Fantasy Football 2015: Value Based Guessing

The Something Awful Forums > Discussion > Sports Argument Stadium > The Armchair Quarterback > Fantasy Football 2015: Value Based Guessing

«‹›411 »

coronaball: Feb 6, 2005; You're finished, pork-o-nazi!

fariz posted:

.5 PPR, 1QB, 2rb, 2wr, 1te, 1r/w
Lacy is the only tier 1 RB not being kept. After him there are guys like Foster / Hill / Anderson
WR options are Megatron, Dez, Julio, OBJ (pretty much every one really)

Why on earth is Lacy not being kept?

# ? Jun 11, 2015 21:35

Adbot: ADBOT LOVES YOU

# ? May 31, 2024 08:50

fariz: Nov 10, 2009; You either die a hero, or you live long enough to see yourself become the villain.

coronaball posted:

Why on earth is Lacy not being kept?

They're keeping Bell.

Hill could fall to my next pick but it'd be unlikely. If probably take him even if I had 2rbs already because in sick in 2005

# ? Jun 11, 2015 21:50

Franks Happy Place: Mar 15, 2011; It is by weed alone I set my mind in motion. It is by the dank of Sapho that thoughts acquire speed, the lips acquire stains, stains become a warning. It is by weed alone I set my mind in motion.

One dude in my keeper league is keeping Jamaal Charles, LeSean McCoy, and Adrian Peterson this year. :stare:

Glad I have a workable tandem of Jeremy Hill and Jonathan Stewart locked up as keepers, because those first two rounds are gonna be an RB wasteland.

# ? Jun 11, 2015 22:09

Leperflesh: May 17, 2007

Benne posted:

Patriots waived Tim Wright

Remember when he was supposed to be the next big star

He barely got to play. The team has Gronk, they're going to throw to Gronk, and Gronk got through the season unhurt. I'm sure there's going to be a fight for him on the waiver wire.

Meanwhile, the Pats dropped Garrett Gilbert, and picked up Matt Flynn.

I bet around the Flynn household they are all tired of people calling up Matt's answering machine to yell "wooo! In like Flynn! Hahaha GET IT?"

Leperflesh fucked around with this message at 23:46 on Jun 11, 2015

# ? Jun 11, 2015 23:44

Matt Zerella: Oct 7, 2002; Norris'es are back baby. It's good again. Awoouu (fox Howl)

The Giants should go after him IMO.

# ? Jun 12, 2015 01:17

Gyshall: Feb 24, 2009; Had a couple of drinks.
Saw a couple of things.

Sataere posted:

Any chance Hill falls to you next round if you grab Dez?

I've been getting Hill in the second round quite regularly, especially starting with Julio in the first.

# ? Jun 12, 2015 01:31

Franks Happy Place: Mar 15, 2011; It is by weed alone I set my mind in motion. It is by the dank of Sapho that thoughts acquire speed, the lips acquire stains, stains become a warning. It is by weed alone I set my mind in motion.

Gyshall posted:

I've been getting Hill in the second round quite regularly, especially starting with Julio in the first.

A Good Move. I can't believe Hill isn't a first rounder.

# ? Jun 12, 2015 01:42

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

So I've been thinking about whether there was any way to tell if "upside-down" drafting strategies (forgoing RBs in the first few rounds) was relatively more effective or less effective depending on your draft position. Is it the right move if you have a late pick? Can it work if you have an early pick?

Couldn't find anything online, so I decided to build a quick simulator. It takes a bunch of modular inputs (scoring, league settings, draft strategy for each player, valuations of positional need, rankings, projections etc) and simulates a large number of fantasy snake drafts.

In a qualitative sense, it's nice to get a feel for how different scenarios play out. For example, here's a drafts from a player at the 1/2 turn in 10-team standard:

code:

'Dez Bryant',
 'LeSean McCoy',
 'Kelvin Benjamin',
 'Mark Ingram',
 'Travis Kelce',
 'T.J. Yeldon',
 'Matthew Stafford',
 'Roddy White',
 'Torrey Smith',
 'Ryan Tannehill',
 'David Cobb',
 'Devante Parker',
 'Reggie Bush',
 'Larry Donnell'

and here's that same drafter's next draft:

code:

