So does Andre.
You'd think a stathead like me would be furiously preparing, calculating correlations and running endless mock drafts. There was a time I did all that, but now I don't even pay attention until shortly before draft time. I don't have specific advice for anyone, like take Aaron Hernandez in the 3rd round or don't take a QB until the 4th round, but there are a few concepts that my study of football statistics over the past few years can be applied to fantasy.
1. Fantasy football is overwhelmingly random, and you are not as smart as you think you are. You don't know anything special. You have no special power of analysis and prognostication. None of us do. Admit that now and you'll be doing yourself a favor. Everyone has access to the same information, the same rankings and the same expert predictions. If you're in a 10 team league, you start the day with a 10% chance of winning it. Injuries and unforeseeable developments will determine who wins your league, not your front office acumen. Even if you are God's gift to fantasy football, you have, what, a 15% chance of winning?
2. Ignore your own analysis and trust the crowd. Fantasy football is a great application for the wisdom of crowds. You have your dumb biases and errors, and someone else has theirs, and another guy has his. Average them all together, and as long as they're not correlated, the biases and errors will cancel out. If your fantasy service of choice shows you where players have been drafted so far, and the player hasn't suffered some cataclysmic injury in the last few days, trust the crowd over your own dumb judgment.
3. Ignore total predicted points and draft players on the basis of Value Over Next Available (VONA) at the same position. If you're in a 10 team league, your worst case scenario for a starting QB is the 10th-rated guy. He's your replacement player at that position. True player value is a function of his value over the replacement level at his position. If a TE is predicted to only get 100 total points this season and a RB is predicted to get 160, that doesn't necessarily mean you take the RB. If the replacement TE is predicted for 40 points, and the replacement RB is predicted for 130, you take the TE. He's worth 70, while the RB is only worth 30. This is why you take defenses in the later rounds even though they tend to score so many total points. Defenses, in general, are closely packed together in terms of predicted point totals. Note that you should make sure you're looking at point values based on your league's particular scoring rules and not a generic rankings.
4. You are the underdog, so you need a high variance strategy to maximize your chance of winning. Say you are pretty good at this stuff, and you have a 15% chance of winning in a 10 team league. (That's really, really good, by the way. It's a 50% improvement over the baseline expectation.) That means the field has an 85% chance of beating you. Just as underdog teams need a high variance strategy to beat a strong favorite, you need to take some chances in your league. Roll the dice. Pick up the players with question marks over their heads in the middle rounds, for example. A guy like Antonio Gates, because of injuries, might be predicted for 70 total points this year. But that's because he has an equally good chance of getting 20 points as 120 points , but not a high chance of getting anything in between. Bet on the 120. Don't bother picking up a guy you know is going to be league-average. Pick up a lower ranked guy with more uncertainty (more upside) instead. This strategy will optimize your chances of being first in your league, at the expense of being 4th through 7th. But it also increases your chances of being 10th.
5. Fantasy value and real-world value of players and their stats diverge tremendously, especially for RBs on good offenses and for average QBs with poor team defenses. RB values are predominantly driven by number of carries, red zone and goal line carries in particular. Trash time allows middling QBs to rack up large amounts of fantasy points. The fact that a half-yard goal line plunge is worth a full 6 points in most leagues warps most fans' perception of real football value.
6. Don't worry about bye weeks. OK, you lose one week but win the other 15. Managing your draft around bye weeks is a great way to end up with a solid 5th place team. You're going to need to gamble. See number 4.
7. Resist the endowment effect. You know that feeling about your team right after the draft? This team is awesome, you say to yourself. Sure it is...for last year. You fall in love. It's a natural instinct to overvalue what you already have. Don't fall in love. This is one reason there tends to be so few trades despite their obvious win-win nature in fantasy sports.
8. Resist other biases, like your favorite team or player, or your favorite team's rival.
9. Ignore preseason statistics. They don't correlate.
- Home
- Unlabelled
- So You've Got Your Fantasy Draft This Week
So You've Got Your Fantasy Draft This Week
By
Brian Burke
published on 8/27/2012
Subscribe to:
Post Comments (Atom)
I agree, especially with 3 and 4, but I like being biased towards my own team's and my favorite players because it makes it more enjoyable to root for them, and I'm willing to sacrifice the percentage point or two in odds to have the greater enjoyment. After all, I have to live with my team for the whole season, and by the time I may or may not win fantasy I'll be much more concerned with my real team making the playoffs (hopefully).
