Fantasy Football - Footballguys Forums
This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!
Let’s Talk About Strategy (1 Viewer)
- Thread starter taylormeetstheismann
- Start date
Kleck
Footballguy
No problem. In the future, using the "report" button to make a request like this is a good idea. Zips me your request and provides me a link right to the thread.Also, on topic... Before you get too far into this I suggest you read everything Doug Drinen has ever written with regard to fantasy football. Lots of the questions in this thread have been addressed very well in his work. He would be the first to tell you that he hasn't answered all the those questions, but he definitely shed light on lots of them.
taylormeetstheismann
Footballguy
I have enjoyed reading many of his articles over the past few years. Taking a very specific topic and scientifically analyzing it in great detail is of interest. In some ways, I'm more interested right now in the direction of the FF basic training articles that came out this year. Looking at broad categories and breaking the key elements down into finer and finer parts. I am attempting to codify my fantasy football knowledge in my own mind (and in the written word) by exercising topics on the message boards.
Last edited by a moderator:
The Ref
Footballguy
I stand by the TE comment "in principal". But I hear where you are coming from, as the top guys will represent good value as they are far out producing the bottom rung of the baseline players. The QB position is very similar in this regard. However once you get beyond TE3 or TE4 the net difference in FF points isn't as great as past years when compared to the bottom rung (or so it seems).I see people "reaching for value" for Manning, Pepper, Green, McNabb, Gonzo, Gates, and Witten (by this years standards).
Hope that quoted right.I don't fully understand the question - seems like you are making this too technical. What you are referring to here appears to be nothing more than a mock draft. You slot the guys based on expected production and the risk taking any of those players versus those around him. For instance, the risk many considered earlier this year was Holmes with his age and likelihood of injury vs Tomlinson's age and reliability - so based on your own assessment, you slot them 1 and 2 and you know the risks. There's your risk - it's more like the pros and cons of each pick and weighing them. And that should be in your valuation of the players anyways.
BigActionMike
Footballguy
How much does the handcuff ruin your chances of drafting a sleeper? Holding Priest Holmes and LJ or Faulk and Jackson could have cost you some good WW selections! After this season I believe quite a bit. I had two roster spots held up between Larry Johnson and A. BOLDIN!! with a reduced roster, this cost me plenty. I was unable to grab the Michael Claytons, Nick GOings, etc. b/c of the reduced roster space and a necessity to play guys each week to manage my bye weeks. This led me to think of ways to try to minimize the impact of bye weeks, the only one that may not be conventional, was to try to grab guys with the same bye week! In a H2H league, you are guaranteed a loss one week, but for the remainder of the year you are at full strength!!in a total points league it almost provides you an advantage in that you could be at the top priority of the WW selection for a few weeks!! (which would be great if your weeks were early) but it also costs you a great deal to start guys like todd pinkston, ricky proehl or even worse take a ZERO. that is where the strategy would be flawed
coolnerd
Footballguy
On a small roster, handcuffing reflects this problem. Esssentially, you are betting that a player WILL get hurt versus MIGHT get hurt. Also, the key is to pick the right handcuff. On the top RBs this yearHolmes- needed handcuff but did Blaylock or JohnsonLT-Chapman played but LT did not totally miss gamesGreen-Inconsistent, but Davenport hurt/Fisher not enoughAlexander- none neededPortis-not needed and was it Betts anyway?lewis- most predictable, but did you get Smith or Taylor/production not special.Duece- only one or two weeks and did you get the right guyIn every case, either the lead back did not get hurt or even with careful research a person could have drafted the wrong back-up. I would argue that in dynasty leagues or large bench leagues then handcuffing a better bet than other botom of the roster options, but if it is a redraft with a handful bench spots then your odds are better finding a fantasy back-up or working the waiver wire than guessing that your guy WILL get hurt.
EBF
Footballguy
When I think of FF, I think of dynasty leagues. Everything else pales in comparison. As for strategy, I could go on for days. In dynasty leagues, I try to build the foundation for a long term contender. I hear tons of talk about youth being overrated, but I actually think it's underrated in most cases. The real key isn't youth though. The real key is talent. Load up on talent and you'll be golden.
