p r i n t     v i e w
fantasy football fantasy football

R o b o

fantasy football fantasy football


Fantasy Football Strategy Soup

By: Dr. RoboStein

Every year I hear many great new fantasy strategy plays. However, every year I also hear many recycled strategy tips that I wish would just go away. I'm confident you've seen the tired maxims listed below because they pop-up again and again like a broken whack-a-mole game. Read on or just click the maxim of your choice for a good debunking.

Take the Robocoach Challenge! If you are a professional fantasy football journalist or representative of a website, publication, application then take the RoboCoach challenge.

False, Misleading, Meaningless Fluff and or Novice Fantasy Football Maxims

Study The Stats Maxim

"fantasy football isn't some little hobby you have with drinking buddies; it's a job," Editor, National Fantasy Football Magazine

Every year I hear at least one expert say that fantasy owners should "study the stats" and "do their homework" and "blah, blah, blah" as if fantasy players didn't have anything better to do. I've seen the marketing research--these people have jobs, mortgages, families and I'm sure they would like to dive into a box score, but they just can't. Telling them to do so only reinforces our industry's geek perception. When the average fantasy player hears such advice they must think something like, "what a bunch of geeks...yeah, I'll study the stats, just as soon as I finish watching the Star Trek marathon on TV Land...not!"

What the average fantasy player needs is a robot to do all the analysis and suggest strategic plays (see Robocoach). Bottom line: the average fantasy player probably doesn't care if they do it themselves--so long as they can act like they did. They want to pick up the winner's check, get their tiny trophy and bust their buddys' butts for not being as smart as they are. If they can do all that and keep their job at the same time then they're happy. Stats? No thank you.

RBs First Maxim

"Get your starting RBs before looking elsewhere," John Doe, National Fantasy Football Magazine

This maxim is really long in the tooth. It works for novices to keep them from doing really stupid things, but fundamentally it is not true. There are so many exceptions that one can easily see there is something else (beside being an RB) that drives strategic value. For example, in 2008 Peyton, Brady and Moss were all non-RB first rounders and the second round had non-RBs' Reggie Wayne and Terrell Owens. What does drive value is whether the player is a difference maker.

Difference makers are players who are in a class of their own. True difference makers have large point differentials from their next closest peer. On RoboCheat these players will be in a thin tier (by themselves or with one or two other players) and the drop to the next tier is very long. For example, few predicted this, but in 2007 Tom Brady and Randy Moss were difference makers, they were so much better than the next best at their position that any fantasy team who had one of them was virtually guaranteed a spot in the playoffs.

Difference makers create positive point differentials versus your opponent's player at the same position. In other words they score more--a lot more. Obeying the RBs first maxim will make you pass over some great difference makers at QB, WR, TE and sometimes even DF.

RoboCheat will steer you right to the difference makers. For example, using RoboCheat I ended up with RB-WR-WR on my first three picks in 2004. I didn't take a RB in round two or three in large part because there were WRs who could create a bigger difference versus the rest of the available players, be they RB, QB, TE or whatever. In the second round all before me took RBs (they stuck to the RBs first maxim). I got the first WR: Torry Holt, a.k.a., Big-Play Torry Holt. By consulting RoboCheat I saw that Torry was a difference maker, i.e., after Torry came a huge ~50 point WR tier drop. By taking Torry I forced my opponents into a lower tier thus creating a, minimum, 50 point/ 14 weeks =~ 4 point per week advantage!

There was an opportunity cost, however, because taking Torry forced me to draft a lower tier player at RB (in round four), but the cost (tier delta) was a a mere 20 points (a loss of only 1.5 points per week). Thus, the net gain was 4 points - 1.5 points = 2.5 points per week and that's nice, but there is something more important going on here. It's a bit dense, but it's worth the read.

How much do you think the excellent tier delta management described above improved my overall chances of winning? A 2.5 point net per week ain't much - even a fantasy football novice knows that. I'd say, in my twelve team expert league my chances of winning improved from one in twelve to two in twenty-three. In other words, I barely moved the odds if at all. What if I told you there was a way for you to improve the odds to one in two? It can be done!

My apologies to the ladies reading this, but think of a beauty pageant to help understand the bigger strategy play going on. Suppose I suggest your state elect a blonde as your next Miss (your State here). Who can argue with that? Nobody, because in general men prefer blondes. However, to understand what beauty is all about, imagine a Miss America pagent where every state elects a blonde to represent them. The winner will certainly be a blonde because there wasn't any non-blonde competition.

Likewise, the winner in your fantasy football league last year probably selected two RBs with his first two picks. I know this because, like the blonde strategy, most fantasy players obey the RBs first maxim and so, statistically speaking, the winner in your league probably implemented the RBs first strategy. People copy success and thus the strategy self-perpetuates (despite my protests).

However, if you don't obey the RBs first maxim (and most of your opponents do) then you can gain major strategic advantages; advantages that make tier deltas seem penny ante. How? Returning to the beauty contest analogy, if Ohio elects a red head and all the other states elect blondes then you can bet (other things being equal) that Miss Ohio will finish either first or dead last! Why? Because the judges scores will be equal for everything except hair color. Miss Ohio separated herself from the pack and thus increased her odds from one in fifty to one in two--that is a major increase in odds and that is what schema shifting can do (more on 'schemas' in a moment).

Likewise, if your competition myopically pursues RBs then they will knock each other out like 49 blondes in a beauty pageant. You should ignore the maxim and look for red-heads, i.e., create competitive differences.

My 2004 draft was a good example of competitive differentials in practice, but before I can describe my 2004 experience I need to give you a little background: As you might have noticed, the NFL exhibits a boom-bust cycle when it comes to passing or running. Some years passing is fashionable and vice-versa for run. Moreover, for reasons unknown to me, some weeks passing is fashionable and vice-versa for run. One week it will be all passing and the QBs and WRs get the points and the next week it is back to the run and the RBs get all the scoring.

I haven't done the number crunching yet, but I'm confident the cycle exists. When I do the crunching I'll be able to predict the cycle, but, to take advantage of it, you don't need to predict it you just need to know it exists. Even if you doubt the existence of the boom-bust cycle, for instruction purposes, assume it exists.

One other background item, in strato-speak your draftee player positions are your strategic 'schema'. Schemas help illustrate built-in advantages that you may (or may not) have. My (simplified) 2004 schema was RB-WR-WR as you can see below (draft round in subscript):
  • RB1-WR2-WR3 = my schema
  • RB1-RB2-WR3 = opponent #1
  • RB1-RB2-QB3 = opponent #2
  • RB1-RB2-TE3 = opponent #3
For simplifcation purposes I'm showing only the top three draft picks. To find competitive differences you must compare schema and cancel out like positions. This assumes that all players in the same position from the same round are of equal value. That isn't true but it helps us see built-in competitive differentials that a particular schema may have. For example, when I play opponent #1 we cancel out like so:
  • RB-WR-WR
  • RB-RB-WR
We can conclude that I will win on weeks where my round two WR outperforms his round two RB. I like that matchup better than if I had just taken a RB in round two (like everybody else). However, at this point you can't say I have a competitive advantage--just a competitive difference. What I've done so far in this draft is what I call a 'schema shift.' In other words, I saw a dominant schema trend and I shifted from the pattern to differentiate my team from the rest of the league. Differences such at this, as I've alluded to, are usually decisive differences.

