Tigerpiper
Footballguy
30 is obviously better. You just have to choose the correct 30.
Favre, Henne and Anderson at QB not looking so good. Go Team 102472!
Favre, Henne and Anderson at QB not looking so good. Go Team 102472!
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!
Subscriber Contest (2 Viewers)
- Thread starter eaglezzz
- Start date
Jayrod
Footballguy
Not necessarily. You can only count 11 scores per week, so the optimal team is likely somewhere in the middle 20's....enough different guys to score the big points in the right weeks, but enough cash available to pay for the big money guys who are the truly big time scorers.
CalBear
Footballguy
In a best ball league, three mediocre guys will score more than one stud.
jdoggydogg
Footballguy
Yes.
That depends on what you mean.I don't think it's true if you mean it for a given week.jdoggydogg said:
For example, this week's projections have Miles Austin projected at 18 points, which is a typical number for a top-5 WR. Now take Mike Wallace, Nate Washington, and Mike Thomas, who are projected at about 11 points apiece. My simulator says that the best score out of Wallace/Washington/Thomas will average about 16.6 points, which still makes Austin the better play. [For the record, three 12 PPG players (Owens, Harvin, Garcon) is about equivalent, on a weekly basis, to an 18 PPG player].
The reason you want a big roster (IMO) is not necessarily because of the weekly advantage in a best ball format. It's because it gives you more chances of ending up with the breakout players.
In other words, I think that three 11 PPG receivers are NOT better than an 18 PPG player. However, if the three guys that look right now like 11 PPG players actually turn out to be a 16 PPG player, an 11 PPG player, and a 6 PPG player, then you are better off with the three than the one (says the simulator).
(Maybe that's what you meant. Disregard if so.)
Dinsy Ejotuz
Footballguy
Could you rerun that three vs one sim as if Austin and Nate Washington both miss the last ten games with broken legs?
The more I think about it, the less confident I am that well-constructed large rosters have a better chance of winning (at the start of the season) than well-constructed small rosters.
Yes, the small rosters took a beating last year, and they continued to take a beating in week one. The more I think about it, though, the more I think this really doesn't mean much at all. Why? Because those results only reflect the results in the actual season that occurred, and not the gajillion other seasons that might have occurred instead.
It's a basic rule in huge contests like this that any strategy that gives you a better chance of finishing first also gives you a better chance of finishing last.
Consider variance, for instance. As has been discussed ad nauseum on this board, the QB/WR hookup increases your variance. This increases your chances of a great week, but it also increases your chances of a terrible week. That's great if you're an underdog and bad if you're a favorite. The key point is that, in a contest like this, everyone is an underdog. A HUUUUGE underdog. If you're an underdog and you pursue a high-variance strategy (as underdogs should), the most likely result is a loss. A bad loss. But that doesn't mean that what you did didn't give you a better chance of winning. It just means it didn't work out.
I guess what I'm saying is this: we do not have a sample size of 26,000 on which to base our hypotheses about big vs small rosters. We have a sample size of one. After this season, the sample size will be two. And that means nothing.
To summarize:
1. I have been a longtime advocate of the majority opinion here: that larger rosters are better. But I've now moved to an agnostic position.
2. I'm not saying this is the case, but I think it's quite possible that a small roster of studs has a better chance of winning it all than a large roster. The small roster also has a greater chance of flaming out early, but that's not relevant to the discussion.
3. I do not believe that the aggregate data on survival rates of different-sized rosters tells us anything about the best a priori strategy.
[DISCLAIMER: all this assumes that what we're interested in is maximizing our chances of finishing #1. I suspect many people, including me, get a lot of gratification from surviving as long as possible, even if they don't win. In that case, the discussion would be much different.]
Yes, the small rosters took a beating last year, and they continued to take a beating in week one. The more I think about it, though, the more I think this really doesn't mean much at all. Why? Because those results only reflect the results in the actual season that occurred, and not the gajillion other seasons that might have occurred instead.
It's a basic rule in huge contests like this that any strategy that gives you a better chance of finishing first also gives you a better chance of finishing last.
Consider variance, for instance. As has been discussed ad nauseum on this board, the QB/WR hookup increases your variance. This increases your chances of a great week, but it also increases your chances of a terrible week. That's great if you're an underdog and bad if you're a favorite. The key point is that, in a contest like this, everyone is an underdog. A HUUUUGE underdog. If you're an underdog and you pursue a high-variance strategy (as underdogs should), the most likely result is a loss. A bad loss. But that doesn't mean that what you did didn't give you a better chance of winning. It just means it didn't work out.
I guess what I'm saying is this: we do not have a sample size of 26,000 on which to base our hypotheses about big vs small rosters. We have a sample size of one. After this season, the sample size will be two. And that means nothing.
To summarize:
1. I have been a longtime advocate of the majority opinion here: that larger rosters are better. But I've now moved to an agnostic position.
2. I'm not saying this is the case, but I think it's quite possible that a small roster of studs has a better chance of winning it all than a large roster. The small roster also has a greater chance of flaming out early, but that's not relevant to the discussion.
