Poly IDP 2012 Draft Thread: No one can replace Ed Hochuli

[Ben Kenobi]

Ben, you're assuming that FF scoring is in a normal distribution. I highly doubt that, given you're the lowest scorer and you're only 1 SD away by your calculation.

Run the numbers yourself. It just looks odd because I have only one win. Slowwy has 5 with the exact same team. Win percentage is very sensitive to how many standard deviations you are away from the mean.

[-Jrabbit]

I like the "if every team played every other team every week" metric. That seems quite indisputably equitable, especially once the bye weeks are done.

[snoopy369]

Run the numbers yourself. It just looks odd because I have only one win. Slowwy has 5 with the exact same team. Win percentage is very sensitive to how many standard deviations you are away from the mean.

You are displaying an incredible lack of understanding of basic statistics here, Ben. 'Standard Deviation' isn't some magic wand you can wave at things... and FF scoring is definitely not a normal distribution (it's probably closer to a uniform distribution, if anything). To predict a record from a score, you'd have to run a regression analysis with a lot more than 12 data points.

[snoopy369]

I like the "if every team played every other team every week" metric. That seems quite indisputably equitable, especially once the bye weeks are done.

I like it also. Not perfect, but certainly better than most ways of looking at it.

[SlowwHand]

It's still just a snapshot in time. It's all interesting in it's synopsis, but won't mean anything this week.

[snoopy369]

I graphed my four leagues normalized to DST scoring scale (not the particulars, but such that they had about the same average score per team). The visual regression line is something about W=(8/500)*(S-1850), though of course the scores are really all over the place (and definitely closer to a uniform distribution than a normal distribution). So my D/ST team (2213 points) for example deserves about 6 wins, not 8. There are some extremes (Albie with 7 wins in Auction and deserves less than 2 stands out, and in one of my ESPN leagues a guy with over 2700 adjusted points should have won 14 games in a 12 game season - but he is 11-1 so nto that far off!) but overall it's not bad (though the predictive value is probably very low).

I refined it a bit more, and normalized everyone to a per-game score (so /12), and then a 100 average per game score. For D/ST and Auction I multiplied by a factor of (100/180.65) to normalize; for my two ESPN leagues it was closer to a factor of one (100/95 and 100/90). The regression line I drew there ended up being (0.33)*(S-82), which is pretty similar to the original line [8/25 or so slope].

I'm curious how close this works for IDP, if someone wants to do the math. For my four leagues, it ends up very close - only off by -1.5 wins total over 42 data points - though I'm not saying anything about its predictive capability, as that's pretty low I'm sure.

[SlowwHand]

I need a Rosetta Stone course for math.

[Ben Kenobi]

I'm curious how close this works for IDP, if someone wants to do the math. For my four leagues, it ends up very close - only off by -1.5 wins total over 42 data points - though I'm not saying anything about its predictive capability, as that's pretty low I'm sure.

Interesting, I'll give that a try.

[Ben Kenobi]

You are displaying an incredible lack of understanding of basic statistics here, Ben. 'Standard Deviation' isn't some magic wand you can wave at things... and FF scoring is definitely not a normal distribution (it's probably closer to a uniform distribution, if anything). To predict a record from a score, you'd have to run a regression analysis with a lot more than 12 data points.

I'm aware of that. Small samples will see larger distortions than a larger sample. I'm just saying, that I have an equally strong team as Slowwhand and 4 fewer wins. My team is better than it's record by about 3 wins or so, not taking any closer look than a quick jot off the back of an envelope here and eyeballing stuff.

[snoopy369]

Just as easy to say Sloww's team is worse than its record by 3 wins or so. I don't know IDP very well so couldn't say exactly, but it seems like you're the lowest scoring team in the league, and with Sparrowhawk's list you're even worse than that by a large margin (perhaps you had a week where you scored a huge amount?). The vs-every-team-every-week metric is pretty solid in terms of identifying how well a team would do if it didn't have bad luck facing harder teams (or good luck facing easier ones), and your team is leaps and bounds below everyone else by that metric (assuming Sparrowhawk's math is right). 22-86 is about a .200 winning percentage, so after 12 games you should be around 2 or 3 wins (2-10 or 3-9).

[Ben Kenobi]

Just as easy to say Sloww's team is worse than its record by 3 wins or so. I don't know IDP very well so couldn't say exactly, but it seems like you're the lowest scoring team in the league, and with Sparrowhawk's list you're even worse than that by a large margin (perhaps you had a week where you scored a huge amount?). The vs-every-team-every-week metric is pretty solid in terms of identifying how well a team would do if it didn't have bad luck facing harder teams (or good luck facing easier ones), and your team is leaps and bounds below everyone else by that metric (assuming Sparrowhawk's math is right). 22-86 is about a .200 winning percentage, so after 12 games you should be around 2 or 3 wins (2-10 or 3-9).