I believe that your 3rd point is the most important. It's incredibly simple - but it's amazing how many people simply don't get this.
As for "trusting the crowd", I agree that it is largely pointless to try to predict a player's performance/stats. You might as well just go with the crowd projections. However, it's NOT pointless to try to predict a player's *availability*. If there is any "skill" in fantasy football, it lies not in prognostication, but in diligently reading camp/team/practice reports and knowing (before your competing owners do) that a given player will be gaining/losing playing time because he is gaining/losing favor with the coaches.
For example, when you read that simple one-liner stuffed at the end of a long team report that says Star Player X received NO snaps in the Thursday practice - even though he's completely healthy - that's a huge danger sign that many other FF owners don't pick up on.
In every draft there are usually 10-15 players that are projected for big stats that I have completely wiped off my board because there are tons of subtle signs indicating that the player's role will be diminished in the offense this year. Conversely, there are 10-15 players that I move up the boards because it is apparent that they will simply receive a lot more playing time than most people are giving them credit for.
RE: Value Over Next Available
IMO The replacement value player should not be the worst STARTING player. It should be ~ the value of the best available player in free agency. If there are 10 teams and in aggregate they hold 18 QBs, then the replacement level player should be the 19th best QB.
Earlier I said the replacement player should be APPROXIMATELY the value of the best available player on waivers because I don't actually choose that player. I rank all players by position and choose the (X+1)th best one as my replacement player, regardless of whether that player is held or not. In this system, some players in free agency have replacement value > 0.
Example:
A team has 2 players:
QB, 10 pts/g
RB, 5 pts/g
QB worst starter: 7 pts/g
QB best FA: 3 pts/g
RB worst starter: 5 pts/g
RB best FA: 4 pts/g
Our team accepts a trade, trading our QB for a RB that is expected to average 9 pts/g. By the worst starter method of measuring replacement value, we read the trade as a net gain of +1. However when we pick up the best available FA QB to fill out our roster, our team is now:
QB, 3 pts/g
RB, 9 pts/g
That's 3 points worse. If we had valued the players using the best available FA, we would have seen that the trade was -2 in value (we still have our old RB on the bench, which is where the other 1 value went).
great stuff. where do you go for player predictions and rankings?
Chuck - Check out www.numberfire.com (we use advanced stats to project fantasy players/rankings, etc...)
I have some concerns about the "risky strategy" (No. 4). Certainly, at 15% you are an underdog to win the league vs the field. If you have a 15% to win, however, that means you have a significant advantage over each other individual owner. That being said, each week is an individual matchup in a head-to-head league. As the favorite, you need consistency (low risk, low variability).
In general, it is extremely important to employ risk analysis for each matchup (if you're the underdog, you need higher variability players and vice versa), but in the case that you actually are better/more informed than the rest of your league (i.e. you have that marginal "moneyball" edge), it's more important to have a consistent team from week-to-week and increases your win probability.
I think it's important to recognize potential biases in the crowd. The crowd, as a group, will tend to forget about things like strength of schedule and regression towards the mean.
Point 1 is why I no longer play fantasy football. It takes too much time for so little skill involved, and even a good analysis can leave you coming up empty. I had a team that was utter garbage due to a huge number of early season injuries, to the point where I didn't have a TE or enough RB for one game. I picked up two unknowns out of desperation and became the team to beat--those two unknowns scored 6 touchdowns between them in their first game for me. Both flamed out the next year, but by then, I'd decided I'd rather play something with some skill involved.
Nonetheless, if you're still going to play, the remaining points are good to keep in mind. I NEVER had a "great" draft, but I usually did fine anyway. Besides, most people I was around wanted week-to-week consistent scorers in a game with high scoring variance, especially when you throw in bye weeks. A Manning won't get "his" "consistent week-to-week points" when his opponent is that week, and you're not getting two guys that caliber anyway. So, my poor drafts usually didn't hurt me very much.
So don't worry about it too much. Just play.