'Odell Beckham Jr.',
 'Dez Bryant',
 'Carlos Hyde',
 'Mark Ingram',
 'DeSean Jackson',
 'Cam Newton',
 'Martellus Bennett',
 'Rashad Jennings',
 'Larry Fitzgerald',
 'Eli Manning',
 'Duke Johnson',
 'Delanie Walker',
 'Hakeem Nicks',
 'Davante Adams'

(you'll notice I haven't bothered with defenses and kickers - I'm assuming you'll be drafting those last, and they're nearly impossible to predict from the preseason anyways)

But quantitatively, I've put together a script that calculates the projected score (using fantasypros averages) for drafted starters as an indicator of how successful the draft was for each player. For instance, here's the sum of the average projected points for starting players in a standard scoring 2RB/2WR/Flex league, over 500 simulations:

pre:

player      pts         std
1            1263.6    27.0
2            1266.7    27.3
3            1271.1    25.7
4            1272.6    25.8
5            1272.0    25.1
6            1260.6    24.2
7            1262.5    23.4
8            1262.2    23.2
9            1261.2    24.2
10          1266.8    22.5

Where "player" is the draft order, "pts" is the sum of the average points of starters by the end of the drafts, and std is standard deviation for pts across simulations.

You can see that picks 3-5 seems pretty advantageous, while picks 6-9 seem to be a bit sub-par. So I asked whether drafting a zero-rb strategy at these spots would improve their outcomes any.

I want to note here that this isn't a general rule: these expected outcomes are specific to the scoring, league settings, and strategies employed here. I had all these players essentially using fantasypros average rankings. Every time they were up for a pick, they would first "consider" their needs, and adjust the rank of all remaining available players by need. Then, they would identify the top four candidates from this combination of need and ranking, and finally, would make a "fuzzy" pick between those four top contenders (where each player ranking had a noise equivalent to the standard deviation of rankings for that player on fantasypros). This struck me as a good heuristic for how people might actually draft (go by rankings, adjust for need, consider the top 4 candidates, and pick their favorite).

So then, for each possible draft position, I simulated 500 additional drafts employing a zero-rb strategy from that position (leaving the other drafters with the default "need+ranking" strategy described above). The zero-rb strategy forced the player to pick non-rbs for the first two rounds, followed by a switch to the default (where they should dynamically "see" that they need RBs).

After those 5000 simulations, here's the difference between the baseline expected points (listed in the table above) and the zero-rb approach from each position:

pre:

1    -1.48
2      -0.42
3     2.15
4      -0.03
5      0.31
6     1.47
7     -2.86
8     0.45
9    -0.67
10    0.22

Positive values mean a zero-rb approach improved the expected points of their starters, negative values means the zero-rb approach was worse. Overall, changes were minimal - it seems like an intentional zero-rb approach for the first two rounds doesn't change overall starter strength on average.

Going zero-rb from 1.01, however, looks like potentially a bad idea. This is really suprising, because whatever they do at 2.10, they can immediately adapt and take whatever is needed at 3.01. It essentially just means that not taking an rb from the 1.01 in this circumstance is too much of a reach - you're passing up too much value. 7 gets dinged a surprisingly large amount for a zero-rb approach too. The biggest benefit from a zero-rb approach seems to come for spots 3 and 6 in this scenario.

So I poked around drafts from the 6 spot to get a feel for why. Here's what a "typical" draft looked like from six:

code:

'Eddie Lacy',
 'Jeremy Hill',
 'Ty Hilton',
 'Carlos Hyde',
 'Russell Wilson',
 'Amari Cooper',
 'Tevin Coleman',
 'Jeremy Maclin',
 'Julius Thomas',
 'Allen Robinson',
 'David Cobb',
 'John Brown',
 'Dwayne Allen',
 'Joe Flacco'

And here's what a sample zero-rb draft looked like from 6:

code:

'Dez Bryant',
 'Julio Jones',
 'Alfred Morris',
 'Carlos Hyde',
 'Latavius Murray',
 'Travis Kelce',
 'Vincent Jackson',
 'Matthew Stafford',
 'Brandon LaFell',
 'Darren McFadden',
 'Devonta Freeman',
 'Percy Harvin',
 'Sam Bradford',
 'Javorius Allen'

While you could always look at actual fantasy mocks for a better idea of how people are drafting, I think this simulator could be useful in asking questions about how particular strategies function over the long haul, including strategies from specific draft positions or strategies against specific opponents, or in custom league settings. I'm hoping to code value-based drafting next, followed by some other permutation (like value over next availabe or VONA drafting), but I'm open to suggestions.