The earlier a schema shift takes place, the more decisive a factor it will be. The reason for this, mathematically speaking, is tier drop-offs, which I sometimes call 'tier deltas'. As the draft progresses tier deltas shrink and schema shifts lose thier power.

Some of you are probably saying, "hey, you took a RB in round one, why didn't you schema shift?!" In 2004, I would have liked to have built-in a competitive difference (a schema shift) in round one, but there was just one RB left, Deuce McAlister, before a major, ~100 point RB tier delta. That was a huge drop-off and I couldn't risk it. Note that when drafting my main focus is on my opponents' schemas (relative to my own), but a 100 point delta cannot be ignored. I wanted to create a competitive difference in round one, but the opportunity cost was too high and I had to bide my time. The longer I waited, however, the less effective my shift would be.

Back to schema comparison (above), if we take into account the boom-bust cycle we can also say that I will likely win on any passing boom year or week. I really like the sound of that, it means, without regard to who I picked at WR in round two, be he great or mediocre, all things being equal I should still win half my matchups thanks to the boom-bust cycle giving me a built-in advantage in passing years (or weeks).

Back to the pageant analogy. Imagine each judge has a scorecard with just two factors on it: height and hair color. If all the contestants are of about equal height then hair color becomes the critical factor in determining the winner. In fantasy football height is analogous to minimizing tier deltas--subjecting your team to big deltas makes you shorter. Hair color is the boom-bust cycle status and if you reasonably manage your deltas then boom-bust becomes the critical factor.

In 2004, drafting Torry was my schema shift and thus I built-in a competitive difference (relative to the dominant schema). All my management of tier deltas just ensured that I would be OK if fate favored neither run nor pass. In other words, my big strategy play was the schema shift because it separated me from the pack, i.e., it put me in a position I where I might lose big or win big. Like Miss Ohio, I had dramatically increased my chances of winning from one in twelve to, perhaps, one in two.

In 2004, the boom-bust cycle tilted to pass and that produced some nice 'upset' (as my opponents called it) wins for me. I beat RB-RB teams regularly and won my league. I managed my deltas by not taking exceedingly large risks (like the 100 point tier drop) and then the schema shift did just what it was supposed to do--it became the deciding factor.

Bye Week Maxim

"Be aware of the bye weeks! You don't want your first and second string QBs to have the same bye week," Sr. Editor, National Fantasy League Football Magazine

This is another fantasy football maxim that drives me nuts. I call it the Bye Week maxim. "Experts" all want you to make sure your starters and backups don't have the same bye week.

@#!$ the bye week! In case you didn't hear that, @#!$ THE BYE WEEK! If all your players have the same bye week that is a GOOD thing. That means you will go 13-1 instead of 7-7. Nothing makes me more upset as when I am missing ONE starter every week for eight weeks especially when I lose a game by one point and wish my starting QB/RB/WR... hadn't been on bye.

I will not spend one minute trying to draft players according to their bye week. There isn't enough time in the world and it doesn't matter. What does matter is that you get a good draft pick when it is your turn. Bye week doesn't matter.

Only Handcuff RBs

Its pretty easy to think of exceptions that prove this is not really true. I could go on for pages, but here are three:
  • In '10, Brandon Marshall left Denver with a big hole to fill. Eddie Royal is the heir apparent, but everybody knows that Eddie is no Brandon Marshall. So, what to do? Read below and do what I did with Roy Williams in '09.
  • In '09 Terrell Owens left Dallas. Roy Williams was the heir apparent, but everybody knew that Roy was no Terrell Owens. We knew Romo was going to get his yards, but Roy? Hmmm. The answer was to handcuff Roy. Miles Austin was the handcuff choice in about week three or four when he exploded with a 200 yard game and if you didn't have the handcuff then...too late!
  • I had Rich Gannon in '04 and you bet your life I had Kerry Collins as his backup. I can't tell you how happy I was that I had Collins when Gannon had his head removed without the benefit of anesthesia by Derek Brooks (Tampa Bay LB) in week 3. If I obeyed the Only Handcuff RBs maxim I would probably have drafted or picked up an incredibly mediocre player, like Kyle Boller, as my new starting QB. I wouldn't have been able to get Collins off waivers--he was highly sought after that week and I was not first on the waiver wire.
Some say this will expose you to problems on the bye week. However, when Oakland's bye week came I could easily pick up a starting QB for the week (starters emerge EVERY week and some starters are available on waivers all season)? Respectable starters emerge every week in the NFL due to injuries and benchings. If you're worried you won't get someone to fill the bye week, then pick a waiver a week in advance to fill in for the bye week.

I admit that typically you don't need to handcuff your QB like I did in the example above. Handcuffing usually only applies to RBs. Fundamentally speaking, however, handcuffing has nothing to do with position. What drives your decision to handcuff? Math of course! You just need to do a simple risk-reward payback calculation:

To Handcuff or not to Handcuff= (chance you'll be rewarded X chance you'll pick the right handcuff X reward of handcuffing) less the (opportunity cost of handcuffing). A positive result means you need to handcuff.

Let me use a non-RB example to show why you needed to handcuff a QB in 2008's draft. Matt Leinart was supposed to be the Arizona Cardinals starting QB for 2008, but everybody knew he had a shaky hold on the starting position. Kurt Warner was the backup. Everybody also knew that the Arizona QB, whoever it may be, would score a ton of points.

If Leinart lost the job and you only drafted Leinart then you would have to take whoever was available on the waiver wire. In my league, when Leinart lost the job, the best available waiver wire QB was Trent Edwards and that was a big drop. I don't know the exact difference between Edwards and Warner's 2008 output, but I'll be generous to Edwards and call it (the reward for handcuffing) 5 points a week.

Other QBs emerged that were better than Edwards, but at the time nobody knew that Flacco and Orton were going to be very good and it took about one month before anybody was willing to call them 'real'. In other words, it is still a 5 point per week reward for holding onto Warner as a handcuff. Flacco and Orton, if you got them at all, would have languished on your bench while you figured out who to start/sit. By the time you started Orton he would have gone down with a season ending injury and so back to the waiver wire for you.

You could pick up Warner during the 2008 draft at about round six so the opportunity cost associated with doing that is equal to the maximum Round 6 vs. Round 7 tier delta. If you look at the RoboCheat tier deltas you'll see that number is around 20 points/14 games = ~1.5 point per week. I use fourteen games as the time horizon, but you can change that to 16 games if you want to assume you'll make the playoffs or if you are in mid-season then reduce it to whatever number of weeks are left in the season. That last point is important to remember, i.e., handcuffing goes beyond draft day and as things change you may need handcuff a starter you hadn't handcuffed before.

The chance you'll be rewarded is the key to understanding why top RBs are typically handcuffed. RBs almost never start all 16 games and so the chance you'll be rewarded is near 100 percent. The point rewards for handcuffing in the RB case are also quite high because the waiver wire is usually completely absent starting RBs and you're stuck with the likes of third-down backs and fullbacks. You can usually get your stud RB's handcuff in the last round of the draft and so the opportunity cost is virtually zero.