3. I do not believe that the aggregate data on survival rates of different-sized rosters tells us anything about the best a priori strategy.
[DISCLAIMER: all this assumes that what we're interested in is maximizing our chances of finishing #1. I suspect many people, including me, get a lot of gratification from surviving as long as possible, even if they don't win. In that case, the discussion would be much different.]
ctriopelle
Footballguy
I think you sprained a bunch of brains there Doug! I take it you're a big fan of Nassim Taleb's alternate reality work? Cool stuff, but tough to get a grasp of, seeing as the probable outcome isn't what really happens, even if it was supposed to, and that's what everyone reacts to and builds further expectations on! Must be nice to have access to a working Monte Carlo simulation generator!
Yeah, I agree to a large extent. The results of varying roster sizes for one season are not definitive at all. This season will certainly be different from last season, and this year's contest features are certainly different than last season's contest.That being said, we do have one data point that larger roster sizes had higher success rates last year. Yeah, variance, yada, yada. But my gut feeling is that in most seasons, for contests with features similar to last year's contest, that larger roster sizes would still have higher success rates in the vast majority of potential seasons.
The question in my mind is whether a roster size of 30 is optimal this year -- compared to 27-29 players, or 24-26 players, or 21-23 players, or 18-20 players. At some point -- 40 players, 50 players, 60 players -- larger rosters will result in poor success rates.
I did 30 this year for the standard reasons -- catching the most break-out players, you get maximum value by packing your roster with value-priced players that have good upside, etc. But I'm having second thoughts that somewhere between 23 and 28 is probably optimal. Who knows? Time will tell. Definitely interesting contest!
ETA: An obvious prediction is that the winner of this year's contest will have the most powerful (high-scoring) combination of high-cost, medium-cost, and low-cost players who become hot and have huge performances during weeks 14, 15 and 16. An 18-player roster can certainly include the low-cost players who break-out and score tons of points in the last 3 weeks, and I think there's no reason why an 18-player roster can't win the contest. It's a matter of odds and what roster size is optimal for success. For this year's contest, I guess I still think it's 30.
Last edited by a moderator:
CalBear
Footballguy
I think it is true for a given week. Look at FBG's own QBBC articles; they've repeatedly shown that even in a non-best-ball format, three mediocre QBs can perform as well as a top QB, just by choosing your starter based on matchup. In a best-ball format, you'd do even better than that. I don't see a great data source for what I want to do, but I bet that if you look at the weekly scores for the players who ended up as WR23, WR24, and WR25 last year, that collectively they performed as a top-5 WR in best ball format. And I bet you could push it further down than that; possibly WR34-36 even.
Marauder
Footballguy
The question is, what is a "well constructed small roster". The problem is that the rosters that have the studs with the highest probablility of doing well are also going to be the rosters with the lowest uniqueness factor. This goes back to your point in the previous post. The way to win this contest is to have players on your roster who produce at stud level but are not commonly picked. Large rosters have a much better chance of doing that than small rosters. Maybe a small roster of high priced but uncommon players would be the highest variance style of all but I doubt many of the small roster guys are doing that.
CalBear
Footballguy
We have a lot of data about what NFL seasons look like. Among that data we have strong evidence that 50% or more of the players identified as studs in the pre-season do not perform like studs in reality, either due to injury or situation. In a world where 50% or less of the top-10 players started out in the top-10, in a best-ball format you're going to be better off with a larger number of players who weren't projected to be in the top 10, than with a smaller number of players who are projected to be in the top 10.The only way this would not be true is if there was a decent likelihood that the pre-season projections are close to correct. Since we know that's false, I think the premise that "studs" could be more important in this format is also false.
And our sample size is based on every week of the season, not the season as a whole. Smaller rosters have gotten killed in every one of what is now nearing 20 samples.
Ignoratio Elenchi
Footballguy
Game Log Dominator would get stuff like this for you pretty quickly. WRs 23-25 last year were Boldin, Sims-Walker, and Harvin. Together in a best-ball format they would have outperformed Andre Johnson 11 of the first 16 weeks last year (ETA: based on standard FBG scoring, which I don't think includes PPR. So per the contest scoring system this might not be entirely accurate).The problem with this kind of analysis, though, is that you're cherrypicking the WRs after the season. Before the season started would you have known to buy those three? Of course I'm not really interested in the answer to that question specifically, but more generally, what is the probability that if you buy 10 "cheap" WRs that three of them will end up being ranked in the top 25 by the end of the year? How does that compare to the probability of buying an expensive WR and having him perform like an expensive WR?
Last edited by a moderator:
Dinsy Ejotuz
Footballguy
Maybe I'm not thinking this through right, but...
Isn't there an argument that we didn't play one season last year? We played 12,000 (each team is a season). The fact that NO ONE managed to pick a combination of studs that worked with a smaller roster is even more damning to me than the fact that smaller rosters lost in every single one of our 20 sample weeks.
If it could be done, someone would have found the combo. It's not like there weren't studs last year - there were plenty.