(more to follow; scripts to be posted in the stats thread within a few days)

Forever_Peace fucked around with this message at 03:57 on Jun 12, 2015

# ? Jun 12, 2015 02:20

Beer4TheBeerGod: Aug 23, 2004; Exciting Lemon

LmaoTheKid posted:

Ride Lacy and Peterson to the championship IMO but I'm still a big believer in RB being the most consistent position in fantasy.

E: In my 3 years of winning championships I had absolute stud RBs (last year my idiot league went nuts with WRs and QBs so I managed to get Forte/Murray/Bell and I rode those ponies all the way without breaking a sweat).

You absolutely grab Lacy and laugh when you get to keep him for years to come.

# ? Jun 12, 2015 02:28

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

(continued)

I mentioned that I made everything super modular. Part of that was so that you folks could easily offer tweaks or suggestions for me to plug in if there's something you want to simulate.

For starters, scoring and league settings are easy. If you want to simulate custom leagues to answer a particular question, just change these values below and post it, and I'll try to run some simulations for you when I have a moment.

code:

pass_att, 0
pass_cmp, 0
pass_yds, 0.04
pass_tds, 4
pass_ints, -1
rush_att, 0
rush_yds, 0.1
rush_tds, 6
rec_att, 0
rec_yds, 0.1
rec_tds, 6
fumbles, -2
QBstart, 1
RBstart, 2
WRstart, 2
TEstart, 1
flexstart, 1
teams, 10
rounds, 15

Everything down through "fumbles" are scoring options (e.g. 0.04 per passing yard means 1 pt per 25 yards for QBs). The "start" settings are for the number of starters you have to play each week (flex is hardcoded as rb/wr/te - sorry! Not that it matters much though, the sim will almost always pick rbs and wrs for it). Teams and rounds are draft settings.

I also mentioned in the previous post that we can ask strategies to consider "need". The way I do that is by adjusting the "value" (e.g. VORP or rank) of potential picks based on the number of players of that position that I already have. That looks like this, if you'd like to tweak the values:

code:

QBval = [0.95, 0.6, 0]
RBval = [1, 1, 0.93, 0.85, 0.70, 0.65, 0.2, 0.1, 0]
WRval = [1, 1, 0.9, 0.85, 0.68, 0.60, 0.2, 0.1, 0]
TEval = [0.96, 0.58, 0]

So, for example, in the first round (before we have any players), we'd value RBs and WRs at full value, QBs at 95% value, and TEs at 96% value. If we take Gronk in the first round, everything else is still at 1st-round value, but now TEs are considered only at 58% value. If we later end up taking ASJ because he was too good to pass up in the late rounds, now we would value all other TEs at 0% and would take none of them. Likewise, the 3rd RB we take is valued at only 93% as the first two, and the 4th at 85% etc. Essentially, what we are doing is forgoing face value (e.g. VORP or rank) because we've already filled that need somewhat. QBs and TEs are on a much steeper curve than RBs and WRs. I made these numbers up, so feel free to make better ones. If everybody in your league drafts QBs late, you can dial back QBvals even further (though remember that they are already fairly low in the rankings on fantasypros - late QB is trendy).

And finally, you are also welcome to recommend a) different combinations of strategies across the drafters, or b) different strategic rules. I mentioned that I'm already pursuing VORP and VONA strategies, but if you want to try out some other weird and crazy thing (e.g. Beer's TE-TE approach from last year), describe exactly what you have in mind and I'll see what I can do.

Forever_Peace fucked around with this message at 02:41 on Jun 12, 2015

# ? Jun 12, 2015 02:37

SouthShoreSamurai: Apr 28, 2009; It is a tale,
Told by an idiot, full of sound and fury,
Signifying nothing.; Fun Shoe

You are an insane and beautiful man.

# ? Jun 12, 2015 03:18

Beer4TheBeerGod: Aug 23, 2004; Exciting Lemon

I love nerdchat.

# ? Jun 12, 2015 04:07

cheese: Jan 7, 2004; Shop around for doctors! Always fucking shop for doctors. Doctors are stupid assholes. And they get by because people are cowed by their mystical bullshit quality of being able to maintain a 3.0 GPA at some Guatemalan medical college for 3 semesters. Find one that makes sense.