I play in quite a few leagues, and one of the leagues I am in with guys from the office for little money on the line. I allowed myself drinking (too much) at this draft since I stay sober at all the rest.I accidently ended up in this very situation, almost all my players were on a bye the same week. I panicked. I had never been so irresponsible! Then I thought I should just ride it out and see what happens.As expected, I lost that week, but I found myself for about 4 weeks of matchups where my opponent had 1 to 2 must play starters on a bye when I was at full stregnth. I won 3 of 4 of those matchups. Does that prove anything? No. But I would love to hear more discussion on this strategy as it is unconventional. For example, If I knew my FF schedule before the draft, I would consider taking the dive against certain owners who are usually strong (assuming the probability of loss was higher anyway). I think LHUCKS and Ridgelake are right on the money. It's all about risk management. Some of this is already inherent thinking with most sharks. We all know we won't win the league with our first round pick, but we know we can lose the league in the first round. However not many of us take that approach in each of the 4 aspects outlined by TMT.The development of a portfolio (team) that minimizes risk while maximizing returns is the utopia. Drinen has written several articles on statistical analysis he has done on QB-RB tandems on your team as well as QB-WR tandems. This is the kind of information that help you build the best risk reward portfolio.I don't claim to know the answer to all of this. Each year I have been trying to refine my strategies using risk management principles, and many of the things TMT is trying to explore will help in refining the model. I look forward to a great discussion!The one thing not yet mentioned that I think is critical is knowing the tendancies, behaviors, and risk tolerance of the other owners in the league. This can significantly affect draft strategy and roster management, and to a much lesser degree lineup selection. It is an area again that we all know about but I think deeper analysis might reveal trends and performances that can be used to our advantage.Thanks TMT!
bueno
In a class by himself
To me, a lot comes down to projections. The past is not the key to the future in fantasy football, making many predictions absolutely worthless. I think we need to look beyond yards gained/three year scoring averages to other things.For example, I never liked Koren Robinson's potential, never thought he would become the #1 receiver in Seattle because he dropped too many passes. Why? His hands are relatively small for one. How can we know that about players?The relationship between RB age, number of touches and perfomance is poorly understood.We don't have any good predictive modelling for DTs and Kickers.
taylormeetstheismann
Footballguy
I've updated and rearranged my four basic categories for decision making with the key question on which decisions are made.1. Player Evaluation - What future statisical performance can be expected from a player?2. Draft Strategy - What players should be acquired?3. Roster Management - Which players should be added or dropped?4. Lineup Selection - Which players should be started?It is my thinking that (simply put) these are the only real questions we ask ourselves.I'm open to suggests that will change this highly simplified point of view.
Uruk-Hai
Footballguy
Agreed. The lynch pin ( & weakness) of most numbers-based drafting systems ( VBD & the like) is projections. But how to improve on their accuracy? Can they be improved? Or is it just the nature of the game that there are too many variables?(Obviously, some are better at projecting performance than others. But I'd bet that even the best predictors miss more than they'd like to admit.)
What makes a successful projection, anyway? If you're within 20% on a particualr player? And is that good enough? You don't have to miss by much on a few players to screw up your whole draft sheet.
GregR_2
Footballguy
I think there's probably a lore more to be gained from discussing the various strategies, tools, techniques, schools of thought, etc, than there is in worrying about what class they fit into.
We can talk about how to better manage risk, how to be better negotiators in trades, things to look for and avoid when scouring waivers, ways of recognizing players with potential, etc... I don't really care what category they fit into. I just care that I can establish whether it will help me or not, and to understand why it helps me and how to best use it.
suspected
Footballguy
I totally agree that applying risk management principles makes sense and is a tremendous opportunity. I like the portfolio approach. Its not just about improving accuracy of projections, but knowing in advance of the draft which player projections are likely to be more accurate and why, and appropriately using this knowledge to your advantage when drafting. What are the real factors that contribute to risk (variability) in projections and how important is each factor (age, #touches, past performance, change of team, strength of offensive line, etc). How do these risk factors vary by position?
Its not sufficient to say we already do this. Yes, we do try to account for these factors but I don't know anyone that has developed an integrated, predictive model that incorporates all the important factors. Yes, there has been some study on individual risk factors but nothing I'm aware of that puts it all together. I think we could do much better if we did more analysis and applied risk management principles to have a better understanding of the relative importance of each factor by position.
Then once we have a better handle on risk, how do we use this information when drafting? How do you best develop player tiers that consider both expected return and variance (risk)? Is it better to draft high reward/risk player in middle or later rounds?
Developing player projections is just one area where these principles can be applied. They can also be used to better understand when backups should be drafted, and in Roster Management and Lineup Selection. They can also be used to help answer questions about the value/risk of drafting multiple players on the same NFL team. 1 1/2 years ago I started a thread that asked the question: Does Diversification Apply to Fantasy Football.
