Read this before starting a "Who Do I pick in Survivor" Thread

[Thremp]

Tyler,

Separation is about percentages, not absolute numbers.

[mblax10]

[ QUOTE ]

3) Do NOT pick a game that features two rivals/teams that hate each other.

[/ QUOTE ]

Never thought about this before reading this thread. This kept me from picking Denver and the ensuing nail bitter this week.

[Levarkin]

New England makes for an interesting dilemma in week 3, since we can assume that pretty much everyone that can be on them, will be on them. They're the dominant favourite since they're the very obvious pick, and no other team's probability of winning is close. Survival or separation? After running some simple expectation formulas, here's a rule I've settled on. If:

Dominant favourite's chance of losing < 1/Number of teams remaining in the pool

It doesn't matter how good another pick might be. Take the dominant favourite.

Notes: this is assuming that you don't have an edge in this pool, and that every other team is picking the dominant favourite. If you do have an edge, or if other teams can't use the dominant favourite, you should be even more inclined to take the dominant favourite.

This situation came up for me since there are 15 people (including me) left in my survivor pool, no one else has used New England, and I'm pretty damn sure all of them will this week. But Pinny has New England's probability of winning at 14/15, which just happens to be the inverse of 1/15, and I have to go with New England, separation be damned.

EV Calcs:

Picking NEP: 1/15 x Pool $ =
Picking another team with a 100% chance of winning: (1-14/15) x 100% x Pool $

So a 100% chance of winning for the other team is the break-even point. Anything less (and of course it is less), and it's best to stick with the Pats.

Having said all that, if:

Dominant favourite's chance of losing > 1/Number of teams remaining in the pool

Then you'd almost always prefer separation. I'd take even a 75% fave if there were 20 teams left in my pool.

[Billy Bibbit]

Levarkin,

I will be picking the Pats in all of my leagues this week, but shouldn't your EV calculations take into account the value you'd get from saving the Pats for another week? This value is probably overstated but it's still worth something.

[DonkeyKongSr]

Yeah, saving the Pats due to "seperation concept" is fine, but when are you actually going to pick them then? A few of the teams I'm playing against have already used their NE pick, so they aren't able to take advantage of what is likely NE's biggest lock win of the year. At this point, I just want the biggest locks. However, I am in a tiny league with terrible players. I don't expect this to last beyond 4 or 5 more weeks.

Here's the Pats schedule. You can maybe pick them over Cleveland in a couple weeks, but after that, it's a long way to their next favorable home game.

Sun 9/23 Buffalo 1:00 pm
Mon 10/1 at Cincinnati 8:30 pm
Sun 10/7 Cleveland 1:00 pm
Sun 10/14 at Dallas 4:15 pm
Sun 10/21 at Miami 1:00 pm
Sun 10/28 Washington 4:15 pm
Sun 11/4 at Indianapolis 4:15 pm
bye
Sun 11/18 at Buffalo 1:00 pm
Sun 11/25 Philadelphia 8:15 pm
Mon 12/3 at Baltimore 8:30 pm
Sun 12/9 Pittsburgh 1:00 pm
Sun 12/16 NY Jets 1:00 pm
Sun 12/23 Miami 1:00 pm
Sat 12/29 at NY Giants 8:15 pm

[Billy Bibbit]

OK, I'm convinced.

Looks like 3. Pats 4. Chargers 5. Colts is the way to go unless the Bucs play surprisingly well in the next two games in which case I'll save the Colts for week 11 against the Chiefs.

[Levarkin]

Sigh. Turns out I was wrong about this; the formula is not as simple as I made it out to be. If anyone's curious, I think this is a more accurate rendition for the conditions under which to take the DF:

OW_P < DF_P / (2 x DF_P - 1 + n - n x DF_P)

Where OW_P is the probability that the "other" team wins (presumably the next best favourite), DF_P is the probability that the dominant team wins, and n is the number of teams (including yours) remaining in the pool. And that's just for the simple case of everyone takes the DF.

Exec. summary would be that even with 10 teams left, the other team would need only a 60% chance of hitting to make the separation worthwhile. The Steelers aren't a good play only if you're down to fewer than 5 teams.

I think.

[Tyler Durden]

Anybody have thoughts on breaking one of the rules by taking the Panthers in Atlanta this weekend? Or I could take the Patriots at home versus Buffalo.

[Jack Bando]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
ok remind me again why inter division is a bad thing when it's the pats at home ?

[/ QUOTE ]

The Pats at home against the Bills might be worth breaking the interdivision rule for. (NE is 11-1 against BUF since Brady became the starter)

[/ QUOTE ]
I still don't understand the interdivisional rule. Is there evidence that upsets happen more often in divisional games or did someone just pull this rule out of their ass? Serious question.

[/ QUOTE ]

According to Football Outsiders's Aaron (who I e-mailed the question to a few days ago), no it doesn't matter. If New England played Buffalo at home this week or some team that was Buffalo's clone but wasn't in the AFC East the outcome would not be likely to be different just because one is/isn't in the AFC East.

So according to FO's data, rule 5 is worthless.

[legend42]

[ QUOTE ]
Anybody have thoughts on breaking one of the rules by taking the Panthers in Atlanta this weekend? Or I could take the Patriots at home versus Buffalo.

[/ QUOTE ]

Is this a serious question?