Sorry, the editor dropped a word: :...when his opponent is BYE that week,..."
The question I always have when I see these lists is, how do you get Points over average from Wisdom of the crowd.
ADP only gives you ranks, but no points and averaging multiple projection systems can get quite expensive.
Also when I dont see where the lots of win win trades come from. Maybe I am to conservative, but if I manage to obtain 3 good RBs, I rather prefer keeping them as security instead of trading away for someone else backup QB.
In most fantasy leagues you don't want a high variance strategy - the single-elimination playoffs add that variance for you.
In a 10-team league where 4 teams make the playoffs, an average team has a 40% chance of making the playoffs and a 25% chance of winning the title if they make the playoffs. If you're good, you have a lot more influence over that 40% than that 25%. The best you can do is maybe a 70% chance of making the playoffs and a 30% chance of winning the title conditional on making the playoffs (for about a 20% chance of winning the title). So you actually don't want to be taking too many risks (at least not risks just for the sake of adding variance).
"Total points leagues", which don't have playoffs but instead just decide the champion by adding up your total points for the season, are the leagues where you want high variance. But they're relatively rare.
On #2 - when you follow the crowd, do not just go by the default rankings on the site that you're drafting on, because that is what all the other owners are paying attention to. If you're in one Yahoo league and one ESPN league, it's better to use the default ESPN rankings when drafting in your Yahoo league and the default Yahoo rankings when drafting in your ESPN league. That way your Yahoo team will be really good according to ESPN's rankings, and your ESPN team will be really good according to Yahoo's rankings; if you just followed the default rankings on the same site then both teams would come out pretty average.
Or, even better, look at the actual average of experts at Fantasypros.com.
Does anyone know where I can find a list of the best players in Value over replacement fantasy? Or one projection this year that shows it?
This is the original sentiment on poker.
I've never heard a good argument for drafting RB "handcuffs." To me that says you like the offensive line and offense more than the actual RB. Marshawn Lynch is worthy of a 2nd round pick probably, but unless you think he has little skill and his success is a result of his O line and offense as a whole, why draft the rookie RB as a handcuff? Let's say Lynch gets hurt for a few weeks. Is it just psychological- meaning you don't want to miss out on those points that Lynch would have gotten if he didn't get hurt?
Forget the fact that a lot of people think Robert Turbin will be good. He's still a rookie and a backup.
I find it hard to believe that your average fantasy football player cares about the offensive line more than the RB. So what is the argument for the handcuff? Or what else am I missing?
The handcuff is a classic low variance strategy. You're assured at least one starter at the cost of possibly 2.
Caveat: I've never played FF.
Re: handcuffs, I've always thought that the idea was that the RB (since that's who I've always heard about with handcuffs) on team X should run for about Y yards and have T TD's. If RB A gets hurt, his backup should get those carries, yds, and TD's. On the few teams that still employ mainly a primary RB (think MIN or JAX), this means (theoretically) that the backup gets those fantasy points. But my personal question has always been--Why are you drafting a player who has a history of injuries???? (think Darren McFadden this year)
Rule number 2 is both inherently silly and directly contradictory to rule number 4.
The crowd is not smart and biases do not cancel out. Moreover, since you are the underdog, you need to outperform the crowd. You don't do that by picking what the crowd does.
2 and 4 don't need to conflict. Most projections systems and rankings do not do rule 4 well because they do not give you the variance number, they just give you the middle stats. So Brian isn't saying don't do rule 4, he's saying don't think you know which rule 4 players are the one's that will win their variation on your own.
I think what Brian said about replacement level is actually the right way you just have to realize that when the top 9 QBs are gone you need to take that 10th one now. You can't simply let the other teams take QB 10-15 and then you throw #16 out there every Sunday. Er, I mean Thursday, er Saturday.... Replacement over the last starter is the right way you just need to buy that last starter for whatever the replacement level price is, which is not Round 15 or $1, it is higher then that.
"The handcuff is a classic low variance strategy." True but that IS a rule conflict. You also noted that trades are hard to pull off. Knowing that, what good is your non-handcuff backup who becomes a starter if you can't trade that extra starter you now have? Sometimes its better to have the handcuff if the handcuff is truly good due to ability or system.