I'm sure there are a multitude of other areas in fantasy football where risk management principles could be applied.
Last edited by a moderator:
I think GregR makes a good point here. What we are talking about is maximizing the skill part of fantasy football---whatever the % may be---and using it to our personal advantage.What is interesting about this dialog, and where it looks to be going, is WHY I am going to make decision "x" or utilize strategy "y", taking it beyond doing something inherently or even HOW to do it. As in my business, the masses understand "how" something is done. How to do something implies one has a plan, but does not have the ability to understand and adapt to change. With all the information out there on the www, more and more individuals who take up this great hobby can follow a road map and apply a little "gut instinct" and be a competitive owner from year-to-year. However, this doesn't really maximize one's skill over another owner and luck plays a role in consistent success.To separate from the masses, one must realize why decisions/strategies are done. This provides the wisdom to recognize and adapt to changes or trends in the ff world. Again, knowing why will maximize the skill portion of fantasy football and tip it to my advantage over my opponents and/or league.Great stuff and looking forward to some thought provoking dialog this off-season.
totally agree with you here.some people are just a better judge of talent than other people. some people like ron wolf can see a guy and for some reason say man that kids going to go far, i better get him. well how come the other gms dont see it? who knows. point is that some people have vision or luck or whatever [research?], and its good to listen to those people. fortunately, dodds bryant and most of the good fbgadmins have the vision most of the time.its usually the visionaries who have some 'zany' idea, and everyone laughs at them until their plot unfolds and then its... why didnt i do that.so i think on the other hand, the only thing we have to combat ppls' natural talent is research, which is what the OP is trying to dig thru. thats also why i watch threads like these. there might be a great idea buried in here.
here's a funny strategy ive found;when you're winning dont change anythingwhen you're losing, try anything to win...i notice when im winning i look for things to change but am hard pressed to make any.when im losing, i'll put players so deep in the roster on the field.why?who knows...and yet i think when u win, there maybe a matchup or two that might be better, but its hard to bench a WR thats on fire...I think GregR makes a good point here. What we are talking about is maximizing the skill part of fantasy football---whatever the % may be---and using it to our personal advantage.What is interesting about this dialog, and where it looks to be going, is WHY I am going to make decision "x" or utilize strategy "y", taking it beyond doing something inherently or even HOW to do it. As in my business, the masses understand "how" something is done. How to do something implies one has a plan, but does not have the ability to understand and adapt to change. With all the information out there on the www, more and more individuals who take up this great hobby can follow a road map and apply a little "gut instinct" and be a competitive owner from year-to-year. However, this doesn't really maximize one's skill over another owner and luck plays a role in consistent success.To separate from the masses, one must realize why decisions/strategies are done. This provides the wisdom to recognize and adapt to changes or trends in the ff world. Again, knowing why will maximize the skill portion of fantasy football and tip it to my advantage over my opponents and/or league.Great stuff and looking forward to some thought provoking dialog this off-season.
Ridgelake
Footballguy
TMT has boiled down FF to its basest elements. I think this provides a good framework from which to start evaluating RISK and RETURN issues.TMT's first modality is Player Evaluation. This area will likely be the easiest to evaluate RISK. His following three modalities, Draft Strategy, Roster Management, and Lineup Selection all involve further steps of RISK involving portfolio RISK.My thoughts on these subjects are in the development stage and are a work-in-progress. So my post may be a bit rambling.Before going further, lets make an assumption that will simplify the measurement of RISK. I’m going to assume that standard deviation (“SD”) will provide a reasonable measurement of RISK. I suspect that there could be better measurements of RISK in FF and I question whether every situation is a normal distribution. But this is a term most people are familiar with and most database/spreadsheet programs can calculate it. If you need a definition or want further clarification, google it.Lets look at the types of risk that are involved in Player Evaluation from a season performance standpoint. Off the top of my head, here are the major risks that I can think of:1) Injury to Player – Obvious choice2) Injury to Teammate – Example, how is Harrison affected if Peyton gets hurt?3) Changes in coaches/schemes. Will Denny Green help Boldin’s numbers? 