I have the lowest variance of all the IDP teams. The team is built, deliberately for 'low variance', it never has very poor weeks, nor does it have great weeks. It consistantly puts up about 130 points or so. I have the highest lowest score and the lowest highest score. The team is 1-11, despite never being blown out, not even once.

138.69 (win)
146.46 (loss) Fitz with 1 point, Turner with 1 point.
131.87 (loss) Rivers with 15 points.
153.94 (loss) Turner with 3 points,
145.91 (loss) Butler with 1 point.
131.94 (loss) Rivers with 18 points.
112.23 (loss) Flacco with 14 points. Vernon Davis with 1.
110.30 (loss) Started Bradford who was my only bright spot. 5 starting players all under 10 points.
120.97 (loss) This was my almost-perfect management, the other way. Sat Rivers, Jackson, Fitz and Nate Washington.
115.80 (loss) Another 'everybody sucks' week. Johnson, Jackson were benched, and nobody but Rice did anything.
141.57 (loss) Better week, same result. Fitz with 1 point. I lose a heartbreaker by 1.
146.12 (loss) Vernon Davis with a 0 costs me a win.

So there you go. 6 times I had enough points to win (140+) points, went 1-5.

[SlowwHand]

And once again I get sucked into a conversation. Why am I so frequently the standard-bearer? Pick on someone else. I'm not exactly happy with how things are going either.

[Ben Kenobi]

Losses by: 1.5, 5, 9, 13, 12, 15, 15, 29, 33, 43, 41.

Out of my 11 losses, only 4 were bad losses, and I've never been the biggest blowout of the week. You would think that if I had the truly worst team that I would get 'biggest blowout' at least once.

[SlowwHand]

Bad fortune, it looks like to me.

[-Jrabbit]

I have the lowest variance of all the IDP teams. The team is built, deliberately for 'low variance', it never has very poor weeks, nor does it have great weeks. It consistantly puts up about 130 points or so. I have the highest lowest score and the lowest highest score. The team is 1-11, despite never being blown out, not even once.

138.69 (win)
146.46 (loss) Fitz with 1 point, Turner with 1 point.
131.87 (loss) Rivers with 15 points.
153.94 (loss) Turner with 3 points,
145.91 (loss) Butler with 1 point.
131.94 (loss) Rivers with 18 points.
112.23 (loss) Flacco with 14 points. Vernon Davis with 1.
110.30 (loss) Started Bradford who was my only bright spot. 5 starting players all under 10 points.
120.97 (loss) This was my almost-perfect management, the other way. Sat Rivers, Jackson, Fitz and Nate Washington.
115.80 (loss) Another 'everybody sucks' week. Johnson, Jackson were benched, and nobody but Rice did anything.
141.57 (loss) Better week, same result. Fitz with 1 point. I lose a heartbreaker by 1.
146.12 (loss) Vernon Davis with a 0 costs me a win.

So there you go. 7 times I had enough points to win (140) points, went 1-6.

If this was a deliberate strategy on your part (a highly dubious assertion IMO), congratulations. You have achieved exactly teh success you deserve. Even if it was possible to draft a team with this intention, I've got to point out that successfully doing so is basically aspiring to mediocrity. WHY WOULD YOU LIMIT YOUR UPSIDE???

And while I don't have time to debunk all of this crap, I'll point out a few assertions that are total bull****:

1. Claiming that you have the league's "highest low score" is a lie. You list your low score as 110.30. My low score to date is 136.45. My guess is there are a bunch of other teams with a higher worst scoring week as well.

2. Claiming to have never been blown out. Really? While you've never faced a truly high-scoring team (174 is the best you've faced), I see one loss by 30+, and three more by 40+.

3. Claiming that 140 points (which you've only managed 7 out of 12 times) is "enough to win." WTF, dude? More like "enough to hope your opponent sucks."
Number of teams scoring under 140 pts., IDP 2012:
wk1 - 3
wk2 - 3
wk3 - 4
wk4 - 0
wk5 - 2
wk6 - 3
wk7 - 4
wk8 - 5
wk9 - 5
wk10 - 6
wk11 - 4
wk12 - 2

Thus, there have been only 3 weeks where a score of 140 would give you a 50/50 or better chance of winning. Recipe for disaster.

Seriously Ben, why not just man up and admit that this team sucks? There's no shame in it. It happens to everyone. This year, 3 of my 4 teams are likely playoff bound, but the other is 3-9.

If it's any consolation, I have no reason to doubt your claim of having the "lowest highest score" in the league.