Wow, really interesting work. What does it look like in a 12 player league with standard scoring?

# ? Jun 12, 2015 04:08

Spoeank: Jul 16, 2003; That's a nice set of 11 dynasty points there, it would be a shame if 3 rings were to happen with it

Goddamn that's some good stuff.

Also I'm doing my fifth mock draft in five nights even though it will be useless in doing any sort of valuation. Just gotta draft.

# ? Jun 12, 2015 05:19

Beer4TheBeerGod: Aug 23, 2004; Exciting Lemon

The takeaway I get from those values is that draft strategy and position are statistically insignificant.

# ? Jun 12, 2015 06:12

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

Beer4TheBeerGod posted:

The takeaway I get from those values is that draft strategy and position are statistically insignificant.

I agree with you about the particular situation as described, though this exercise isn't really about drawing generalizations (it's more about testing the consequences of a specific circumstance).

But there are plenty of circumstances where I could see draft strategy mattering a great deal. It actually could be the case that working from aggregate rankings and adjusting for need is a fairly optimal strategy for 10-team standard. But that's not always how people actually draft. For instance, here's the single solution that would occur in the same league if everybody took the highest ADP player each turn (after adjusting for need):

code:

1     1306.9
2     1283.0
3      1283.2
4     1197.5
5     1235.2
6      1240.9
7     1276.4
8     1280.8
9     1285.2
10    1257.7

team 1 makes out like a bandit, while team 4 sucks.

Team 1 in an all-ADP draft:

code:

'Jamaal Charles',
 'C.J. Anderson',
 'Jordy Nelson',
 'Russell Wilson',
 'Andre Johnson',
 'Vincent Jackson',
 'Martellus Bennett',
 'Darren McFadden',
 'Kevin White',
 'Knile Davis',
 'Allen Robinson',
 'Montee Ball',
 'Sam Bradford',
 'David Johnson'

Team 4 in an all-ADP draft:

code:

"Le'Veon Bell",
 'Calvin Johnson',
 'Randall Cobb',
 'Todd Gurley',
 'Drew Brees',
 'Amari Cooper',
 'C.J. Spiller',
 'Antonio Gates',
 'Isaiah Crowell',
 'Eric Decker',
 'Philip Rivers',
 'Cody Latimer',
 'Charles Sims',
 'Duke Johnson'

I guess your starting RBs are Spiller and Crowell for the first few weeks? They also have to rely on Brees and Gates (vs. Russell and Bennett on team 1).

But if everybody is going by ADP, here's what happens if ONLY team 4 uses the aggregate rankings instead (100 sims):

code:

1     1314.2
2     1263.9
3     1237.1
4     1265.7
5      1232.2
6     1252.9
7     1284.8
8     1266.5
9       1283.1
10    1260.1

In a 12-week regular season, team 4 just increased their expected performance by over 5.5 points per week.

Here's a sample team from drafter 4 in this case - we expect to see lots of folks with a higher rank than ADP, capitalizing on available value that other teams are passing by.

code:

'Marshawn Lynch',
 'Calvin Johnson',
 'Randall Cobb',
 'Lamar Miller',
 'Latavius Murray',
 'Ben Roethlisberger',
 'Tevin Coleman',
 'Jarvis Landry',
 'Zach Ertz',
 'Brandon LaFell',
 'Philip Rivers',
 'Charles Sims',
 'Anquan Boldin',
 'Delanie Walker'

And here's one where they fell into an upside-down draft that looks pretty good, because they were able to nab Stewart and Miller that were being undervalued by ADP:

code:

'Antonio Brown',
 'Calvin Johnson',
 'Randall Cobb',
 'Jonathan Stewart',
 'Lamar Miller',
 'T.J. Yeldon',
 'Cam Newton',
 'Isaiah Crowell',
 'Michael Floyd',
 'Jason Witten',
 'Philip Rivers',
 'Anquan Boldin',
 'Delanie Walker',
 'Duke Johnson'

In a 12-team league, the "exact ADP draft" solution looks like this:

code:

1     1228.5
2      1216.7
3     1212.5
4     1217.7
5     1193.6
6     1207.8
7         1238
8     1199.2
9     1230.6
10     1223.9
11    1250.4
12    1186.8

Team 12 is the odd-one-out here. But by using rank instead of ADP (100 sims):

code:

1     1224.6
2     1208.7
3     1217.1
4     1214.7
5     1210.8
6     1223.1
7     1194.4
8     1226.5
9     1229.0
10    1211.4
11    1234.9
12    1235.2

they improved their team by about 4 expected points per week.