4) Change in teams – How will Dillon do in New England?5) Regular player variability – Is he steady eddy (Alexander) or rather variable (Moulds)?Injury RISK is something that no player can avoid. If you drafted David Boston this past year, you were screwed. Same with Steve Smith and Charles Rodgers. But this is an area that can be analyzed. Earlier in this thread, a poster mentioned some analysis that showed players with past injuries are less likely to finish all 16 games but the difference in games played is about 1. Finding this analysis would be a good start to understanding the amount of risk that injury represents to players. It should be fairly easy to calculate the average and standard deviation of games played by player. Looking at numbers based upon positions, age, injury type, etc. will help understand the injury RISK issues. The effect of Injury to a Teammate is likely harder to judge. Someone with a good database could probably cook something up here. But I suspect that it would require a lot of individual attention to give consideration to factors such as: 1) the capability of the backup (Davis/Foster vs. Peyton/Clipboard); 2) the position hurt and the position affected (QB hurt affecting WR, etc.); and 3) the reliance of one person on another (Moss’ effect on Cpep). Hard statistical data would probably only provide gross parameters. A lot of subjective judgment will probably be needed to estimate Injury to Teammate RISK. Changes in coaches/schemes would probably have some statistical relevance. An extreme example of variability here is KC’s defense still sucks under Gunther Cunningham. But Washington’s D is hugely better under Greg Williams (and some new players). It should not be too hard for someone with a good database to evaluate changes in performance with changes in head coaches and/or coordinators. It could at least provide some insights into the degree of change (standard deviation) that could be expected. Change in Teams should also be meaningfully analyzed with standard deviation. No one knows for sure how a new player will do. Who would have thought that New England would turn into a running team with Dillon? How wide were peoples’ projections on Portis’ first season in D.C.? Garcia in Cleveland? T.O. in Philly? It should be fairly easy to measure average and SD of the change in performance from one place to the next.Regular Player Variability is the catch-all. It also becomes more subjective on what measurement to do compare against. For the others, the analysis is more straight-forward. With injuries, its pre versus post performance. For changes in coaches and teams, its one versus the other. But for Player Variability, what do you measure against? Last season? An average of the two prior years? Perhaps it makes sense to compare the last set of Draft Dominator projections against actual performance. All kinds of drill downs could be performed on subsets. How much variability among the top 10 RBs? How about WRs ranked 20 to 40? Based on projections or actuals?I am going to stop here and let these ideas on Player Evaluation get bandied about. I am sure that the great minds here at FBG will be able to take these ideas and improve upon them. After we digest some of this information, we can move on to the next step of how to deal with the risks we are estimating.Regards
LHUCKS
Footballguy
I don't know where you came from but when I read your posts I feel like I've found my long lost twin brother
BostonFred and GregR are two other guys that I definitely see eye to eye with on this topic as well, so look out for their posts as well.Anyhow, we definitely can put our heads together over this offseason. I wish I could elaborate more on my theories but I just don't have the time right now as I've started a new career professionally that is very demanding. I'll be here all offseason though as usual....beating up on the FBG rankings in the Spring 
Last edited by a moderator:
We also need to look at what's cutting edge regarding FF itself:IDPDynasty leaguesAuction - unsure on thisWhat's cutting edge analysis for QB/RB/WR is beyond cutting edge for PK/DEF and that's beyond IDP & Dynasty.I'm trying to think of things that I know how to use from more basic leagues and apply them elsewhere, specifically in my large roster IDP heavy (start 11) dynasty league.Someone also said that schedule management is the same as roster management, but in one guy's case it wasn't he set himself up to play weak teams based on what division he was in - it had nothing to do with rosters and everything to do with evaluating other owner's skills. I'm not sure what I'd call this: league management/rules management?
RoarinSonoran
Footballguy
What's wrong with a .500 team? My 7-7 team won it all this year, and my 6-7 team finished 2nd.
