In post 1064, PokerFace wrote:Ok Yahoo fantasy projections say Manziel should outscore Stafford. I'd like to revisit my earlier argument that Yahoo's projections are the worst fantasy projections and state this as further evidence
I agree, absolutely. They rely way too much on past results as an indication of future success. Stafford did poorly the first half of the season with a hardish schedule, so he MUST do poorly against an easy schedule the rest of the way. Giovani Bernard has been awful lately and is clearly at the bottom of a two-man committee, but he did really great earlier in the season and was projected as an RB1, so he MUST get to at least 7-8 points.
Wrong and wrong.
I imagine it's simply cheaper to just have a computer algorithm spit out numbers than to pay someone to meticulously examine each player.
Yahoo pay several people to do that. Like all the major fantasy sites, they have a bunch of people putting out rankings each week and to do that they have to go through the vast majority of, if not all starters. I suspect they have a model to do it too but I doubt it has anything to do with the fact that it would be hard for them to get a person to do it and more to do with the fact that it is far more mathematically sound to employ an approach that is entirely objective and logically consistent from week to week.
To borrow an example from Rob, he thought Kerwynn Williams could easily score 2 or 15 points. How do you put a number on that? How much do you factor in the fact that the Rams D hasn't allowed a touchdown for the previous two weeks (and still hasn't)? What about the Cardinals saying it was still a committee - do you trust the touches? You can go on and on like this. My guess is that the model starts with relatively few pieces of information, such as the player's scoring history, the strength of the opponent's run/pass D (or offence if projecting D/STs) and gives you an expected average point total based on those. Someone will probably manually adjust aspects like playing time in cases where the model doesn't know something, like a player being named the starter or being hurt. It's possible they run the model thousands of times with Williams getting the start in some and being the backup in others, then get an average value from all those runs. 6.90 was a reasonable estimate based on the available information and was not too far off the actual score.
Manziel being ranked above Stafford is probably based on three things: the model thinks the Vikings D/ST is actually pretty good (if Yahoo's matchup strength ratings are reflective of what the model uses, which is also stupid but a different argument); the model knows that Stafford's season average is not very good; and the model knows that in a very limited sample of playing time, Manziel actually put up a relatively large amount of fantasy points because he ran for a TD. This is probably one of those rare scenarios in which a new starter will actually have what looks like quite an aggressive projection because there is at least some evidence of Manziel being able to score points and there is a high level of confidence in the workload. The variance will be huge for a player with so little history.
Generally speaking I suspect the model is slow to adjust to breakout stars because it needs a certain sample size of scoring history to project a high score, whereas it's working from a big sample with established players and isn't going to drop them way down the rankings just because they have 2 or 3 bad weeks; it will be an incremental change. The model probably had Bernard relatively highly projected to start the season and he has five double-digit games, including two over 20 points, so even though he's had 3 disappointing weeks, the model knows that he is still getting touches and won't completely discount his history. Us fantasy players, on the other hand, are often perfectly happy to declare a player a star or drop them in a flash after just a couple of weeks of good/bad performance.
On an entirely different note, mith's new ideas sound like they would be great fun to try.
If ya smell what The Locke is cookin'!
"Locke Lamora and Andrius, defying all logic since 1081."
I think they actually got Kerwynn Williams right for their projections. 7pts was a fair projection. He basically hit the projection - he ended with 8.60. Which is a fair result, and one that keeps me in the running. I would have liked to see a Rams TD, though - the total from the Arizona defense isn't awful for me, but it doesn't help.
They should have gone for it in the fourth at the goal line on fourth down. Seems pretty unlikely that they'd have had another good shot at a touchdown given the way the game went.
If ya smell what The Locke is cookin'!
"Locke Lamora and Andrius, defying all logic since 1081."
Worst case scenario, you give the other team the ball on their 1 yard line and have a very realistic chance at a safety. Not a bad deal, with huge potential upside.
In post 1052, reinoe wrote:I have no hope of moving on anyway. I took a pass on Robert Morris and kept Eddie royal way past his exp date. I had holes in my team but didn't fix them until it was too late.
Everyone on my team has underperformed except three players. This is including evry bench player. Losing to wraith is my destiny and I'm glad he decided to go the extra mile. Wraith wore a condom, gave me a pillow to bite, and used lube.
Scum typically need to fabricate reasons for scumreads...
And of course my team overperforms for only the second time this season now when i am in consolation bracket
When I joined this site, I was a software tester for mobile business applications and the song PokerFace was not yet written by Lady Gaga
Now I test lottery and gambling software as my job. It's funny how my life has turned out. Somewhere a Time Traveler is laughing madly
This was definitely not the week for everyone to underperform
jdodge1019: hasjghsalghsakljghs is from vermont
jdodge1019: vermont is made of liberal freaks and cows
jdodge1019: he's not a liberal
jdodge1019: thus he is a cow
reinoe (4) is down 76.50-101.50 vs. Wraith (1). Reinoe has Cutler remaining and needs him to score more than 25 points, as Wraith's players have all played.
Flameaxe (2) is up 94.58-57.90 vs. gamsimbre (3). Flameaxe's team is done playing, so gamsimbre is down 36.68 points with Kenny Stills remaining for Monday.
My QB situation this year was mediocre to sometimes below and my WR situation was sub par as well this year
At start of year I had choice between Foles and Cutler in the draft. I grabbed foles and reinoe grabbed Cutler on the next pick.
Kinda wish I had got cutler so that would have been above mediocre but don't know how else i would have done
Still want winner of Flameaxe and Gamesbre to win it all
As seed 8 I am likly to be jahudo as I am up by 40 with some people on my team yet to play. Ingram won't be able to get 40-60 points
Titus or Juls will face me for 5th
When I joined this site, I was a software tester for mobile business applications and the song PokerFace was not yet written by Lady Gaga
Now I test lottery and gambling software as my job. It's funny how my life has turned out. Somewhere a Time Traveler is laughing madly
Tally voodoo clearly trumped Cow's good fortune to this point. Looks like you might get a big advantage if the pessimistic reports from the team on Murray are true, KMD.
If ya smell what The Locke is cookin'!
"Locke Lamora and Andrius, defying all logic since 1081."