategy" you are determining based up starting requirements/scoring system/etc. what is the best course of action to take in a draft before it starts. Do I go RB/WR to start or take Brady in the 1st or....? That's what I think this study would accomplish. What is the success rate of certain draft strategies and from that you could maybe determine the best way(s) to attack a draft.

Right. Something like this could potentially prove or disprove VBD. What would happen if a certain draft order for the first 4-5 positions had an overwhelming # of playoff berths/championships? If VBD is telling you to do something different, which is more correct?Or, what if teams that took a QB in the 1st round had an abnormally low # of championships compared to other draft orders? These are the kinds of results, if given a large enough sample size, could help substantiate if what we think is the best way to draft really is.

 

[Jeff Pasquino]

While I admire the effort, step back and take a look at the big picture for a moment.

The biggest influences last year were:

Tom Brady

Randy Moss

ADP

Having 2-3 of these guys would have been a recipe to go to the postseason.

None of them were 1st round targets.

Odds are that even just 1 of them meant playoffs for that squad if they had a decent rest of their team.

The point is, if most teams were carried by a few common names that were NOT first round picks, then it really was more about identifying those players (ADP beyond Round 1 who could carry a team) than learning about draft position.

 

[Dinsy Ejotuz]

Somewhere this was done by one of the hardcore programmer/stats guys - maybe Drinen or Stuart. Looked at past seasons and told you based on position in the draft what the optimal draft strategy was on average, round by round. Look at some of the old 2006 or 2007 FBG articles.

 

[gianmarco]

While I admire the effort, step back and take a look at the big picture for a moment.The biggest influences last year were:Tom BradyRandy MossADPHaving 2-3 of these guys would have been a recipe to go to the postseason.None of them were 1st round targets.Odds are that even just 1 of them meant playoffs for that squad if they had a decent rest of their team.The point is, if most teams were carried by a few common names that were NOT first round picks, then it really was more about identifying those players (ADP beyond Round 1 who could carry a team) than learning about draft position.

I agree somewhat. Or at least, that's what we all THINK. But, is there any evidence that this is REALLY the case or is this just something that SEEMS to be true and really isn't? What if this were done and the results came back surprisingly different? I think, if this were done, that if the results came back that there really is no specific order/orders that stood out, then there is something else that determines success such as what you pointed out. That could be part of the analysis at the end of the year. If this were done, another data point that could be added is determining which teams had these key players for a given year and their respective finishes. So, if this had been done for 2007, it could be easy to ask to include which teams had these players and where they finished. This could be applied to that theory as well as looking at specific draft order. I just wonder without something like this ever having been done if we can truly say what is the KEY components of success.

 

[zanti8]

i think the most important thing for a title is the sequence that occurs once in the playoffs . what i mean is how many had brady/moss last year with a dominating team all through the season . then got beat in the playoffs . adding stats could help calculate a higher % for "points leaders" but ff always comes down to sequence . what guys have a good game in the playoffs . i had my best year last year with win % and point leading in leagues but all those teams lost in the playoffs . a couple teams that were "wildcards" won me titles . there will always be this "random" element because the sequence is to unpredictable and is what determines champions, not points assembled durring the season ( although obviousley you need those to make the playoffs) .

 

[BertJones]

I love the idea and would like to see the results.

However...

This doesn't take into consideration waiver pickups (picking up a Ryan Grant last year would have paid off big time during the stretch run) or lineup submission blunders (player x goes off while he's on your bench).

 

[Neil Beaufort Zod]

If you're going to do this, I suggest getting all participants to give you everyone's breakdown from their league and agree to reveal how they finished at the end of the year. That way you have 12 X more data. It sounded like you were just going to get that owner's data only.

 

[Sidewinder16]

Sure, doesn't have to just be one league type. Thing is, the more league types you have, the more difficult it could be to keep track of. Definitely something that could be expanded in the future. I think if there were something like 5 different "standard" or "common" league settings, then you could get enough information to analyze the data. And yes, this is data after the fact, but it's actual real data that you accumulate and then analyze. Prospective studies like this are actually some of the best ways to get significant results.

Also, after quickly thinking about this, we could actually get some pretty large data sets since you wouldn't just have to submit info on your team. For every redraft a person is in that qualifies, they could submit a draft order for all 12 teams (or however many) and their final positions. Thus, even if only 100 people in only 1 league participated, we would have 1200 data entries. That would be easy to achieve, I think. Getting that increased 10 fold should be quite doable. Then, run this for 3-4 yrs, and there may be substantial data to work with.

I'm sure there are enough of us FF here that keep track of league records for multiple years that you could probably get a decent amount of data from the past 3-4 years (or more).Unfortunately, I can't help you out as I haven't participated in a redraft in years (just keeper and dynasty)

 

[gianmarco]

I love the idea and would like to see the results.However...This doesn't take into consideration waiver pickups (picking up a Ryan Grant last year would have paid off big time during the stretch run) or lineup submission blunders (player x goes off while he's on your bench).

There are going to be confounding variables, however, you're forgetting one thing. If you get a large enough sample size, these types of things should be eliminated from affecting the outcome. If we were to get 1200 team data points, 1/12th of those teams picked up R. Grant and benefited. However, with a large enough sample size, it should be evenly distributed as to not affect the final #'s. If you were to have 120,000 data points, his effect on final standings as it relates to what we're looking at would be minimal. Same goes for other teams, players, luck, etc. That's why this would require a very large sample size to be effective.

 

[gianmarco]

If you're going to do this, I suggest getting all participants to give you everyone's breakdown from their league and agree to reveal how they finished at the end of the year. That way you have 12 X more data. It sounded like you were just going to get that owner's data only.

Right. I said this up in post #6 as I realized this shortly after. It really would not take that many people to get a pretty substantial data set to work with.

 

[jwb]

While I admire the effort, step back and take a look at the big picture for a moment.The biggest influences last year were:Tom BradyRandy MossADPHaving 2-3 of these guys would have been a recipe to go to the postseason.None of them were 1st round targets.Odds are that even just 1 of them meant playoffs for that squad if they had a decent rest of their team.The point is, if most teams were carried by a few common names that were NOT first round picks, then it really was more about identifying those players (ADP beyond Round 1 who could carry a team) than learning about draft position.

But that why the proposed plan will only work with several years worth of data (say 5+) - that minimizes the one year explosion from the third round (etc).

 

[crush304]

I'm actually working on a project similar to this but I'm going abut it in a different way. I'm building a fantasy simulator, basically matching current players to players in previous years based on a range of adp or position rank. Use that to generate a pool of possible seasons for each current player and then play the league out by randomly picking one season out of the pool for each player. Do that a bunch of times and I can track how often each fantasy team wins, average points, how often each player will start, etc. I'm setting it up to work based on a full season or by game using a best ball format. Hopefully that makes sense.

I'm very early in the project at this point, I just started a few weeks ago but it's starting to come together and I'm pretty excited about the data I should be able to produce

on another note, if you want fantasy team data you can go to http://www.myfantasyleague.com/fantasy-foo...ft-analysis.php

pick the year you want, and at the bottom of the adp page there is a list 'Leagues Involved In This Draft Analysis' that takes you to the individual league pages.