I think this is too simplified, since it doesn't take into account the interactions among more than one item. As an example, #1 and #3 in combination raise an issue that I think is equally important: Timing. Namely:- At what point during the season can you adequately judge that a player is better/worse than you originally projected?- At what point during the season can you adequately judge that a player is better/worse than the league consensus?This is the fundamental decision as to whether to try to acquire or release a particular player, whether through trade or FA.--If you want to keep this list, I would delete #1, since player evaluation is ultimately the driver of all other decisions. Let's reframe the categories like this:1. Preseason a. Accurate projections - point estimates b. Accounting for (variations in) uncertainty c. Strategy -- determining where to find the best value, to guide your decisions d. Contigency planning for the day of the draft/auction2. In-season: decisions affecting a single week a. Evaluating opposition -- is the matchup better or worse than average for a player, and by how much? b. Finding one-week wonders c. Risk management -- multiple starters on the same team (or in the same game) i. in your lineup i. across your and your opponents lineup. d. Weather and other externalities3. In season: decisions affecting more than one week a. Revising your player estimates i. How much to increase/decrease your expectations based on actual results ii. Addressing uncertainty about future performance -- how sure are you? iii. When to take action based on new estimates - don't jump too soon, but don't wait until every other player makes the same adjustments b. Planning for the future i. Bye weeks ii. Playoff matchups c. Maximizing value throughout roster i. Frequent starters -- identifying your relative weakness, evaluating possible risks ii. Infrequent starters -- solid backups vs. handcuffs vs. potential breakout players. iii. Trade-offs between both -- improving depth vs. improving your expected lineup. d. Risk managementThere is a lot of repetition in each category -- with a little work, this could become a matrix of sorts, with one axis reflecting the circumstance (preseason, short term, long term), the other reflecting things to consider (making an estimate, judging uncertainty in estimate, planning ahead, identifying relative value, risk management). --A big challenge in all of this is trying to quantify probabilities given small sample sizes. It is easy to use rules of thumb to guide us in ff decisions -- someone mentioned "Don't draft rookie WRs." These rules are both good and bad. Good because such rules are usually preferable to the absense of such a rule (rookie WRs rarely meet popular expectations), but bad because there is no measurement of degree. How much worse do rookie WRs do than other WRs in similar situations? How much more are rookies overvalued by the ff public compared to other players?
Uruk-Hai
Footballguy
I like the way you've laid this out. Each point listed could/should have its own discussion ( I realize there's tons of overlap). We've had staff/member articles in the past addressing many of these, but I think that an in-depth discussion on different items may help get us further along to something that's truly "cutting edge".
Wilbur Wood
Footballguy
Rookie WR's can and do produce and can win championships. However, it is usually very uncertain as to which rookie WR's will produce in mid-August.IMO, the key to success with rookie WR's is to target them via trade or WW during the first 6 games of the season.
Colbert and Evans were available on and off on the WW in most leagues early. Moreover, either one could have obtained "cheaply" via a trade if one targetted them at the right time. Clayton would need to have been targetted earlier.
Rookie WR's can win championships, but the lower risk/higher return strategy is to go after these guys in Sep/early Oct and not August as by then more uncertainty has been removed from the landscape but still before Mainstream Amercia has embraced guys like Evans and Clayton as studs.
crush304
Footballguy
I think we can greatly improve the way we us ADP for Drafting.Currently we just guess when we should wait to draft a player based on his ADP. For example we have a player with an ADP in the 6th round and have him ranked in the top 20 overall we would wait until, what, our 4th? pick to draft to draft him. If we know the ADP and the Standard Deviation for each player we could calculate the percent chance the he will be avaliable at each of our draft positions. This alone should be a big improvement to what we do now.A step further you can use this to decide if you should take a player now or wait until your next pick. You need to compare the loss you are taking this round by not selecting your highest ranked player to the potential gain you will receive from your next pick if that player is still avaliable. You can calculate the loss for the current round by subtracting the value of the lower ranked player from the value of the higher ranked player you are trying to slip through. You can calculate the expected value of your next pick by comparing the player you are going to slip through with the worst player you would be forced to draft with your next pick. The calculations are easiest if the player you are trying to slip through is the same position as the worst player that you would have to draft for your next pick. To calculate the value for the next pick you multiply the percent change that you highest ranked player will still be avaliable by the number of points you project him to score. You then add to that (1-percent chance he will be avaliable) times the number of points you expect the worst player you'd have to draft to score. You can then compare that to your baseline for that position and come up with your VBD # for your next draft slot. You subtract the value of the worst player from the value we just calculate and that give you the on average expected gain by letting the top player slide to your next draft slot.If the gain is better that the loss you let the player slide, if the loss is more than the gain then you pick the player now. Sorry if the math is a little confusing, I'll come back and post a better example later. Let me know what you think
Jackal King
Footballguy
Crush,I've been doing something similar for the last three seasons, but I do not use it as precisely as you have outlined. But...the concept is the same.I extract the mock drafts from ANTS each year leading up to my draft. I have a program that essentially let's me enter a name and a draft position and it returns the percentage likliehood that that player will be available at that spot.It will also tell me the percentage liklihood that the named player will be available in each of my picks in the draft. Just for simplicity...if Favre is drafted 3.2, 3.8, and 3.12 in three mocks, and I queried my database for Favre at 3.9...it would return a 33.33% chance of my ability to draft him with the 3.9 pick. I have found it to be pretty accurate...and it keeps me from reaching for players earlier than I might otherwise.I have found it especially useful with tiering. If I have a tier or bucket of players all ranked together...I'll keep picking other spots as long as I've got at least a 75% chance of getting one of the players in the bucket.I'm looking forward to your further comments here, as I think it might allow me to improve my techniques.