Back to my 2008 QB handcuff example, the chance that Leinart was going to lose his starting role was quite high. Most experts I spoke with put it at about 90 percent. Let's be generous and call it a mere 50 percent chance of losing his starting job.

The chance you'll be rewarded is also influenced by how confident you are that you can pick the right backup. In the Leinart case you could say with 100 percent confidence that Warner was the correct handcuff. Given that, the chance you'll be rewarded remains the same, i.e., 50% * 100% = 50 percent. With RBs the confidence factor usually drops below 100 percent--I call this the Shanahan factor. Whenever the starting RB went down in Denver you could never be sure who Shanahan would put in. Sometimes he would completely shock everybody and stick in a fullback. Grrrr! When this is the case, your confidence might drop to 50 percent or lower.

That is the last number we need for our equation and so plugging in the numbers we see:

To Handcuff or not to Handcuff= (.9 X 1.00 X 5) - (1.5) = ~3.5 points per week payback. The payback is positive and so in 2008 you should have handcuffed Leinart to Warner.

When the Shanihan Factor is near or below 50 percent you sometimes should get all the backups to your franchise player. Doing this may require three or even four roster spots, but you might need to do it to control the Shanihan Factor and get a positive payback.

In 2007, I had Brandon Jacobs and that meant I needed to fill three roster positions in order to hold onto the NY Giants tailback position: one for the Jacobs and two for the possible backups (Ahmad Bradshaw, Reuben Droughns and Derrick Ward). As these players got injured I just swapped the appropriate players from waivers into the two roster spots I designated for the backup to my franchise player (Brandon Jacobs). During two weeks of the season I had a whopping four roster slots occupied with NY Giant RBs, but it was worth it--these guys were money. If the payback calculation ever flips to negative, however, you should change your strategy.

How does it flip? You might have already noticed a missing opportunity cost to consider. As you begin to fill more and more roster spots with handcuffs you are passing up on upside players. Upside players, unlike forgone mid-round draft picks, have a risk factor associated with them. In other words, they couldn't start for you right away, but they might become starters. The same can be said of your handcuff players and so from that perspective upsided players and handcuffs are the same thing. If you pick up the upside waiver player, releasing a handcuff, you will increase the Shanihan factor and reduce your payback. If the payback of the upside player exceeds the payback you gave up then do it. It's a balancing act, but you can see that it is about math and payback calculations--not a player's RBness.

There is another mid-season scenario to consider: if you are first or second on the waiver wire there might be a player that is not an upsider, maybe a player that somebody forgot to handcuff that has now has become a clear cut fantasy starter. In a situation like this you might need to grab that player, drop the extra handcuff and allow the Shanihan factor to jump. Essentially, this becomes like the mid-round draft opportunity cost because the waiver wire player isn't an upsider--he's a clear cut fantasy starter with no associated upsider risk.

With the 2007 Giant tailbacks the payback never flipped. More roster slots meant increased opportunity cost for handcuffing Jacobs, but the opporuntity cost never got so high as to flip the payback into negative territory. I kept winning (in large part due to the NY tailbacks) and I was usually at the end of the waiver order where upside was low. If however, I found a player (or combination of players) on waivers that, factoring in the upside risk, had a payback exceeding my handcuff payback then I would have dumped the handcuffs.

DF and K Last Maxim

"Never, ever, draft your kicker and team defenses before the last round," Sr. Editor, National Fantasy League Football Magazine

This maxim is just the flip-side of the RBs First maxim and it is just as strategically vacuous. This year there are several defenses that will be drafted in the middle rounds and RIGHTFULLY SO. According to the WEIGHTED projections in my league two defenses will be top 10 in scoring at the end of the year. TOP 10! And this happens every year in almost all leagues.

Does this mean you should draft the DFs in the first round? No, you can get DFs in the later rounds because we silly humans go after DFs then, not because they are of lower value. Some smart players will schema shift and draft their DF in the middle rounds and get great value while others are wasting picks on upside players.

I was so happy that I drafted NE's DF in '04 in the middle rounds--on several occasions NE dropped 20 points on my opponent. And not that I am saying to draft your DF in the middle rounds always, but if there is a DF out there that will outperform the available talent and won't be around for your next pick--THEN GRAB THEM! Better yet, just get Robocoach's customized cheatsheet and find out who to take and exactly when to take them, i.e., what round and what pick in that round.

Trade Overachievers Maxim

"If you're 5-0, trade your overachievers for proven stars on the brink of breaking out," Writer, National Fantasy Football Magazine

Don't do this--I've done the analysis and found that this is incorrect. In the physical sciences they say objects in motion tend to stay in motion and this is true for NFL football players. In the NFL, players tend to show what they will be all about in the first few weeks of the season (or the first few starts if the player comes in mid-season). If a player 'overachieves' during the start-up then, per my analysis, generally speaking, he'll continue to overachieve.

It might seem backwards to you, i.e., you might say something like, "but this player came out of nowhere. He isn't proven and so-and-so is." It's not backwards, it's forwards, i.e., more often than not what you are seeing is the beginning of a new trend. You have to play the odds which I've found favor it being a new trend. RoboCheat incorporates this principle and thus you can be confident that RoboCheat will tell you what player is more valuable.

Additionally, don't gamble when you trade and only trade when you can find a win-win deal, i.e., a deal that will improve the output of your team and the other team. If you really want to do a deal, and you should, then just have RoboTrade do a trade analysis for you.

Robo will find you a win-win deal that generates value for you and your trade partner. Remember, even if you don't use RoboTrade, win-win is the only way to go. People won't trade with you again if you do much better in a trade (even if they come out better too).

Winning Illusion

"We've won X experts contests," Owner and Operator, Major Fantasy Football Website

Experts who are fond of talking about all the "experts" leagues they have won should be approached with caution. I call this the Winning Illusion because players will always tell you about the time they won, or the great stock they picked, but they never tell you about their losers--and "experts" have many losers. The true quality test is not if you win, but your winning percentage versus quality competition.

Robocoach has competed thrice in the "experts" leagues and he has a 33% winning percentage, i.e., we won the championship 1 our of 3 years. Not bad, but still not enough data points so you won't catch us blabbing about our winning percentage being high.

The Robocoach Challenge

If you are a professional fantasy football journalist or representative of a website, publication or application then RoboCoach wants to play you! Robocoach will manage 11 teams and team 12 will be you, the expert. You can change the scoring setup and league setup too--Robo isn't tied to any particular scoring system or league setup.

RoboCoach is an open API XML web service and will integrate easily with any league management provider who would like to sponsor the contest. It will be like Gary Kasparov versus Deep Blue and should generate a lot of publicity for the sponsors, you (the expert) and Robo.

Robo will play as many experts as want to compete, so invite your friends! Robo simultaneously strategizes for approximately 200,000 teams in 20,000 leagues, so multiple leagues are no problem for Robo. BTW, I don't even think Deep Blue could do that, i.e., Deep Blue was designed for one game at a time.

3 RB Maxim

"Have three RBs on your roster after the first three rounds...grab a TE and/or a QB [by round 5] and then start drafting heavily for receivers [after round 5 or 6]," Owner and Operator, Major Fantasy Football Website.