Isn't there an argument that we didn't play one season last year? We played 12,000 (each team is a season). The fact that NO ONE managed to pick a combination of studs that worked with a smaller roster is even more damning to me than the fact that smaller rosters lost in every single one of our 20 sample weeks.
If it could be done, someone would have found the combo. It's not like there weren't studs last year - there were plenty.
Last edited by a moderator:
CalBear
Footballguy
Obviously there's cherry-picking going on there, but it's also going on for Andre Johnson. It so happens that Johnson was projected to be a top WR and turned out to be a top WR, but a bunch of guys projected to be good WRs turned out to be mediocre or bad. Hypothesis: The best finish of the WRs projected as WR23-25 before the season will be better than the average finish of the WRs projected as WR1-10.
Edit to add:
Pre-season, here's what we had:
WR1 Andre Johnson
WR2 Calvin Johnson
WR3 Randy Moss
WR4 Miles Austin
WR5 Roddy White
WR6 Reggie Wayne
WR7 Brandon Marshall
WR8 Larry Fitzgerald
WR9 Greg Jennings
WR10 Marques Colston
WR23 Michael Crabtree
WR24 Mike Williams (TB)
WR25 Hines Ward
Last edited by a moderator:
CalBear
Footballguy
This is not exactly correct, because our 12,000 seasons were not independent. Everyone who chose Matt Forte last year got hosed, and everyone who chose Chris Johnson did well.
nzranger
Footballguy
I'm not so sure about this.. Just doing some fuzzy math here but even if you relegated the number of eligible players down to just the "studs" in each positional category, and remember you have to knock 12,000 down to just the number of teams that took 18 players (on my iPhone too much of a pain in the ### to look up), I believe that would leave you with far more possible teams unused than used.In short I think I could make a roster of 18 players that beat last years winner. Doug i'd be interested to know, do you know if anyone has already done this?
Ignoratio Elenchi
Footballguy
The next problem with this, as it relates to the contest, is that (IIRC) Crabtree, Williams, and Ward would cost much more than Andre Johnson. If you want three WRs that cost the same as one "stud" then you have to move further down the list. But it would definitely be interesting to see how it plays out.What I'm interested in doing, but have never had the time to do, is reverse-engineering the optimal team... and then the 2nd best team, and the 3rd, etc. all the way down the line. Obviously there will be a lot of redundancy (e.g. the optimal roster might be composed of 19 players who cost $231 - there would be a large number of variations on this roster that include that core but spend that remaining $19 on other players in various ways) but I'm only interested in those core players. Will the most optimal rosters consist of 14 "studs" or 24 cheaper guys, or an even mixture? If the best possible roster every year consists mostly of guys that were projected to be preseason studs, then a small stud-heavy roster may be the way to go - as Doug pointed out, odds are you won't pick the right ones, but you have to take the chance that you will. OTOH we might find that every year the optimal rosters consist mainly of cheap guys that come out of nowhere, in which case a large roster of cheaper guys is the way to go...
Ignoratio Elenchi
Footballguy
This is something I've always meant to do. In my mind, it's not too hard, because you only have to optimize the roster for weeks 14-16, and then tweak it to make sure it would've survived weeks 1-13. But I think I'm probably underestimating how much work it would actually be. And of course, once we start introducing hypothetical rosters, the idea of "would have survived weeks 1-13" becomes slightly less well-defined. I'll probably take a stab at it at some point, but not any time soon.
Ahmad Rashad
Footballguy
Yes, WR's 23-25 in the contest cost $59 total, while AJ cost only $32 (http://thefantasystar.com/fbg35k/2010.php ). If you can put together a roster with 30 players with nearly double the budget of an 18-player team, I agree that would be a better roster.
For 3 consecutive WR's to have the same value as AJ this year, you'd have to go with WR's 50-52 (last year it would have been WR's 38-40 to reach the cost of WR1). If AJ performs as WR1 and WR's 50-52 perform as WR50-52, and all stay healthy, I think AJ will outscore the cheapies. The problem is any of the 4 might get injured and might outperform or underperform their slot. If you look in hindsight and pick the right 3 cheapies that total $32, I'm pretty sure those 3 will outperform WR1, but they will end the year ranked much higher than WR's 50-52. The key isn't so much roster size but picking the right cheapies and studs. With this year's rules, studs were much cheaper and cheapies were much more expensive, so smaller rosters should have a much better chance than last year. Everyone, even those with the top studs, will have a bunch of cheapies on their roster -- the ones who do the best will be the ones who pick the right cheapies and whose "studs" perform at least as expected.
With last year's rules, it made sense to go with the larger roster because there were so many high-value cheapies. With this year's rules, it's not so clear. If Chris Johnson goes for 2500 yards like he predicts, he'll end up being a great value at $40, even though he's the most expensive player this year. Combining him with cheapies like Foster and Brandon Jackson might prove to be a winning combination, even on an 18-player roster with similar success at other positions, probably blowing away all the 30-player rosters with no studs.
Last year's most expensive player was $50, and there were far more $1 and very cheap players, so there was a much bigger advantage to the rosters loaded with cheapies. Last year's results do NOT prove that it's much better to go with a 30-man roster rather than an 18-man roster this year. It's a different contest.