Ridgelake
Footballguy
Well, my feeble little mind has been continuing to work through the problem of measuring risk in Player Evaluation. Over the last couple of days, I have had a BFOTO. Blinding Flash Of The Obvious.
Projections are screaming out for MULTIPLE REGRESSION ANALYSIS. MRA takes a number of explanatory variables (i.e. last season's results, new coach, etc.) and calculates their effect on a dependent variable (i.e., upcoming season's performance). It gives you an exact equation and tells you how much error is left unexplained by those variables. It makes so much sense that I can't believe it hasnt been done by tons of people. I just googled it and didnt find anything. I am truly amazed by this. I also googled MRA and here is a decent link that explains it in more detail. Clicky
This tool could be used for both annual projections as well as weekly projections. It will take some work to play around with determining reasonable explanatory variables. But it should be doable. I just looked in Excel, and my older version can handle 16 explanatory variables. That should be more than plenty to get us started on getting some reasonable estimation formulas.
What does everyone think of this?
I would be willing to play around with the data. But can someone point me to where I can get good data to use in Excel?
Regards
Projections are screaming out for MULTIPLE REGRESSION ANALYSIS. MRA takes a number of explanatory variables (i.e. last season's results, new coach, etc.) and calculates their effect on a dependent variable (i.e., upcoming season's performance). It gives you an exact equation and tells you how much error is left unexplained by those variables. It makes so much sense that I can't believe it hasnt been done by tons of people. I just googled it and didnt find anything. I am truly amazed by this. I also googled MRA and here is a decent link that explains it in more detail. Clicky
This tool could be used for both annual projections as well as weekly projections. It will take some work to play around with determining reasonable explanatory variables. But it should be doable. I just looked in Excel, and my older version can handle 16 explanatory variables. That should be more than plenty to get us started on getting some reasonable estimation formulas.
What does everyone think of this?
I would be willing to play around with the data. But can someone point me to where I can get good data to use in Excel?
Regards
crush304
Footballguy
ADP with Standard Deviation should take into account homers as long as the numbers have been taken from enough mock drafts. But you may also need to take into account tendencies in your league. For example lets say you want to draft Lee Evans and you calculate that he has a 90% chance to fall to your next pick from the ADP data and that on average you are better off letting him fall. It's likely that the 10% of the time he gets drafted before your next pick it is by buffalo homers so if you know your league has several buffalo homers you either might want to lower the percent chance that you think he will be avaliable say to 50% and then see if you still have value be letting him fall or just not take the risk and draft him now.This system will not gaurentee you the best player everytime. But taking the highest ranked player everytime prevents you from having a much better draft if some of you highest ranked players will fall another round or too.
crush304
Footballguy
Excelent, it sound like you are further along than I am. I like the idea of using tiers. How do you get your data, do you have to go through a lot of individual drafts and import them into the computer?I'll show some examples and see if they can help you out. I'm just making up rankings here so don't worry about those.Lets say you have the eighth pick in a 10 team draft so you are going to draft 8th and then again at 13. when it comes to your pick you have Domanick Davis as your highest ranked player but his ADP is 13, right at your next pick. Edge who has an ADP of 5 also fell to you but you have him projected to score 10 less points than Davis.D Davis 220 points ADP 13Edge 210 points ADP 5So if we pick edge now instead of D Davis then we loose 10 points for this pick. Lets also say that the worst player we would have to draft at our next pick is Kevin Jones and we have him projected at 195 points. If we let Davis fall then there is a 50% chance he will be avaliable at our next pick so half the time we draft Davis and half the time we draft Jones so on average our next pick is worth 207.5 (.5*220 + ((1-.5) *195) points if we let davis fall. We know we lose 10 points from our first pick by letting davis fall but we also gain on average 12.5 points (207.5-195) from our next pick by letting davis fall. It is also important to remember that in this example it is very close and you don't get 207.5 points for your second pick if you let Davis fall you get 220 points 50% of the time and 195 50% of the time. In this case since it is very close you might choose not to take the risk. That you still have to decide. Some other case will not be close and the payoff can be very high by letting the player fall.