I'm looking forward to seeing how the VORP and VONA strats do. Aggregate rankings already perform pretty good, but they don't translate well to weird scoring rules. It'll be interesting to see how much advantage you can get by customizing your own rankings using projections if everybody else is still using the standard rankings.

edit: ADP values in this case come from averaging together fantasyfootballcalculator mock ADPs with fantasypros mock ADPs, then removing everybody without a listed ADP (because they tend to go undrafted more than not at those two sites).

Forever_Peace fucked around with this message at 14:45 on Jun 12, 2015

# ? Jun 12, 2015 14:18

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

cheese posted:

Wow, really interesting work. What does it look like in a 12 player league with standard scoring?

Here's what happens if everybody uses aggregate rankings (adjusting for need, making fuzzy picks between the top 4 candidates each turn) in a 12-team league (500 sims):

code:

1     1211.0
2     1211.6
3     1212.7
4     1211.6
5     1210.6
6     1220.5
7     1221.7
8     1228.5
9     1230.1
10    1223.7
11    1229.8
12    1226.5

Looks like if everybody is equally "good" at drafting, picks 1-5 are marginally worse (by about an expected point per week) than picks 6-12, with a peak at 8/9.

And here's a 14-team league with everybody using that same strategy (500 sims):

code:

1     1169.8
2     1171.3
3     1177.3
4     1170.5
5     1174.8
6     1177.6
7     1176.6
8     1181.2
9     1182.3
10    1186.6
11    1181.3
12    1182.5
13    1186.2
14    1179.6

Looks like a minor benefit to being in spots 8-13, but it's not as pronounced as in 12-team.

# ? Jun 12, 2015 14:40

Gyshall: Feb 24, 2009; Had a couple of drinks.
Saw a couple of things.

Great stuff.

# ? Jun 12, 2015 14:55

Varg: Jan 13, 2007; A friendly face.

that's a lot of words and numbers that end up meaning nothing as soon as the draft actually starts and you just go by instinct after thinking about the perfect draft strategy for the last 6 months. fwiw I've had the 1st spot once and 2nd spot twice in the past 3 years in 10 teams and have won all those years. I really don't know if draft position means all that much when you have a mix of people who know and don't really know what they're doing.

Varg fucked around with this message at 16:27 on Jun 12, 2015

# ? Jun 12, 2015 16:24

cheese: Jan 7, 2004; Shop around for doctors! Always fucking shop for doctors. Doctors are stupid assholes. And they get by because people are cowed by their mystical bullshit quality of being able to maintain a 3.0 GPA at some Guatemalan medical college for 3 semesters. Find one that makes sense.

Forever_Peace posted:

Here's what happens if everybody uses aggregate rankings (adjusting for need, making fuzzy picks between the top 4 candidates each turn) in a 12-team league (500 sims):

So if I understand this right, the "adjusting for need" thing is basically that as a person's team fills up, they start giving more value to spots that are unfilled? I.E. If I go RB-WR-RB in my first 3 picks, my TE "need" goes up because I have not filled it? So in that case, maybe I draft Travis Kelce instead of Lamar Miller, even though Miller has a slightly higher ADP, because my "need" for TE pushes Kelce up a few spots?

# ? Jun 12, 2015 16:25

Fight Club Sandwich: Apr 29, 2006; you want a piece of me???

Varg posted:

that's a lot of words and numbers that end up meaning nothing as soon as the draft actually starts and you just go by instinct after thinking about the perfect draft strategy for the last 6 months. fwiw I've had the 1st spot once and 2nd spot twice in the past 3 years in 10 teams and have won all those years. I really don't know if draft position means all that much when you have a mix of people who know and don't really know what they're doing.