WisWolvrns
Footballguy
great discussions on roster size and such. interesting stuff. ( I thought bell curve, but I'm not a math guy. )
as for the top scorers for a week having less chance to survive, I look at it that a top scoring team most probably had multiple breakout players that week. in any given season how many times does a player have a breakout scoring week? depending on the player, it could be from 0-4?. The studs score decent week to week, with less down weeks than the average guys but also less #1 weeks in some cases, but the average to low guys are more shoot the moon in score or score minimal. The timing is everything. If a given week has your roster score multiple shoot the moon players at the same time, you get a top score for the week. The problem is, then other weeks they will more likely burn out, giving your team a big hole to try and fill with other player scores to make the cut. So the luck factor is to spread out those shoot the moon scores around 1 or 2 per week with average scoring for your other players to fill out the score and make the cut. Then at the end, total points week 14-16, the teams which end up with the right shoot the moon players those weeks have the best chance of winning in my view. It's skill to select players that fit the category, but luck to having them not all score big all the same week (or more specifically, all crash in the same week!)
swinging this back to the roster sizes, I lean toward the larger rosters to have more shots amongst the shoot the moon type players, while paying some stud money for 1 player per position to give some backbone to my roster. Albeit, so far I haven't hit quite right to make the final weeks, but injuries and some poorly placed $'s have cost my previous years....
as for the reverse engineering idea, that would be cool. the core players for week 14-16 would work and possibly just a top 10 scorers for each of the 1-13 weeks and cherry pick the cheap shoot the moon players would work in making sure each cut line is passed. it wouldn't surprise me if starting with the top 5 scorers for weeks 14-16, taking any common players and filling in with as many of the others as possible would be enough to do it, since those shoot the moon players in weeks 14-16 are the ones which will make the difference in winning the whole contest.
as for the top scorers for a week having less chance to survive, I look at it that a top scoring team most probably had multiple breakout players that week. in any given season how many times does a player have a breakout scoring week? depending on the player, it could be from 0-4?. The studs score decent week to week, with less down weeks than the average guys but also less #1 weeks in some cases, but the average to low guys are more shoot the moon in score or score minimal. The timing is everything. If a given week has your roster score multiple shoot the moon players at the same time, you get a top score for the week. The problem is, then other weeks they will more likely burn out, giving your team a big hole to try and fill with other player scores to make the cut. So the luck factor is to spread out those shoot the moon scores around 1 or 2 per week with average scoring for your other players to fill out the score and make the cut. Then at the end, total points week 14-16, the teams which end up with the right shoot the moon players those weeks have the best chance of winning in my view. It's skill to select players that fit the category, but luck to having them not all score big all the same week (or more specifically, all crash in the same week!)
swinging this back to the roster sizes, I lean toward the larger rosters to have more shots amongst the shoot the moon type players, while paying some stud money for 1 player per position to give some backbone to my roster. Albeit, so far I haven't hit quite right to make the final weeks, but injuries and some poorly placed $'s have cost my previous years....
as for the reverse engineering idea, that would be cool. the core players for week 14-16 would work and possibly just a top 10 scorers for each of the 1-13 weeks and cherry pick the cheap shoot the moon players would work in making sure each cut line is passed. it wouldn't surprise me if starting with the top 5 scorers for weeks 14-16, taking any common players and filling in with as many of the others as possible would be enough to do it, since those shoot the moon players in weeks 14-16 are the ones which will make the difference in winning the whole contest.
WisWolvrns
Footballguy
actually, it is probably quick to figure out the best week 14-16 roster. since it is cumulative scoring, one would just need the total points scored per player over weeks 14-16 and the cost of each at the start of the season. then just go down the list and pick the best qb, 2rb, 3wr, 1te, 1flex, 1pk, 1def and that would be it. moving down some if $'s don't stay below the minimum, but figuring some of the top scorers those weeks were cheaper bust/boon players, I doubt the roster cost will be to crazy.
to confirm each week's cut survival, all that would be needed is the scores of the players selected above for each week 1-13 and what the cut line was for each week 1-13. tally the totals to see if they made the cut. any weeks they didn't, a top 10 scorers list for the week probably has a cheap boon/bust player in the list, who's points would be enough to bump the roster over the cut.
to confirm each week's cut survival, all that would be needed is the scores of the players selected above for each week 1-13 and what the cut line was for each week 1-13. tally the totals to see if they made the cut. any weeks they didn't, a top 10 scorers list for the week probably has a cheap boon/bust player in the list, who's points would be enough to bump the roster over the cut.
WisWolvrns
Footballguy
PS> cool feature with the bye's in the contest entry page for each roster. albeit... not so cool about the out?'s [ not your guys fault though! lol just is a word I hope to not see when I check my team page! The only words on the page that matter to me are: This entry is still alive. for as long as it lasts! ]
jdoggydogg
Footballguy
I agree. The reason I couldn't spend a lot of money for a top five guy at any position is there's too much risk. Yes, there will be guys (Austin? Chris Johnson?) that kill it every week. But for every uber stud season, there are 10 or 20 guys that are very inconsistent.
jdoggydogg
Footballguy
I think I'll always opt for a large roster, but your points are fantastic. I love it that you're getting into the minutiae of probability theory. Fascinating.