This is like the RBs First schlock. However, it takes it a step further by saying not only should RBs be first, but three of your first four picks should be RBs. Woe to the man who heeds this advice because by the time the fourth round ends you would have passed up on so many WRs and QBs of higher weighted value that your chances of winning are nil. In the 2005 draft you would have passed on Peyton, Chad Johnson, Torry Holt and many more.

By using your first three picks (or three out of four) on RBs you have just guaranteed that at least one of your top picks will ride pine. Even if you successfully trade away your extra RB to get production at say WR you will have probably let him sit for the first few weeks, maybe longer, to figure out if he was any good.

This is really bad advice because you are assuming that one of your RBs picks won't pan out (this advice is targeted at two RB starter setup). One of your RBs won't pan out--that part is correct and you have to hedge against that risk. However, the 3 RB hedge is the wrong way to mitigate that risk!

The appropriate hedge is to take the immediate backup of your 1st and 2nd round picks and you can do that (usually) with your final two picks in the draft. I call this the Franchise Hedge, but most others just call it handcuffing. For example, my first round pick in 2004 was Deuce McAllister. My final round pick was Aaron Stecker and I can't tell you how happy I was when McAllister went down that I already had Stecker because I would never have got him on waivers, he was too hot!

Additionally, in the second and third round of the same draft, I took Holt and Horn, which is completely against the expert's "strategic" advice. I took the WRs because they were difference makers and had the highest weighted value per Robocoach's cheetsheet and RoboCheat also showed if I didn't get them then I wouldn't get them at all (and he was correct, of course). To be more precise, RoboCheat showed that I wouldn't get them (Holt and Horn) and what I would be a bigger tier drop than if I chose the RBs.

Robocoach was right because that trio (with the help of Stecker, my Franchise Hedge) beat my opponents about the head and shoulders all season long. I won my league championship mainly because of these first three picks (Robocoach's suggestion to pick up Antonio Gates in the fifth round didn't hurt either (note, this was his breakout year--Robo was one of the handful who predicted it!).

This maxim makes matters worse by saying you shouldn't take a WR until, at earliest, round 5. My earlier example of drafting Holt and Horn in rounds two and three poke a hole in this maxim, i.e., the maxim advises against selecting WRs in rounds 2 and 3, but I won anyway. You will miss out on many difference making WRs if you heed this maxim.

Remember, being a difference maker is what matters. RB numbers are generally higher than WRs, but you shouldn't base decisions up generalities. Why? Because strategically speaking, fantasy draft moves are like chess moves. You want to maximize your value with every move in chess, likewise you want to maximize your value with every fantasy football draft move. However, maximizing is much more complex than simply making a judgement that RB so-and-so will produce more fantasy points than WR such-and-such.

In chess, if your strategy play is a single move and you have the chance to take a knight or a rook then you take the rook. However if your strategy horizon increases to two moves, then the right move might be the knight. Why? Suppose in the two move scenario taking the knight puts you in position to take a bishop on move two. Knight plus bishop is greater than rook and thus changing your strategic horizon changed the move you should make.

Here is a fantasy example, in the second round of my 2004 draft I had the chance to take Torry Holt versus Corey Dillon. Dillon was the 'right' pick because he was in a higher tier (according to RoboCheat) and he wasn't going to be available in the third round. When extending the strategy play out to include both second and third rounds the numbers favored Torry. Dillon, as RoboCheat showed me, was in a wide tier and if I passed on Dillon then I could still get a player in his tier in the third round (Stephen Davis, Brian Westbrook and Travis Henry). However, Torry was in a narrow tier and RoboCheat showed me that I would incur a mongo two-tier-drop and end up with someone like Derrick Mason, yuck!

RoboCheat will show you the likely tier deltas, but a shameless plug for RoboDraft is appropriate here. When you're drinking a beer at the draft you probably don't want to be looking at RoboCheat and envisioning all the scenarios. RoboDraft will do all that for you and let you enjoy yourself. RoboDraft will also tell you how many scenarios he analyzed. I just visited RoboDraft to check how many scenarios there are in a typical first round decision: a little over 70 quadrillion (that is 70 with fifteen zeros) which is less than I thought it would be, but enough for me to conclude beer is better than math.

Note also that the human factor is key in calculating scenarios. For example, in my 2004 league the value of one defense, New England, exceeded the value of ALL those players I just named, but I didn't take New England in the 2nd or 3rd round. I had the luxury of waiting until later rounds because, despite the high value and delta, nobody goes after DFs until later rounds. I was risking a big tier delta if I didn't get New England, but if I waited patiently I could make big gains elsewhere. In summary, you should follow the dominant schema, but look for ways to outperform it and when the time is right make your move.

Relatedly, RoboDraft knows when the position runs will begin and RoboCheat also shows you when to draft player so as to grab difference makers before they drop off the board.

Remember, don't obey to the DF Last maxim that I debunked earlier. As I said above you should grab your DF one round before someone else grabs them and no sooner. The top DFs are a pile of gold, but nobody will jump on the pile because the other piles are being jumped on first. Whatever, just make sure you follow Robocheat's advice. Robo will get you on that pile of DF gold at the CORRECT TIME, which is NEVER in the last round.

Set Your Lineup

"Make sure you set your lineup on time," Owner and Operator, Major Fantasy Football Site

This is good advice, but do you really think players need to hear this? I think this quote is indicative of the level for which most fantasy strategy advice is geared. That's too bad because there is so much to talk about. Strategically speaking fantasy is more complicated than chess and telling fantasy players to put their lineup in is about the same as telling a chess player to protect their king. Its good advice, but if I'm a chess expert I wouldn't read another word of that article.

Even if you were a novice would you really want to take this simple, but beatable, advice versus just getting a robot like Robocoach to tell you what to do? If you want simplicity then we guarantee RoboCoach is simpler and will make you look more like an expert than will KISS strategic advice. Robo knows how to manage your team after draft day.

Group Think

"If you get the first pick take Larry Johnson," Every analyst, every site and every magazine, August 2006

I call this Group Think because fantasy analysts, myself included, are an incestuous bunch. We know who the other analysts are and we will modify our picks if we like someone's analysis. We're copy-cats. Yes, and we get paid for it (it's shameful ;-)).

You shouldn't pay attention to us: we have no integrity. Yes it's true, despite our best intentions we will all converge upon the same solution because we will read each other articles and debate each other at conferences like the FSTA Fantasy Conference. Eventually all this analysis sharing results in us coming to the same conclusions. Thus everyone, and I mean everyone, picked Larry Johnson as the top draft pick in 2006. And this is not to say that is incorrect--LJ had a career year in 2006. In some (but not most) leagues Robocoach's Cheatsheet had Johnson as the top pick, but my point is that Robo didn't go and talk to some other analyst to figure it out and he doesn't modify his picks because of Group Think.

BTW, perhaps I should explain that I am not Robocoach. Robocoach is an algorithm and I frequently disagree with Robo's picks. I am sometimes right, but more often wrong. I can change Robo's algorithm and thus I make him susceptible to group think, but as I explain below, there is a very high hurdle to clear before Robo's algorithms change.

