[sjelveh]

Very interested to see who it will be this week

Week 1
13 votes
Bears -3.5 win
Bengals +3 win

Week 2
9 votes each
Saints -2 win
Giants +3 win
Texas -31 win

week 3
32 votes
The Bears -3 push

week 4
12 votes
The Bears -3 win

[begolf25]

I am starting to see another trend developing here with da bears.

[sjelveh]

I am starting to see another trend developing here with da bears.

i see it too just bet on them

[the boss]

Bears look really good!

[aca]

[Ganchrow]

Very interested to see who it will be this week

Week 1
13 votes
Bears -3.5 win
Bengals +3 win

Week 2
9 votes each
Saints -2 win
Giants +3 win
Texas -31 win

week 3
32 votes
The Bears -3 push

week 4
12 votes
The Bears -3 win

Would it help to point out that out-of-sample the "system" is only 1-0-1?

[ivyconniver]

Would it help to point out that out-of-sample the "system" is only 1-0-1?

can you explain that a bit, Gan?

[tacomax]

There was no clear "best bet" in the first 2 weeks - there were teams tied for the lead but no winner, as it were. The only two clear best bets were the ones in week 3 and week 4 which have a 1-0-1 record.

[rolemand]

The system wasn't noticed until after week 2. So you can't count those.

[sjelveh]

say what you want but 6-0-1 and still going is impressive and ill play it this week for the hell of it

[Ganchrow]

Would it help to point out that out-of-sample the "system" is only 1-0-1?

can you explain that a bit, Gan?

Just as rolemand suggests, you shouldn't use the data same data that caused you to first become aware of a "trend" (such data is known as "in-sample" data) as evidence of the trend's predictive power. What's needed is independent confirmation based on data that was not used in formulating the hypothesis (known as "out-of-sample data").

For example, I'll point out that over the past two week, BTP contest participants whose usernames have exactly 4 letters have a record of 85-49-14 for +46 units. Now let's suppose that next week the 4-letter contestants go 36-36 for 0 units, implying a total record of 121-85-14 +46.

Now tell me ... does that rather impressive 3-week record imply anything meaningful about our 4-letter contestant trend?

[pags11]

I'm gonna try to re-start my own trend this week...

[ivyconniver]

just bumping this up, the top 6 "best bets" were 1-4-1 in week 5. If you're only doing the top best bet, it was ok state and it lost.

[Ganchrow]

just bumping this up, the top 6 "best bets" were 1-4-1 in week 5. If you're only doing the top best bet, it was ok state and it lost.

And after 5 weeks the players with 4-letters in their names 'trend' is now 118-85-21 for +37 units.

[tacomax]

And after 5 weeks the players with 4-letters in their names 'trend' is now 118-85-21 for +37 units.

That's the type of stuff we need here. I'll be all over the 4-letter-names next week.