LittlePhatty
Footballguy
I don't think it is that simple at all.Let's say you find the "stud" RB that adds up to the most points over Weeks 14 - 16. Mr. Stud scored 63 points total in those 3 weeks, more than any other single RB.
Stud RB1:
w14 = 23
w15 = 19
w16 = 21
= 63 for you
That doesn't mean he's the guy to own.
It's absolutely possible that several other RBs in combination would have gotten you more points. And if you look at the top entries, you'll see they didn't use the points from the same stud guys every week.
You'd actually get more points by owning 3 different RBs that had big weeks independent of one another, even if none if them came close to scoring 63 points individually over those weeks. Like this:
Scrub RB1:
w14 = 34
w15 = 7
w16 = 0
Scrub RB2:
w14 = 11
w15 = 29
w16 = 7
Scrub RB3:
w14 = 3
w15 = 9
w16 = 31
= 94 for you
None of those guys got 63 points individually total in those three weeks, but when combined on your larger roster they boosted your points total significantly.
These are random numbers for the example, and they are exaggerated.
It's been pointed out before that the winner last year had breakout guys like Jermichael Finley, Jamaal Charles, etc. but a LOT of the final 250 had those guys. The real difference in being #1 and being #2 could have come down to any of the other picks on his roster like the Lions TD or Earl Bennett - both of which contributed to his points in the final three weeks, but they were no where near the top scorers for the entire season.
Ignoratio Elenchi
Footballguy
Optimizing the roster for weeks 14-16 would be pretty easy - if I had all the players, their prices, and their points for those three weeks in Excel, I could run Solver to find the optimal lineup. The trickier part would then be determining whether such a roster would have survived the first 13 weeks to get to the final 250. It would also be a pain to then find the 2nd most optimal roster, and the third most optimal roster, etc. I suppose I could turn players on and off with a binary switch but there are so many combinations it would be prohibitive in Excel, and I don't have the programming expertise to do it in a more elegant way. But finding the optimal 2009 roster core wouldn't be too hard if I had all the prices and scores at my fingertips (which, currently, I don't).
Actually your sample size for last year is 16. The large rosters dominated every single week. The smaller rosters made up the most entires last year and only held down one of the final ten spots as I recall.
Your sample size is one. The 16 weeks are not independent of each other (ie. Ronnie Brown getting hurt in week whatever, and is out for the season, affects the remaineder of the weeks). But either way , it's not nearly enough samples to accurately conclude anything because of all the possible combinations of events that could occur.
Think about something as simple as a role of the die. Let's say you role it 1 set (your sample) of 16 times and 4, 5 or 6 comes up 12 times. 1, 2 and 3 come up the other 4. Does that mean that betting that the 4, 5 or 6 will occur is the better strategy than betting on that the 1, 2 or 3 will occur? It isn't, we know this for a fact.
Now extend this simple example into the impossibly complex world of football and all the unknown probabilities. You say only 1 small roster team held the top 10 spots, but how much would that change if ADP doesn't get stuffed 2x at the GL vs Baltimore in week 6. That's an additional 6 points for all ADP owners, how many got ADP owners got eliminated that week because of that? And that's just 1 play of 1 game of 1 week. For all we know that single play could have made 2 or 3 smaller teams make the top 10.
What we have is one sample, and to draw absolute conclusions on it, is simply put, foolish.
CalBear
Footballguy
Look, no matter what happens, studs will have off days; we have 80 years of data which shows that. Take Marshall Faulk in 2000; he had less than 10 points twice, despite having one of the greatest fantasy RB seasons of all time. Or Priest Holmes in 2003; five weeks under 13 points. In 2002 he had two weeks under 12 points before he missed the last two games of the season. Tomlinson in 2006 had three games under 11 points. So, even when you look at the studliest of studly seasons, there are down games. Yes, Tomlinson "could have" scored two TDs in week 5 of 2006, but he didn't, and what's more, it is completely predictable that he would have down weeks no matter how many NFL records he broke that year.In a theoretical season where studs score 15+ points every game, you should clearly take studs. But we have plenty of data to show that such a season does not exist.
Look, no one was saying that stud's never have down games and I agree 100% that we have plenty of data showing that they do even in their best seasons. However, this argument doesn't address what I was and what DD was saying. The point is, there are an infinite possible paths for a season to take, and to look at a single one and try to use those results to conclude something is beyond silly. I understand that ADP didn't score that TD in week 6 and Tomlinson didn't score TD's, the point is that if he had, the results could have been very different and "proven" that smaller rosters were the way to go. We don't know, but we only have 1 sample point to look at, and as you said, those events didn't happen. The problem with most of the studs vs lesser players debates is that the arguments made are being made in a vacuum. We don't have the luxury of knowing who will finish in the top 3, or top 5, or top 25. Most of these arguments don't seem to take into account the thought that by taking a couple stud RB's they're free to take less of them and thus spend more money and rosters spots at other positions.The truth is that I understand completely the logic behind those that say the larger rosters are better, and I completely understand those that say it's the lower rosters. I think both have valid points, and DD said above, one is trying to prolong their life in the contest, while the other is swinging for the fences thinking they want to maximize their chances of winning, even if it decreases their chances of finishing in the money.I'm actually of the group (probably a small one) that believes that there probably isn't an optimal strategy, and that even in the event that there was it'd be impossible for anyone to prove (in this lifetime at least).