Larry Johnson is a good example of Robo's integrity and my lack thereof. In 2007, after his career year, group think pushed LJ back near the top draft slot. He was usually the third or fourth pick suggestion. Robo, however, had LJ at the back end of the first round and, in many leagues, well into the second round!

"Uh-oh," I thought, "there must be a bug." I spent a lot of time that preseason trying to get Robo to fall in line. I figured there must be some sort of problem with the 2006 statistical data that Robo bases part of his projections on. However, it wasn't the stats, I found that several of Robo's downgrade factors kicked in on LJ. For example, among other downgrades, the 'number of games' downgrade applied to LJ (games played is inversely related to the following year's production). Number-of-games is a principle that I've analyzed myself and have scientically proven to a 95% confidence level.

The bottom line is that I was falling victim to group think and I was even considering reprogramming Robo because of it. Had it not been for a set of scientifically proven, rock-solid principles commited to code I would have fallen victim to group think.

I'd guess that >95% of analysts do not have rock solid principles commited to code and thus group think will sway them. There were a few scattered analysts (the 5%) who downgraded LJ to the second round that year, but they, like RoboCoach, were going against the wind of convention--anyone who dared put LJ in the second round was loudly mocked.

LJ ended up handing the keys to Kolby Smith in 2007 around mid-season because of injury. LJ is now the poster-boy for number-of-games downgrading. Even if he had a great year, he would have been the exception.

Not to belabor this point, but the 2007 Superbowl was another classic example (in my opinion) of group think. If you picked the NY Giants to win you were roundly mocked. RoboTout didn't pick them to win, but rather, to lose by 1 point (btw, you can see RoboTout's predictions on the main menu). Vegas had a 10 point spread and so, once again, I checked all the code to see if I could get Robo in line and once again I found no algorithm flaws and no bad data.

On Superbowl Sunday I was doing a home improvement project and so listened to the radio all day and the NY Giant backers (celebrity guests and callers) were immediately branded fools by ESPN radio analysts. You had to have a lot of guts to stand by your NY Giants prediction because all the media types had, apparently, prepped for New England and sang their praises all day.

The Giants won and of course none of these media types burned their journalism degrees. Nor should they because they did their job. I've said this before, but it's worth repeating: forecasting is about math and journalists are bad at math (but good at talking, writing, etc.). I've never seen a scientific study proving this, but I'd lay odds it's true. Bottom line, journalist enable group think and that moves the Vegas line. However, you shouldn't listen to journalists for anything other than entertainment.

It will never happen, but it would be a lot better if analysts didn't share their analysis. Some might think all this sharing is a good thing, but it results in everybody being right or everybody being wrong. If everybody had integrity and stuck with their original picks then some would rise to the top instead of what we have now: the poor analysts drag down the good analysts and we get mediocrity and homogeneity. Essentially, your choices are RoboCheat or everybody else. I suggest you take the former.

BTW, if you get the first pick and Robo says the best back is really in the 10th spot then you should trade the first pick (or whatever pick) for the 10th pick. I describe this in more detail below, but there is no sense waiting until the 24th pick (in a 12 team league where you got the first pick) to get your second pick. You can trade for the 10th spot, get the true top pick, and then pick up another player with the 15th pick (assuming snake draft style). Thus, you get the pick you wanted in slot 10 and you get another pick 9 slots before you would have drafted your second pick (had you kept the #1 spot). See the Master Power Slot analysis below for more discussion of this topic.

Do The Opposite

"Do the opposite," Publisher, Major National Fantasy Football Magazine

This is just as foolish as doing what everyone else is doing. 'Doing the opposite' is like driving your car north in a southbound lane or, to use a football analogy, it is like throwing a bomb when you are on the 1 yard line. Sure, everybody runs short plays from the 1, but you'll do the opposite and go for the bomb! My point is that sometimes there is a very good reason people are drafting certain players, be they RBs, QBs or whatever, and if you 'do the opposite' then you better have a good reason to.

This 'expert' had the third draft pick in his mock draft and selected Peyton Manning. This is just wrong. If he wanted to draft Peyton he could have done so in the, say, 8th slot. He should have traded his 3rd slot for a 8th slot and that way his second round pick would have been better. Don't do the opposite just for the sake of it--in fact, don't do it at all, use the optimal strategy (discussed below).

This fantasy football 'expert' is a nice guy and does know a lot, but he was doing a lot of bragging because, apparently, he had won a few expert leagues in the prior year using this strategy. Busting your buddies butts is good form and I admire people willing to try new strategies, but blind squirrels find nuts once in a while. This expert won because his strategy happened to work not because it was optimal.

This expert was on to something, but he couldn't define it. To have a 'strategy' at all, you need to define what it is and it has to be reproducable. He's saying that everybody is drafting RBs in round one, that part we can understand, but what is the opposite of a RB? He took a QB, but why not a WR or maybe even a TE? How do we do the opposite in other rounds? He found the strategy worked, but he couldn't explain why and, for all he knows, he might just have been lucky.

Any strategy might work at any given time, but there is only one strategy that will optimize your winning percentage: On any given turn you must:
  • Select the player that:
    1. Is in the position that will experience the greatest tier delta between this pick and your next
    2. can fill an open starting position.

I think the thing most FF players and even most FF experts don't understand is item #2 (above). If so, you need to read the Upsider's Gambit. There is only one exception to this optimal strategy: you must select handcuffed players and franchise player backups if those players won't be around for your next pick. Typically, only RBs need handcuffs because it is the only position that (1) has a high propensity for injury and (2) you can accurately predict where the points will go if there is an injury. As a rule of thumb, if you drafted a RB in round one then you need his backup--even (perhaps especially) if it is a RBBC.

I should also mention one other exception to the above. It's not really an exception, but when the opportunity cost (the point drop, a.k.a, tier drop-off, penalty for not following the optimal strategy) is reasonably small you should try a schema shift gambit.

So there it is and as you can see the optimal strategy is very simple, but it doesn't exactly fill up a magazine or the airwaves with fluff now does it. It also has that word 'delta' in it and that means math and math doesn't sell.

The mainstream fantasy experts, good guys all, need to sell stuff, be it advertisements or subscriptions or whatever and will avoid math like my kids avoid vegtables. Listen and read mainstream experts, I do, but when they start talking hard-core strategy you might just want to tune out.

I know some of you probably just thought, "hypocrite, you sell subscriptions too." True, but FFD is different and different is not a good thing for FFD's bottom line, at least not in the short term. You can ask anyone in the industry whose taken the time to get to know FFD--they'll tell you they respect what we're doing, but it's rocket science and it won't sell. I don't think that anlalogy applies though--it's true that behind the scenes FFD implements sophisticated mathematics and algorithms, but that is just what the description implies: 'behind the scenes'. RoboCoach users don't have to know and can't see the complexity--they get their Robo reports and thus they get all the benefit of the sophistication without having to know it. To turn the analogy around: astronauts don't know jack about rocket science, but they know they want it.