Fantasy Football 2015: that's a lot of words and numbers that end up meaning nothing

Instincts are cool and I'm not a fan of any method that uses projected points as a barometer for success

# ? Jun 12, 2015 16:31

sourdough: Apr 30, 2012

Fight Club Sandwich posted:

Fantasy Football 2015: that's a lot of words and numbers that end up meaning nothing

Instincts are cool and I'm not a fan of any method that uses projected points as a barometer for success

Agreed. My instincts say pick the funniest names and hell I won 69% of the leagues I played in last year.

# ? Jun 12, 2015 16:35

Gyshall: Feb 24, 2009; Had a couple of drinks.
Saw a couple of things.

fantasy football is 25% analytics/stats, 25% knowing football/football IQ/knowing which guys actually suck, and 50% luck

# ? Jun 12, 2015 16:41

Franks Happy Place: Mar 15, 2011; It is by weed alone I set my mind in motion. It is by the dank of Sapho that thoughts acquire speed, the lips acquire stains, stains become a warning. It is by weed alone I set my mind in motion.

Gyshall posted:

fantasy football is 25% analytics/stats, 25% knowing football/football IQ/knowing which guys actually suck, and 50% luck

In-season roster management is way more important than drafting, too. Even the best drafted team will fall apart if you don't make the right moves.

# ? Jun 12, 2015 16:44

cheese: Jan 7, 2004; Shop around for doctors! Always fucking shop for doctors. Doctors are stupid assholes. And they get by because people are cowed by their mystical bullshit quality of being able to maintain a 3.0 GPA at some Guatemalan medical college for 3 semesters. Find one that makes sense.

Franks Happy Place posted:

In-season roster management is way more important than drafting, too. Even the best drafted team will fall apart if you don't make the right moves.

If your end of the season roster looks just like your draft roster, you hosed up because it means you didn't 1) make good waiver wire pickups and 2) didn't make any advantageous trades.

Varg posted:

that's a lot of words and numbers that end up meaning nothing as soon as the draft actually starts and you just go by instinct after thinking about the perfect draft strategy for the last 6 months. fwiw I've had the 1st spot once and 2nd spot twice in the past 3 years in 10 teams and have won all those years. I really don't know if draft position means all that much when you have a mix of people who know and don't really know what they're doing.

All the stats go out the window when your buddies coworker that you got at the last minute to make it 12 decides to draft only 49er's "because they are going to win the super bowl this year".

# ? Jun 12, 2015 16:47

Beer4TheBeerGod: Aug 23, 2004; Exciting Lemon

Fight Club Sandwich posted:

Fantasy Football 2015: that's a lot of words and numbers that end up meaning nothing

Instincts are cool and I'm not a fan of any method that uses projected points as a barometer for success

Projections are notoriously terrible. 40% of the time you're wrong 100% of the time. IMO the benefit of projections lies within the aggregate. How much better is player A expected to be relative to player B? How many people really think this guy is a stud? Which position is deep and which position is scarce?

Stuff like that is what helps me.

# ? Jun 12, 2015 16:51

Sataere: Jul 20, 2005; Step 1: Start fight
Step 2: Attack straw man
Step 3: REPEAT

Do not engage with me

I think the way to really test this out would be to use last years projections against that draft-based simulator, and then measure the results that actually occurred.

# ? Jun 12, 2015 16:53

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

cheese posted:

So if I understand this right, the "adjusting for need" thing is basically that as a person's team fills up, they start giving more value to spots that are unfilled? I.E. If I go RB-WR-RB in my first 3 picks, my TE "need" goes up because I have not filled it? So in that case, maybe I draft Travis Kelce instead of Lamar Miller, even though Miller has a slightly higher ADP, because my "need" for TE pushes Kelce up a few spots?

Basically yeah, and you can set how steep the valuation dropoff is for each position. Using the values I posted earlier:

code:

QBval = [0.95, 0.6, 0]
RBval = [1, 1, 0.93, 0.85, 0.70, 0.65, 0.2, 0.1, 0]
WRval = [1, 1, 0.9, 0.85, 0.68, 0.60, 0.2, 0.1, 0]
TEval = [0.96, 0.58, 0]

If you've already drafted 2rbs and 1wr, and you're deciding between Miller and Kelce, we'd reduce Miller's value to 93% of what we actually expect it to be (because he'd be the 3rd RB and 0.93 is the 3rd RB multiplier value in our settings) and Kelce's value to 96% of what we actually expect it to be. These are pretty small adjustments at this point near the beginning of the draft (on purpose, because we don't want to pass up on value when "need" hasn't become critical yet), so if Miller starts with the higher value, he'd probably still have the higher value after the adjustment.