CalBear
Footballguy
The point is, taking expensive players doesn't constitute "swinging for the fences"; it constitutes reducing the number of chances you get to swing for the fences. With last year's contest parameters, it was obvious that 24-team rosters had a better chance to win than 18-team rosters--both a better chance to make the final 200, and a better chance to win once there. This year may be a little different, but I am certain that 18-team rosters will be similarly disadvantaged. The only question is whether 21, 25, or 30 is closest to optimal.
Well I think it's pretty easy to conclude that any injury to a stud player will impact the short roster teams more. Ditto the bye weeks. Larger rosters will have all the studs covered in some combination. A short roster would almost have to be perfect to win. Given all the possible outcomes, perfection is unlikely.
This is what we're debating. I don't believe this was so obvious. What we have is that in one single scenario (out of an infinite number), this turned out to be the case. Literally a few inches here or there in a game and a Stud scores an extra TD or a "value" player doesn't and the results could possibly change drastically.
This is really going off my point, but I'll debate it with you:
Maybe we have 2 different definitions for the team "swinging for the fences". But to me, it means taking a substantial risk for a huge reward with a low probability of success. Using that definition I'm not sure I'd say the larger rosters are swinging for the fences...
Smaller Rosters: Taking the risk that the studly team will a) not get injured (substantial risk), b) the great games where the studs perform studly will offset the games where the studs perform poorly and it's enough to carry them through each week (substantial risk). This, to me, and I believe in your thinking too, has a low probability of success. The huge reward is that in the small chance that they make it to the top 250, they have a stacked team and no longer need to worry about elimination. They believe this will give them a strong advantage.
Larger Rosters: Taking the risk that a "value" team will have enough players a) take advantage of injuries, or b) "break out". According to the numbers, 50% or so of the top players at each position weren't there last year. This doesn't not seem like a substantial risk to me that you're taking a bunch of guys who weren't in the top players last year who history has shown that some of them are likely to be in the top class this year. The reward is not as great either because even if you've hit some of those players, you're going up against a bunch of other large rosters who undoubtedly have many of the same players. This puts the chances of your winning strictly into the hands of one or two unique player combinations you have. This doesn't not seem like a "swing for the fences" gamble, but rather a I'll take the less risky route (which you agree it is saying that you think that larger roster players have a clear advantage) and hope that my uniqueness will win it for me.
Oh, it's definitely unlikely. But given 13,000 entries, it's essentially required.I should just put this in my sig: I'm not saying small/stud rosters are better. I'm just not completely sure they're not.
That's definitely true.But the expensive players are swinging for loftier fences. And it'll take about an 800-foot homer to win the thing.
Point 1: Agree 100% (injuries)Point 2: Agree 100% (byes)Point 3: Agree 100% (studs covered by larger rosters)Point 4: Agree 50% (short roster would need to have a perfect roster to win)Point 5: Agree 100% (perfection)My take on point 4: 1) Wouldn't really any roster need to be perfect to win? Large or small, you need to really nail it on the head to beat out 13000 other entries. 2) Looking past that, bc I don't really think that's what you were getting at. I agree 50% because you don't have to be perfect, but you have to be near perfect. You can't load up a team with 18 "studs" ( or whatever combination of stud/value guys you want to reach 18) and survive multiple injuries and weeks with multiple byes. The bye's we have near perfect knowledge on (trades) and can easily account for this. Injuries, I guess the jury is out on if we have any predictive capability on this ("injury prone players"), but I think even if we do, it's small. So yes, a small roster team would need to be nearly perfect to get to the Final 250. I say nearly because you have to account for the fact a team may be able to survive a significant injury, maybe even 2. I mean, they have many of the top pre-season players, and the thought isn't necesarily that they'll survive all injuries, but that they'll avoid significant injuries (out multiple week injuries) AND that the scoring potential of the other "studs" on their rosters can make up for poor weeks and injuries and carry a team. They probably need a bit more of luck to go their way in the elimination weeks, and run the risk of points wasted, but on the chance that they make it to the top 250, you'd have to say that they'll probably have the advantage of any other larger roster team. Does it mean they'll win? Of course not. Does it mean they'll finish high? Nope, they still need their players to perform like everyone else, but over an infinite scenarios, they figure that should they get to the top 250, they'll have a higher chance of winning than any other single larger roster team.
I went with a relatively small roster.
To win, you're going to need studly performance at each starting position. So let's just assume that my team and your team will both have studly perfomances at each starting position (because if they don't, our teams are irrelevant).