Back to my point, here's the proof the aforementioned strategy is the optimal strategy. Imagine you had a perfect cheatsheet and it was your turn to draft. Would you 'do the opposite' of what everyone else was doing? Would you implement any positional strategy, e.g., RB First? No, you would select the difference maker, i.e., the player that gives you the largest positive tier delta (as shown by your magic cheatsheet). The player would be at a position that lacks a starter. In other words, there is no need to draft upside players to your bench when you have a magic cheatsheet!

So, given perfect knowledge you wouldn't have any positional strategy at all with your picks, i.e., if everybody were selecting RBs you wouldn't do-the-opposite. You would check your magic cheatsheet and pick according to the aforementioned rules. The same is true without a magic cheatsheet which I explain below, but first a little aside about the value of draft position.

The 'perfect knowledge' scenario also proves that no draft position is any better than another, intrinsically speaking. With perfect knowledge the optimal draft position isn't the first draft position. It can be, but it isn't necessarily. For every draft position there is an optimal selection. If everybody makes the optimal pick then who has won the draft? The draft slot that won is the team that put the most fantasy points on the field for the season. So if you add up all the expected points then the winning draft slot might have been the 12th slot, or the 10th, or the xth. Do you see my point? It doesn't matter. You can read more about this the fantasy draft slot debunking section of this page.

You have to treat your cheatsheet as if it were magic! For all intents and purposes it is magic. Nobody has perfect knowledge and a cheatsheet by Robocoach is as close as you will get. Even if you don't use Robo, you have to play your cheatsheet as if it were perfect. Doing otherwise can sometimes work, but it isn't optimal strategy to assume your cheatsheet is flawed and draft bench players to hedge against the flaws. This is what everybody does, however, and you can use that to your advantage if you know how.

That said, I do have to admit some things tend to be true. As you move through higher draft slots the player value differentials do drop, i.e., the drop in value from the 1st to the 2nd pick is significantly more than the drop from the 12th to the 13th and so on. FF players understand this and you can use thier knowledge against them and trade up to the Master Power Slot. Suffice to say, first out of the gate isn't the winner--there is a whole track to run.

Per my analysis, the snake in the draft does just what it is designed to do: it nullifies the advantage of early draft picks. I've only seen one instance where the advantage wasn't nulified (Marshall Faulk in 2000 and maybe Priest in 2003).

Another admission, doing the opposite does creates value, but it isn't what DTO 'expert' proponents say it is, i.e., getting high output from positions other people aren't drafting. DTO creates 'value' because you are creating trade differentials with the other teams.

Have you ever noticed that you can't seem to find good trade opportunities with other teams? Its probably because you and the other team had the same draft day strategy, e.g., 'coincidentally' you and the other team are stacked at RB and you're both weak at WR, QB and TE. If you DTO then you create trade opportunities because you'll be strong where others are weak. The guy who drafted three stud RBs at the draft will suddenly give you a stud to get one of your OK (but not that great) WRs. That is what DTO really does, it necessarliy creates trade opportunities, but it doesn't necessarily create higher team output.

Are you surprised the 'experts' didn't notice the true 'value' of DTO? You shouldn't be surprised by now. I'll repeat, these 'experts' are just journalists. In case you can't remember high school: writers were the ones who were no good at math. Math is the basis of winning strategy games like fantasy football and so don't listen to fantasy journalists for strategy tips.

I do have to note one exception to the 'journalist suck' rule that I just laid out. A really good journalist is reliable. You can define a really good journalist as someone who knows a ton about football, but can't really translate that knowledge into formulas. That journalist is what a computer scientist would call a perfectly weighted neural network. They are hard to find and change every year, but you'll know them when you find them.

These journalist don't understand why some of the things they say work--they just know they work. They can give you a fish (sometimes), but they can't teach you a fish. When the environment changes, e.g., when everybody starts DTO, these 'experts' typically can't adjust because they don't know why their strategy was working. They don't understand the mathematical fundamentals that belie their strategies, but Robocoach does!

Case in point, the dude who wrote the 'DTO' article was very successful one year and so half the guys in his draft were implementing DTO the next year. In the latter year, he probably had to take Peyton where he did take him (the third slot) because his opponents (drafting in the 4th, 5th, 6th and 7th slots) said they were implementing DTO too and so would have taken a QB (Peyton). Even when Peyton went off the board some still stuck to DTO and grabbed a QB.

This is strategic evolution in progress--copy the winner. The funny thing is that he thought he was doing the opposite (taking a QB early), but so was everybody else. if everybody is DTOing you should be DTOOing in order to be DTOing. Ouch my head hurts.

Do you get it? So, the aforementioned 'expert' who drafted a Peyton with the third pick thought he was DTOing, but since half his league was DTOing he was actually doing what everybody else was. He's wasn't being flexible. The river used to flow north, but, in the middle of the boat ride, it started flowing south and now he's going with the flow when he thought he was going upstream. In summary, novel draft strategies work sometimes, but people catch on and over the long term your best bet is to use the optimal strategy presented above.

Also, your best bet is to avoid using strategy tips from journalists. They are selling yesterday's winners! Please listen to RoboCheat he'll steer you right on draft day. He can't teach you to fish, but if you've got a RoboFisherman who needs to fish anyway?

The First Draft Slot is The Best

If you're lucky enough to get the first pick you should draft so-and-so, Every Fantasy Analyst, Every Media, bar none

Draft slots don't have intrinsic value, e.g., the 1st slot is no better than the 12th slot. The 'snake' in a Snake Draft does what it is designed to do, i.e., equalize slot value, but most (if not all) fantasy players don't trust the snake mechinism--they still think lower picks are better. They should trust the snake and better yet they should learn to ride the snake. Riding the snake is easy if you know what draft slot to mount your saddle on.

First a little background: in a twelve team snake draft, if you have the 12th pick then you also have the 13th pick, but the 1st slot has to wait until the 24th slot for their second pick.

For example in 2006, the owner with the first pick in one of my drafts took Tomlinson (350 weighted points) and I picked up Culpepper (320 weighted points) with the 12th. My second pick had to make up the 30 point differential (350-320). I was able to cover the differential because I had an advantage, i.e., I had the first pick in round two. Culpepper stunk that year, but that doesn't change the point, i.e., the first slot is no better than any other slot.

That said, some slots are better, but it has nothing to do with the slot number, it depends on who is the biggest second round sleeper. BTW, I think this is a Dr. RoboStein original so if you adopt the following analysis please refer to me in some way. Maybe you can call it the Master Power Slot (MPS) which requires finding the McSleeper or something stupid like that. Back to my point, to find the MPS you need to find the MasterSleeper which is merely the biggest sleeper in the second round. You can find the McSleeper by looking at the Delta column on the RoboCheat cheatsheet. The McSleeper is the second rounder (per ADP) with the highest positive Delta (ADP-RDP). For example, in 2008, the biggest sleeper in my twelve team league was Willis McGahee. As you can see in my RoboCheat cheatsheet table (below) Willis was a second rounder (12 < ADP <= 24) and (after looking at the other positions) I found he had the highest Delta (16) and thus he was my McSleeper.


To find the MPS you need to examine his WUP (Wait Until Pick) which tells us the earliest slot we should use if we wish to draft him. The MPS is the first round draft slot that gets Willis on the wrap so we just subtract twelve from his WUP. Mathematically: MPS = McSleeper WUP - No Teams In League or MPS = 15 - 12 = 3. The third slot has an advantage--it is the MPS.