A backup RB is like a lottery ticket that pays off if the starter gets injured. Felix Jones is probably a low-end RB3, which is worthless in most fantasy leagues, unless DeMarco Murray gets hurt in which case he suddenly becomes a solid RB2. If you also own Murray, then Felix Jones becomes insurance - a lottery ticket that pays off precisely when you need it to.
If you don't own Murray, then in order for Felix Jones to be startable you need a Murray injury (to give Jones RB2 value) plus you also need one of your starting fantasy RBs to struggle (to create a slot in your lineup where you can start him). If you own Murray as well as Jones, then all you need for Jones to be startable is a Murray injury, since that will both increase Jones's value and create a hole in your lineup.
That's why it's better to handcuff than to just draft some random backup as a lottery ticket. Drafting Matt Forte & Felix Jones (or DeMarco Murray & Michael Bush) is like taking out insurance on your neighbor's car.
Good post. I agree in drafting "upside". Every year it is those mid round picks or waiver wire pickups that usually win you a championship, cruz,nelson,stafford,jamal charles,fred jackson etc...
While it is true that fantasy sports is highly random and essentially gambling, using a VONA strategy to draft takes all the fun out of it which is the actual point of playing the game. You lose a sense of gratification when successful if you have used an arbitrary opportunity cost system based on someone else's predictions to draft your team. I think it is very possible for someone to use their own intuition and reasoning to give themselves and higher chance of winning the league relative to others.
That said, I think the best advice in the column is to roll the dice and take the high risk high reward players. Just like in Black Jack, money come in, money comes out. No one gets rich sitting at the $5 tables. The Darren McFaddens and Cam Newtons will carry teams to victory.
Re: Joseph and Darren McFadden - when McFadden's been on the field he's been one of the best fantasy backs in the game. You draft "injury-prone" players like him if (1) you think injuries are random and not a skill, and/or (2) go for the high variance strategy and hope he stays healthy.
Talk about random, I have had the consensus best team in my league for 5 years running (it is a auction/dynasty league where players pay rises each year if you keep them). I am god with math so I just kill the other players on salary allocation. This year I have Rodgers, 3 of the top 10RBs and Andre Johnson, other years it has been similar.
Yet each year I have some bad luck, either with injuries (Brady ACL year), or with losing in the playoffs. Doesn't bother me since I do it for the entertainment, but I am well acquainted with the vicissitudes of lady luck.
I am a big believer in the "crowd effect" that you suggested for any very liquid market, and it seems at first glance that fantasy player values would fall into this category. however, you neglect the fact that an overwhelming majority of fantasy football players obtain their biases from a single source. There are many ranking sites and magazines out there, and many various sites from which to run a draft, but in general, yahoo is the gold standard. Few fantasy football players are unaware of the yahoo rankings (which in my opinion tend to be fairly good, but not great)--and while they still inject their biases from that rubric, the biases deviate from the same starting point.
It stands to reason, then, that if you are a better handicapper, or have access to someone who is better than the analysts at yahoo, you will have an advantage.
In a draft there is still incredibly high variance, and I agree that the very best will probably win 15% in a field of 10. Draft order, injuries, and waiver priority in key weeks are huge factors. However, in salary cap leagues, many of these sources of variance are mitigated. In the last 4 years on the yahoo cap league my percentiles were 99, 99, 98, and 88. Facebook cap leagues 99% and 99% the previous two years. In more serious competitions for cap leagues with buyins ranging from $100 to $250 I have been outside of the top 80% only twice in 14 tries (and never in the bottom half), with 2 firsts out of 400+. I believe that there is less luck in these than you might think....or perhaps I am running INCREDIBLY hot.
I wouldn't recommend numberfire. It hasn't cracked the top 15 in either of the past 2 years: http://www.fantasypros.com/nfl/accuracy/
" I believe that there is less luck in these than you might think....or perhaps I am running INCREDIBLY hot. "
More likely, you're just in a weak league...
"The handcuff is a classic low variance strategy. You're assured at least one starter at the cost of possibly 2."
You only handcuff top RBs. The reason it works is that you can be assured that your handcuff will have his very best games the weeks you most need him to (when your #1 is out). That is worth a ton, compared to most bench players, whose production is mostly unpredictable and not correlated to your team's need.