The question then is: which studs are likely to be more studly — my higher-priced studs, or your lower-priced studs?
We're both going to have to get very lucky to have a chance to win. But if I get very lucky with my high-priced studs, I think I'll outscore your team that gets very lucky with its low-priced studs.
(A big roster might last longer, on average, than a small roster; but busting in week 10 instead of week 4 isn't worth very much. I think small rosters are relatively more likely to bust out early; but also relatively less likely to bust out in the middle [since if they survive the beginning, it means they did land studs rather than busts], and relatively more likely to win the whole thing.)
It's a hard question, though, and I'm not very confident that I'm right.
To win, you're going to need studly performance at each starting position. So let's just assume that my team and your team will both have studly perfomances at each starting position (because if they don't, our teams are irrelevant).
The question then is: which studs are likely to be more studly — my higher-priced studs, or your lower-priced studs?
We're both going to have to get very lucky to have a chance to win. But if I get very lucky with my high-priced studs, I think I'll outscore your team that gets very lucky with its low-priced studs.
(A big roster might last longer, on average, than a small roster; but busting in week 10 instead of week 4 isn't worth very much. I think small rosters are relatively more likely to bust out early; but also relatively less likely to bust out in the middle [since if they survive the beginning, it means they did land studs rather than busts], and relatively more likely to win the whole thing.)
It's a hard question, though, and I'm not very confident that I'm right.
Last edited by a moderator:
Ignoratio Elenchi
Footballguy
I spend way too much time thinking about questions like these. I went into last year's contest with basically the same ideas you outlined here, and submitted a 20-player roster. I made the final 250, but I was very fortunate to do so, and I ended up coming in like 241th place during weeks 14-16. Over the course of the season and the offseason, I changed my mind. If we accept the premise that we both get "studly" performances at each position, I don't think a smaller roster of higher priced studs is likely to be more "studly" than a larger roster of less expensive players. I don't know if I agree that a smaller roster is relatively less likely to bust in the middle of the season (I'm curious why you think it is), and in general I think a larger roster has a better chance of surviving through week 13 than does a smaller roster (which is all that matters). I also don't know that a smaller roster of "preseason studs" provides any advantage in weeks 14-16, let alone enough of an advantage to offset the disadvantage from weeks 1-13.
Like you, though, I'm not extremely confident that I'm right, and am continually open to having my mind changed.
Last year was only one year, but which players appeared most frequently on teams that made it to the final weeks?
I'm sure that Chris Johnson was one. Who were the others — were they mostly high-priced guys or low-priced guys? (Maybe the answer has already been posted somewhere . . .)
I'm sure that Chris Johnson was one. Who were the others — were they mostly high-priced guys or low-priced guys? (Maybe the answer has already been posted somewhere . . .)
I edited my prior post with a very brief explanation. But essentially, I think the performance of small-roster teams will be more polarized than the performance of big-roster teams. Small-roster teams are more likely to be either very bad or very good. The very bad ones will be knocked out early. The ones that aren't knocked out early should, from that point forward, last longer than the big-roster teams that didn't get knocked out early.
Last edited by a moderator:
CalBear
Footballguy
That's certainly not what happened last year; small rosters got knocked out at a higher rate all year long, and in the finals.
Ignoratio Elenchi
Footballguy
I'm not sure where we're drawing the line between "early" and "later" but the byes range from weeks 4 through 10 - I'd think a smaller roster wouldn't be more likely to survive this stretch than a larger roster. If we're talking weeks 11-13 then maybe I could buy the argument that a smaller roster has a better chance of making the final 250 than does a larger roster, but I'm still not convinced of that, nor am I convinced that a smaller roster of more expensive players has any kind of advantage in weeks 14-16 (although I used to believe it would, and am still open to the idea that it does). I still think variance is too high, particularly at RB and WR, to go in with a small number of "proven" players. Even if you happen to correctly predict which of the expensive players will end up as RB1, RB2, WR1, WR2, and WR3 by the end of the season, I don't know if you could survive all 13 weeks with no real depth behind them.Does anyone still have a table of last year's available players and their prices? I thought maybe I did but I can't find it...