I didn't get the first pick in my draft, but if I had I would have traded up to the third slot. I would have tried also to get my trade partner to throw in something extra. Your league opponents will gladly trade down because they all think lower picks are better. They are wrong--the MPS is better and the MPS might be slot 12 in a twelve team draft! Additionally, don't forget that if you have the first pick or second pick you can usually trade up and get your trade partner to give you an extra pick, say, in the fifth round. This might seem like wheeling and dealing, but it is really just plain math. Your friends will laugh at you, but if you find yourself in this situation then the winner's check already has your name on it.

There is a rare case where the MPS doesn't apply. I call this the 'Faulk' exception because Marshall Faulk at his best in, I think 2000, invalidated the MPS. In terms of Weighted Value (the 'Robo' column in the image above) Faulk was so far ahead of anyone that you shouldn't have traded up to the MPS if you had the first pick, second pick yes, but not the first (Faulk). Essentially, what you gave up by leaving Faulk behind couldn't be made up by any second round sleeper. I've not seen this scenario since and Faulk was such an exceptional producer (pass-run threat) that I may not ever see it again. BTW, in case you couldn't tell, like many fantasy players who won their leagues because of Marshall--I have a huge man-crush on him.

I'd guess the Faulk exception kicks in when a player is forecast to be, say, an extra 100 weighted points above the next best player at their position, but that is so rare that you would be safe to ignore it. Priest Holmes, in 2003, fit this profile, but at the beginning of the season nobody really could have projected it, i.e., we all (Robocoach included) expected Priest was going to be good, but not that good! In any event, you couldn't have planned for Priest's big season at the time of the draft and so the MPS was a valid draft day strategy.

Note also that MPS isn't really limited to a first round-second round strategy play, but for simplification purposes I'm presenting it as such. For example, you can increase your horizon to include the third round and then the MPS becomes the slot that gets the highest weighted point sum over your first three picks (instead of the first two). Likewise, MPS IS not limited to the first round and can be implemented at any time in the draft, although your commish and commish software would probably throw a fit if you started trading slots in mid-draft.

If you want to take a load off your mind then just use RoboDraft as all this is part of the RoboDraft algorithm and all you'll have to do is calculate the 'simple' MPS described here. I only mention the other possiblilites for MPS here to show you how much math is really going on with every pick and, despite the remarkable ability of humans to apply their personal neural networks and approximate the right answer, it's a lot more fun to sit back, have a robot tell you what to do and drink a beer. Give your synapses a rest.

In 2006 Julius Jones was my league's McSleeper and I know most of you are saying that Julius stunk up the joint in 2006. That is true, Marion Barber emerged and Julius lost production that was his for the taking. BTW, this was a scenario that some predicted and Robo's algorithm should have and now does handle situations such as this. Bad code aside, the point is still valid: if you have a McSleeper, that is, a first round player that is going in the second round you must trade up to the MPS so you get the McSleeper without drafting them way to early.

WUP is very important in all of this regardless of MPS or McSleeper. If the do-the-opposite analyst considered WUP then he would have known that Peyton wasn't the appropriate pick in the third slot because there are too many players that will return a higher weighted point total (that won't be around with his next pick) than will Peyton. In the eigth slot that isn't necessarily true and thus Peyton's WUP was eight that year. Ignoring MPS you still can trade up from third to eigth if you want Peyton, i.e., trade your slot--don't be stupid and draft Peyton third (well before his WUP).

Avoid RBBC Players

"Avoid RBBCs: running backs who share the backfield" Every Publisher, Every Major National Fantasy Football Magazine

Just plain wrong again--embrace RBBC, don't avoid it. Folks, you are going to have to get used to RBBC because it works. I've done the analysis and in almost every case RBBC production exceeds that of a traditional one-running-back backfield.

Most NFL coaches know RBBC works, I suspect, and that is why we see the RBBC strategy increasingly employed in the NFL. Think about it, all four teams in the 2007 AFC-NFC Championship games employed RBBC (Patriots, Colts, Bears and Saints)! I think the only thing holding this strategy from full NFL deployment is coach-RB loyalty. But the days of a coach being loyal to one RB are numbered.

Note, I am saying RBBC has greater position production, not player production! Did I just contradict my earlier statement? No! You don't avoid RBBC, RBBC is just another factor to weigh when determining a RB's value. Most fantasy gurus' simple spreadsheets can't handle the RBBC factor--they don't know how it will affect production and so they just add a fudge factor and drop RBBC backs way down in their cheat sheet order--waaaaaaay down. Consequently, great RBs in a two back system like Joe Addai, in 2006, get undervalued. The gurus are just guessing, if they are wrong they can say, "well nobody thought he was going to be that good. (see group think). They should say "nobody, except Robocoach" because Robo knows how to correctly value RBBC backs and he will gladly draft/pick-up RBBC backs because they usually are undervalued.

Go for upside players in the middle rounds (upsider gambit)

Most people I know can't resist grabbing an 'upside' player and I lovingly refer to them as 'upsiders'. For example, an upsider might already have their QB slot filled and they draft a second QB before getting their second WR or first TE, etc. Resist the temptation: don't draft a player to your bench when you have open starting slots. Upside players will be on your bench (at least for a while) and bench players score zero points every Sunday. There are exceptions to this rule of thumb (discussed below), but if you just avoid upsiding and read no more then you'll be in good shape.

The payback of the 'upside gambit', as I call it, comes when a starter doesn't pan out and you stick in the upside prospect. However, drafting an upside prospect puts the upsider behind other teams that are diligently filling starter spots. In other words, upsiders are moving themselves to the end of the line when it comes to filling thier open starting slots--and that increases the risk that those positions will under-produce. Under production is the risk the upsider takes, but from a numbers perspective the payback for that risk appears rather slim--even when allowing the best case scenario.

The payback of the upside gambit is:
  • (chance starter is a dud) X (chance the upside player becomes a stud) = chance upside strategy will work

Some reading this are thinking, "what about the scenario where they both are studs." I'll discuss that below, but essentially that scenario forces the upsider into trading one player to prevent their new stud's points from languishing on the bench.

Back to the formula, to be generous assume there is a 50-50 chance for each parameter above and so, baring trades, the reasonable expectation is the stategy will work 50% * 50% = 25% percent of the time.

What is the payback for taking the risk? Not much. If the gambit is to pay off then (baring trade) the starter must reveal themselves to be a dud thus opening a starting slot. After it becomes apparent the starter is a dud, and that can be many weeks, the upsider begins to get their payback. Being generous again, assume the starter reveals his dudness early, by week three. The payback is the sum of the following from week three to the end of the season:
  • (Upside Player Production) - [Higher of: [Dud Starter Production or Best Available Waiver Player Production])

Let's be really generous and assume a Reggie Bush like best-case scenario. I originally was thinking Adrian Peterson would be a good best case, but AP studded and dudded (with injury) in the same season. Reggie was somewhat of a dud himself until about mid-2006-season, but that is when the upsider expects the production to begin. In any event, starting at week three (2006) the delta between Reggie and the best available waiver player was large-- about 110 (weighted) points (in my league) or approximately 10 extra points per week with Reggie versus the likely alternative. Baring trade, the payback of this gambit was:
  • Payback = chance of payoff * payoff

  • Payback = 25% * 110 points =~ 25 points

That's a somewhat small payoff, but every little bit counts and 25 points can make a big difference depending on when you get it. However, in a most-likely case scenario the payback would be cut in about half, and 12 to 13 weighted points just isn't much. However that's not the end of the payback calculation. The upsider must also subtract the points they forsaked. The points the upsider forgos is the sum of the production of the starters the upsider would have had (sans-Reggie) less the production of the starters the upsider did have.