Plenty of great information and here is all I can add. It is simple but works.If you have been playing with the same group of owners for a period of time, 3 years or longer, then begin to track drafting tendencies of each owner. It works far better than the ADP information and is far more accurate. Plus, it is historical data and trends from YOUR league versus a consesus. True, you can only plug in positions but, if you are using your own rankings and trust the trends from your league, then you should be set. You can back track and fill out names to positions based on your personal rankings. Log when certain owners take certain positions. Log what owners are willing to reach; the owners that will wait and the owners that will gamble. I have been playing with the same group of guys in a dynasty format for 17 years. Like clockwork every draft owner will drift back to their bad habits. This is a great compliment to any ADP information you have.
Wilbur Wood
Footballguy
While it may be helpful to come up with some historcial "decision rules" that can be used at the margin relative to factors that may influence production, I don't believe that projecting football statistics (which is what this all boils down to) can be measured and projected with anywhere near the same level of multi-factor dependency as contrasted with baseball projections.
Ridgelake
Footballguy
Wilbur,One of the great things about MRA is that it tells you how much error is unaccounted for by the explanatory variables. And until the numbers are run, we can't say for sure how accurate such a technique would be. That said, it would not surprise me if there is a lot of unexplained variance. But again, until the numbers are run, we won't know for sure.
I think that taking a look at you draft position each round and comparing it to ADP is an overlooked idea. I was in a 10-team league this year and knew I would be picking 9th and 12th in the first two rounds. I also knew that in all likelihood that I would not like the options left at those picks. I waited too long to get a trade together and ended up suffering the consequences. You should also be able to get a general idea of who will be available through most of - if not all - the draft with your picks. If it appears that you will be consistently missing out on certain tiers of players based on their ADP you really need to do what you can to trade out of this situation.
Sounds great until you begin applying it to other examples. Lets take Curtis Martin.Over 30 running back with declining touchdown production in previous year made him a terrible risk. The FBG expert ranked him the #19 running back in preseason, presumably because of his this factor.However, if you took Martin in the 3rd or 4th round you probably made your league playoffs. The same analysis could be applied to Muhsin Muhammad as well. If you had both Martin and Muhammad you most likely won your league championship. Risk analysis would have steered you clear of both players.
Uruk-Hai
Footballguy
But what does "nail 60-65% of your evaluations" mean? Does it mean that you have 60% of the players ranked in the correct slots? Or that you have them ranked within 5 spots (for example) of their EOY rankings? Or that your projections for players are within 10% for 60% of the players? Or does it mean that 60% of the players you "ended up with" were "good"?These questions aren't necessarily directed at you, abrecher, but we talk all of the time about the players we were "right" or "wrong" about. What does that mean? And, by extension, what is a good projection?
coolnerd
Footballguy
IMO, most of the ---- "sux" post come from the fact people misunderstand this concept. If you draft a WR in 6th or 7th round as the 2nd Wr with updised to be a number, but he turns out to be inconsistent but still average in the top 20 or so then you have drafted correctly and that is a win. Too many people feel like a win is taking a guy in 13th round and having him end up as the number 3 WR in fantasy scroing for the year is the only win. What i can't do is give a numeric variance as to describe that number. Say the 6th rounder is preseason expectd to average 8.5 points (standard). Is a ten percent negative variance (7.8) enough to say he was a bust. It is not that simple. if this guys hot weeks come when your number 1 was on a bye, or helped you beat the league header in a shoot out or was on championship Sunday then that guy is not a bust. One of the problems with the idea of a good pick is that the guy on your fantasy roster is like a player on an NFL roster. A part of his evaluation is as a part of a greater whole. For example, that 4th RB who is performing above what you expected looks like a win, but if your first three backs are still better and you don't trade him for something else to help your team then your evaulation is correct, but then his value for the owner is limited. On a team with a couple poor slection and/or injuries at RB that same guy could be the difference between being compeitive or being a fantasy bum. this is what a trade should attempt to resolve and sometimes why on paper a trade seems unbalanced just looking at the players/pick involved, but is often a good trade for both sides. Of course even if I had a clear way the describe what a evaluation win is then who many does it take to win a championship. there maybe some general rules but we have all either been or seen the 13-1 owner go into the fantasy playoffs against some 7-7 guy with who is next to last in scoring slip in because the rest of his division is awful. The 13-1 guy has his only bad week and may even the points he need sitting on his bench but he loses to a team that is held together by threads. Is there a percentage that will get you a win all the time? Of course, not, but is there a percentage that tilts things more in your favor, probably. Finding that number and definition what a good evaluation is seem to be what's elusive.