But it makes a big difference after the first 5-6 rounds. If we've gone 3WR and 3RB, the next pick has a better chance to be a QB or a TE unless we get passed something pretty juicy (we'd be willing to take up to roughly a 10% reduction in perceived value in order to fill a need).

I also have it set to make a big difference if you go WR-WR in the first two rounds. You're really likely to draft 2-3 RBs from there before your next WR.

quote:

Instincts are cool and I'm not a fan of any method that uses projected points as a barometer for success

Well, I happen to have historical data for ADP and projected points etc., I think. If we assume ADP reflects a greater share of "instinct" than projected points (in a wisdom of the crowds sense), we could probably just go see if it's superior over the course of the season.

quote:

that's a lot of words and numbers that end up meaning nothing as soon as the draft actually starts and you just go by instinct after thinking about the perfect draft strategy for the last 6 months. fwiw I've had the 1st spot once and 2nd spot twice in the past 3 years in 10 teams and have won all those years. I really don't know if draft position means all that much when you have a mix of people who know and don't really know what they're doing.

FWIW I have no idea what draft spots were good in previous years because I haven't simulated them (and I don't think it would actually contribute much because we could theoretically just look at which draft position actually won the most leagues, because we have the benefit of hindsight). Again, I'm not making generalizations to all drafts on purpose - I'm seeing what happens given particular circumstances of league settings, scoring settings, and draft strategies at different starting positions.

Though you bring up a good point about there usually being a good mix of opponents who know what they are doing and opponents who don't. Fortunately, I could probably simulate that uncertainty if you like! (e.g. each drafter samples from a pool of draft strategies at the start of each draft)

# ? Jun 12, 2015 16:53

Matt Zerella: Oct 7, 2002; Norris'es are back baby. It's good again. Awoouu (fox Howl)

cheese posted:

If your end of the season roster looks just like your draft roster, you hosed up because it means you didn't 1) make good waiver wire pickups and 2) didn't make any advantageous trades.

Or you drafted correctly to begin with :colbert:

But generally my last 6 picks are gone by week 6 due to add/drops.

# ? Jun 12, 2015 17:29

Spoeank: Jul 16, 2003; That's a nice set of 11 dynasty points there, it would be a shame if 3 rings were to happen with it

LmaoTheKid posted:

Or you drafted correctly to begin with

But generally my last 6 picks are gone by week 6 due to add/drops.

If you ain't churnin, you ain't tryin.

# ? Jun 12, 2015 18:45

Matt Zerella: Oct 7, 2002; Norris'es are back baby. It's good again. Awoouu (fox Howl)

Spoeank posted:

If you ain't churnin, you ain't tryin.

Too much churnin means you need some learnin

# ? Jun 12, 2015 19:19

Forever_Peace: May 7, 2007; Shoe do do do do do do do Shoe do do do do do do yeah Shoe do do do do do do do Shoe do do do do do do yeah

Finished a version of a value-based drafting strategy. It finds the need-adjusted top6 ranked players available and always takes the one with the highest value over replacement (VORP).

Spot the VORP drafter in this 16-team league:

code:

1     1152.362
2     1154.197
3     1170.378
4     1140.524
5     1139.398
6     1134.015
7     1125.697
8     1127.011
9     1129.141
10    1123.864
11    1120.862
12     1112.22
13    1122.029
14    1123.066
15    1123.719
16    1143.496

It almost invariably took 4 RBs in the first 5 rounds (even though only the first 3 "counted" as potential starters in a 2RB/2WR/flex, so team 3 actually had really great RB depth too that isn't reflected above). I'd say this definitely illustrates why fantasypros rankings don't necessarily scale up to bigger leagues, and you can get quite and edge by adapting your strategy if other people don't.