To draft Bush his rookie year (2006) the upsider had to draft him in about the third round, but lets be generous again and say the upsider got him later--the fifth round. What an awesome steal the upsider thinks! Assume in rounds one through four the upsider filled their roster in typical style, e.g., RB, RB, WR, QB. In a typical league setup that means the following starter slots are open in round five: WR, TE, DF, K.

Instead of filling an open slot the upsider selects Bush in round five and thus should get different, and presumably worse, players at the remaining starting slots because they'll all be filled a round later than was possible. Mathematically the upsiders loses the sum of:
  • The Round 6 WR they would have got - the Round 7 WR they did get
  • The Round 7 TE they would have got - the Round 8 TE they did get
  • The Round 8 DF they would have got - the Round 9 DF they did get
  • The Round 9 K they would have got - the Round 10 K they did get

I call this part of the equation the 'tier delta' summation. Tier deltas kick in when the gambit forces the upsider to draft players from a lower tier than they might have. A little background on tiers might be appropriate here, tier is player value and it's appropriate to think of players' value tied to tiers as opposed to some number calculated by Robo or your spreadsheet. Tiers are more appropriate than numbers because predictions have a large margin of error. For example, the cheatsheet table shows Tomlinson, Jackson and McGahee in the same tier with a value of 535. Tomlinson's numerical forecast, however, is much higher than the tier value (590 vs. 535) so the question arises, "is Tomlinson worth more?" Probably so, but because of inherent ambiguity in forecasting, your best bet is to say, relative to all the other players, these three players are about the same and place more confidence in the tier value. There is some randomness in calculating tiers and some may argue whether, in 2008, McGahee should belong to same tier as Jackson and Tomlinson. Regardless of your position in that debate, the point about tiering players is valid and you should place more confidence in the tier value than any other forecast value.

Continuing this example and refering to the cheatsheet table, the tier changes with the heavy black line and you can see Westbrook belongs to the next tier valued at 430. That is a huge, 535-430 = 105 point tier delta! However, in the later rounds the tier deltas narrow and, by rounds five or six in my league, the tier deltas were more like ten or twenty points.

Long story short: each of the above items in the summation might penalize the upside gambit at about 15 points per position (more before round five and less after). I'll get to why I said 'might' instead of 'will' below, but for now lets think like an upsider and move numbers around like the CFO of Global Crossing.

Upsiders can try to workaround this sorry situation, but it all nets the same. For example, the upsider might skip the WR in round six and (because the run on TE had yet to begin) they get a really good TE in round six and no tier delta penalty. The upsider can continue that workaround with the other open positions (DF, K), but eventually they've got to fill that open WR position and now they have to face a single big tier delta at WR (in round 9) instead of smaller deltas spread over each position.

Now I'll address why I said upsiders 'might' be, but typically aren't, penalized. The clue to upsider penalty avoidance is shown in the tier delta items (above). Note, for DF and K the draft rounds are rounds 8 and 9 respectively. Most of the experience FF players reading this will immediately say that you never take a DF or K until the last round. In practice, that is true (btw, if you truly believe that you should read the DF and K Last Maxim debunking). While it's true that prevailing practices allow K and DF in later (last) rounds, that doesn't mean my upside gambit analysis is flawed. Prevailing practices simply tell us that everybody, plays the upside gambit at some point in the draft. Some start early as in this example and some start later. The later you start the less you risk, but anytime you take an upside player before filling a roster position you are employing the upside gambit.

So what does all this mean? It means that, in most drafts the sum of the items above is usually zero and the upside gambit payback remains positive! With all those deductions you'd think the payback would flip to negative, but these sum to zero because:
  • If everybody does it there is no tier delta penalty for upsiders.
  • Tiers are somewhat wide, about 3 to 5 players wide, per tier per position, and that cuts the upsiders chance of dropping to a lower tier unless a similar number of teams draft 'straight'.
  • It usually takes two or more tier delta penalties to offset the payback and the above facts combine to reduce the chance of incurring a tier delta penalty.
  • Upsiders band together and perpetuate the strategy. Ever wonder why you were mocked for taking a kicker or defense in the middle rounds? Now you know. Upsiders know there is an opportunity cost for their strategy--they don't know exactly how much, but they know it's less if everybody follows their lead.

Bottom line, the payback is positive, but only when upsider opponents don't make the upsider pay the penalty for their gambit. If enough teams, probably around half, play it straight (no upside gambit) then upsiders get burned. Joining against upsiders creates more tier deltas across multiple positions and the gambit payback flips to negative.

So far, this analysis assumes a single team universe, i.e., upsiders keep the prospect player just for thier team to hedge against a starter dudding-out. In reality, there are more teams and the upsider can trade the player if need be. Allowing for trades, most of the same numbers still apply, that is, there is still (at best) a 50% shot of the prospect player being a stud. Assume the upsider gets fair value in a trade, and still assume that the hedge is worthless unless a starter duds-out. In other words, if all the upsider's players are good then there is no room for trades--the points will go to the bench (and bench points don't count).

The trade scenario is slightly different from the single-team universe scenario in some respects. Both hedge against a dud with the difference being the single-team universe hedges only against a dud at the one position (the prospect player's position), but trades enable the upside prospect to hedge against duds at any position. This changes a parameter value in our first equation (shown again below)
  • (chance starter is a dud) X (chance the upside player becomes a stud) = chance upside strategy will work

Remember, for this strategy to have a payback the upsider needs a dud. In our earlier example we generously assumed chance starter is a dud to be 50%. Including trades allows the upside strategy to hedge dudness at more positions and so chances the hedge works for a dud increase too, let's say to 75%.
  • 75% X 50% = chance upside strategy will work =~ 40%
So you can conclude already that allowing trades improves the payback, but by how much, i.e., is it significant?

Returning to our best-case Reggie Bush scenario and assuming perfect trade liquidity where you get back exactly the (pent-up) value that you traded away.
  • Payback = chance of payoff * payoff

  • Payback = 40% * 110 points =~ 40 points (best-case)
That improves the payback versus the single team universe, but not so much (15 points). Once again, this is best case and so most likely case would be about half that.

Bottom line: when adjusting for trade ability, the upside gambit payback increases a bit. The added value appears to be about equal to the value of an average (middle round) tier delta (15 points in my league).

In summary, going for upside in the middle rounds is STILL bunk. We found positive payback, but the payback is small and is positive only if most of the other drafters play the upside gambit too. If enough teams draft straight, i.e., fill their position needs first, the payback will flip to negative and the upsider will need to draft straight. Don't do it.