Ignoratio Elenchi
Footballguy
Regarding reverse engineering, here's a quick cut at the top players in weeks 14-16 last year (2009):QB:95.2 pts - Eli Manning (total points for all 3 weeks)89.1 - Donovan McNabb83.4 - Peyton Manning82.4 - Matt Moore81.1 - Matt Schaub79.2 - Joe Flacco78.2 - Philip Rivers78.1 - Kyle Orton76.6 - Aaron Rodgers75.8 - Drew Brees75.6 - Ben RoethlisbergerRB:80.4 - Chris Johnson72.9 - Adrian Peterson72.5 - Jerome Harrison65.6 - Jamaal Charles62.0 - Frank Gore58.8 - Ryan Grant58.8 - Ray Rice57.4 - Jonathan Stewart55.1 - Maurice Jones-DrewWR:89.3 - Andre Johnson85.5 - Brandon Marshall69.7 - DeSean Jackson64.1 - Steve Smith, CAR62.9 - Miles Austin59.1 - Vincent Jackson55.8 - Derrick Mason47.1 - Randy MossTE:73.5 - Dallas Clark57.3 - Brent Celek50.6 - Todd Heap49.6 - Antonio Gates49.4 - John Carlson47.4 - Fred DavisI bet the top-10 winning teams from last year's contest had some combination of the above players. Assuming you used 2 slots for kickers and 2 slots for D/STs, I bet an 18-roster team with the following would be among the top-10:QB - Eli Manning ($16), Matt Moore, Joe Flacco ($14) -- Note delete Moore because he was not a listed playerRB - Chris Johnson ($38), Jerome Harrison ($4), Jamaal Charles ($7), Ray Rice ($21)WR - Andre Johnson ($43), DeSean Jackson ($24), Miles Austin ($6), Vincent Jackson ($21), Derrick Mason ($9)TE - Dallas Clark ($18), Brent Celek ($7)D - pick 2K - pick 2Total cost = $228 (excluding kickers and defense)In Week 14, this team would have had 260.6 points (excluding K and D/ST):QB - 42.4 pts (Eli Manning)RB - 38.1 pts (Chris Johnson)RB - 29.9 pts (Ray Rice)WR - 42.6 pts (Andre Johnson)WR - 29.8 pts (DeSean Jackson)WR - 20.4 pts (Derrick Mason)TE - 29.8 pts (Dallas Clark)Flex - 27.6 pts (Jamaal Charles)In week 15, this team would have had 255.2 points (excl. D and K):QB - 40.2 pts (Joe Flacco)RB - 48.8 pts (Jerome Harrison)RB - 24.0 (Jamaal Charles)WR - 28.6 pts (Andre Johnson)WR - 27.8 pts (Vincent Jackson)WR - 26.9 pts (DeSean Jackson)TE - 32.0 pts (Dallas Clark)Flex - 26.9 pts (Miles Austin)In week 16, this team would have had 154.4 points (excl. D and K):QB - 18.4 pts (Eli Manning)RB - 25.4 pts (Chris Johnson)RB - 20.8 pts (Jerome Harrison)WR - 18.1 pts (Andre Johnson)WR - 16.9 pts (Miles Austin)WR - 14.7 pts (Derrick Mason)TE - 24.1 pts (Brent Celek)Flex - 16.0 pts (Ray Rice)Excluding points scored by kickers and defenses, this team scored 681.6 points in weeks 14 to 16. Each of the players were used at least one week (weeks 14-16). By further reverse engineering, you could pick the defenses and kickers to ensure that this team would make all the cuts for weeks 1 to 13 (this could be a problem). If this team survived to the final 250, I bet it would have placed at least in the money and hopefully in the top 10.ETA: Whoops! Matt Moore was not a listed player last year. I deleted him. The above team cost $228 (excluding kickers and defense).
Last edited by a moderator:
CalBear
Footballguy
Here are the dollar totals for those players (bolded are "cheap" players for their positions):
So it looks like half of the leaderboard at each position from weeks 14-16 came from cheap players. This is not an isolated result; it is absolutely the expected result given what we know about fantasy football. It's pretty obvious that buying Ben Roethlisberger and Eli Manning was a much better deal than buying Tom Brady or Drew Brees for about the same amount.I think the major flaw with the logic that you need studs to be able to produce in weeks 14-16 is that our definition of who is a stud changes radically over the course of the season. This year, half of the expensive players will not be on these lists, again.
LittlePhatty
Footballguy
I bet reverse-engineering a winning solution with 18 players would mean leaving $ on the table at the beginning of the contest. And who would do that?Let's say your choice 18-player roster of "after-the-season-is-done" cherry-picked guys adds up to $225.
Now imagine you're back at the beginning of the year (without knowing what the future holds) and you have those exact 18 players selected, adding up to $225.
Are you going to roll with those guys and leave $25 on the table? No.
You would either:
A) Select a few other cheap guys, increasing your roster size and defeating the 18 player theory.
or
B) use the extra money for an "upgrade" by un-checking Ray Rice and putting Matt Forte in his place. Ouch. Season over.
I think this is evidence against 18 player rosters. Even if you are successful finding 18 higher-costing players that turn out to be the perfect picks for this contest, there are additional odds against those 18 players adding up to be exactly $250.
Last edited by a moderator:
I agree with this. But this also means that half of the expensive players will be on the list again this year, and there is nothing to suggest that a person couldn't have put together a 18 player teams that had players from both high and low. And you bringing up Ben Roethlisberger leads me to a question that I don't think anyone has quite answered: How are we defining studs? Because even if we take it to the extreme and say that 18 man roster has 2 K/2Def and spend the min $4 on them, he still only has an average of $17.57 to spend on each of his remaining players. So I think we can argue that anyone over $20 (making an adjust for TE's being less costly) would have to be considered a "stud" for this contest. That's:6 out of top 11 QB's (2 were at $19 and 1 not available for purchase) were "studs"7 out of the top 9 RB's were "studs"6 out of the top 8 WR's were "studs"