A typical VORP team for player 3 here in this 16-team league:

code:

'Jamaal Charles',
 'Mark Ingram',
 'Lamar Miller',
 'Vincent Jackson',
 'Giovani Bernard',
 'Allen Robinson',
 'Anquan Boldin',
 'Colin Kaepernick',
 'Larry Donnell',
 'Malcom Floyd',
 'Coby Fleener',
 'Jaelen Strong',
 'Daniel Herron',
 'Andrew Hawkins'

Coming up with what defines a "replacement-level player" was actually kind of fun. I ended up embedding a little scrip that does a few quick mock drafts ahead of the actual simulation with the given settings to get an estimate for "what kind of player might still be available in the 11th round and below in this circumstance" (where 11+ round players tend to be pretty interchangeable with "free" players from the WW in a 15-round draft). Then for the draft, it defined VORP as "advantage over estimated replacement-level player".

This is sort of the classic definition of VORP, and I'm not sure how much I like it. I might change the replacement-player cutoff to round 6 to see how that changes things (sort of like "replacement-level starter", ie the last few rounds where you'd get an rb/wr comparable to what other folks are starting).

# ? Jun 12, 2015 20:03

VDay: Jul 2, 2003; I'm Pacman Jones!

Figuring out who you should drop early and who to hold on to all season is one of the more important and underrated fantasy skills imo. Way too many people get attached to their drafted players and have them sit on the bench for 9 weeks not doing anything before finally dropping them for a second defense or bye-week kicker. It can definitely bite you in the rear end sometimes, but I've found that being super aggressive with drops/waiver pickups pays off like 80% of the time, and I end up getting a league-winning player from the waiver wire almost every year that I win.

# ? Jun 12, 2015 20:06

pangstrom: Jan 25, 2003; Wedge Regret

VDay posted:

Figuring out who you should drop early and who to hold on to all season is one of the more important and underrated fantasy skills imo. Way too many people get attached to their drafted players and have them sit on the bench for 9 weeks not doing anything before finally dropping them for a second defense or bye-week kicker. It can definitely bite you in the rear end sometimes, but I've found that being super aggressive with drops/waiver pickups pays off like 80% of the time, and I end up getting a league-winning player from the waiver wire almost every year that I win.

Yeah in the league I play in every year this is one of the more stable "attribute" spectrums. There are guys who cut bait fast and those who wait it out and everywhere inbetween, and usually the former don't spring for speculative plays/flavors of the week and the latter do. I am more aggressive than most but to this day the bad moves I remember most are the ones where I cut bait on established people on a bad run who then get back on track for someone else.

# ? Jun 12, 2015 20:12

Silly Burrito: Nov 27, 2007; SET A COURSE FOR
THE FLAVOR QUADRANT

VDay posted:

Figuring out who you should drop early and who to hold on to all season is one of the more important and underrated fantasy skills imo. Way too many people get attached to their drafted players and have them sit on the bench for 9 weeks not doing anything before finally dropping them for a second defense or bye-week kicker. It can definitely bite you in the rear end sometimes, but I've found that being super aggressive with drops/waiver pickups pays off like 80% of the time, and I end up getting a league-winning player from the waiver wire almost every year that I win.

I'm looking directly at you, Cordarrelle Patterson.

# ? Jun 12, 2015 20:19

MacheteZombie: Feb 4, 2007

Silly Burrito posted:

I'm looking directly at you, Cordarrelle Patterson.

My favorite bust of the season.

# ? Jun 12, 2015 20:26

Metapod: Mar 18, 2012

Silly Burrito posted:

I'm looking directly at you, Percy Harvin.

# ? Jun 12, 2015 20:59

Spoeank: Jul 16, 2003; That's a nice set of 11 dynasty points there, it would be a shame if 3 rings were to happen with it

Like half of y'all lit a roster spot on fire for 2014 Josh Gordon :lol:

That's an all-time fantasy football waste of a roster spot

# ? Jun 12, 2015 21:01

VDay: Jul 2, 2003; I'm Pacman Jones!

Silly Burrito posted:

I'm looking directly at you, Keenan Allen.

# ? Jun 12, 2015 21:12

Adbot: ADBOT LOVES YOU

# ? May 31, 2024 08:50

Varg: Jan 13, 2007; A friendly face.

I drafted and held onto Jeremy Hill all last year apart from 1 week when I dropped him out of frustration and then picked him back up the next week before he became super useful

basically if you're going to hold onto a player, make sure it's an RB that has a shot

# ? Jun 12, 2015 21:30

The Something Awful Forums > Discussion > Sports Argument Stadium > The Armchair Quarterback > Fantasy Football 2015: Value Based Guessing

«‹›411 »