crush304
Footballguy
Ridge,thanks for the link, this is probably a good place to start for stats.
pro-football-reference.com
Where do you see this going? It seems like this would be a good tool to analize a lot of the subjective information that would be used when doing projections. Say for a rb.
what is the players oppertunity to play?
how comitted is the coach to running the ball?
how good is the teams defence?
what is the players injury risk?
how difficult is the teams schedule?
what is the overall tallent level of the player?
...
I would think a lot of us would go through these and many more question when projecting a player. It would be interesting to use MRA to see how our perceptions of a players situation affect our predictions. The difficult thing is that the info would have to be recorded when the projections are made so I don't know if it would be possible to run a historical analysis but it would be a good reason to start recording all of this information when making projections.
Am I way off on this or do you think that this is something that could be viable.
Ridgelake
Footballguy
Crush, Where do I see this going? Well, the answer will depend on what we find.
But seriously, a lot will depend on what data is available. Assuming good data is available, we could run MRAs to find what inputs are the best estimators of future performance. I can envision MRA being used for both season performance and weekly performance.One of the by-products of the analysis will be an understanding of what players and variables are correlated and by how much. Correlations will help form the basis of the next step of dealing with risk, portfolio risk. In the last couple of days, I have been digging up my old textbooks to relearn how portfolio theory can be used to minimize overall risk. One of the key factors in portfolio risk management is understanding the correlations in performance between the assets held. So that is where I see this going. First to understand TMT's Player Evaluation. And then moving into his team and roster management areas.I'll check out that link to the stats data. It will be fun to see where this might go.
Spartans Rule
Footballguy
An idea had for projections that I'm still playing with in my mind would be to break each team's stats down in to situations (down and distance, time remaining and score, field position, etc), then project how often they are likely to get in these different situations and who will get the touches when they do. The data dominator would be great for this.I think you could get a lot better handle on each team's offensive philosophy and the touches players are likely to get this way.
taylormeetstheismann
Footballguy
The reason I categorize decision making into these four areas is that I can only think of four decision types we make as owners.What future statisical performance can be expected from a player? (Player Evaluation)
What players should be initially acquired? (Draft Strategy)
Which players should be added or dropped? (Roster Management)
Which players should be started? (Lineup Selection)
Obviously, I am being purposely simplistic in my distillation of the essential questions. There are, of course, a myriad of other questions one can ask themselves. But I believe that these are "the" fundamental questions that must be answered and the answers are what trigger decisions.
Part of my objective in this thread is to collect cutting edge ideas or even just different viewpoints that would enhance our mutual thinking on the subject of strategy. So I'm open to the suggestions of anyone that can add an essential question that I haven't accounted for in my thought process on decision making. Maybe I will agree, maybe I won't. I'm certainly not demanding anyone accept my way of thinking a correct even I continue to promote it in the dialogue. I plan to re-read this thread several times during the off-season as I prepare for next year. I also plan on creating a few new threads (a few of which I'm working on now) to focus on more specific topics that may be of interest to others. It's a long off-season until draft day..........
HULLOBUDMAN
Footballguy
Forgive if this has been covered somewhere in here. In addressing category 3. Lineup Selection is there any statisical analysis concerning players preformances after being listed on the injury report? What average decrease in numbers players produce when listed Ques, Prob, Doubtful?Possibly even the percentages that each position plays full games as opposed to partial games in conjunction with each position and status.One step further, an average decrease in production in conjunction to certain injuries. Randy Moss, for example this year with his hammy. Would it be possible to look back 3 years and see what the decrease in WR #s is following a hamstring injury after missing one week? after playing the 1st, 2nd...etc.?I typically will not start a player until he has been back from the injury and had average production one week. This usually works well but I don't have anything to back it up with.
Last edited by a moderator:
The #1 thing that helps me is my one league that drafts in early july.I am forced to sit there an analyze all 16 rounds. I make mistakes in it, but I take it seriously. Its not some silly mock that probably peters out towards the end. It gives me a way to grade myself and recognize my mistakes. My drafts later in the summer are much, much stronger.
Similar threads
