Lucky Ladder database mining

[pawnking]

It occurs to me that with the lucky ladder, you have a perfect vehicle for some serious data mining. Here are dozens and dozens of picks, ranked in order of most certain to least certain. Can you do an analysis of how frequently consenses #1 picks are correct? I see a lot of people picked the NE/Oak game correctly, with maybe a 90 to 95% success rate.

Many who picked that game as a low certanty picked Oakland to win. Everyone I saw with my scanning who picked that game high picked NE to win.

Can all of this information be used to generate some kind of database whose total predictive power can be compared to actual results? If we attribute a high value to #1 picks and a low value to last picks, can we come up with a factor which might have a strong correlation to actual wins and losses. And can this predictive power of the ladder outperform Vegas' return on straight picks? Can regession analysis provide any kind of corrolation between being a high certainty pick and beating the point spread?

Just wondering...

[kashikis]

That would be interesting. Good idea I wonder if they could do this.

[Mercer Boy]

Dang pawnking, I wish I knew statistics!!! You most likely have a great analytical mind for things like this, being a chess player and all.

This does sound like a cool thing to work on, but I would have no idea where to start. I will tell you this, though...it is MUCH easier to pick football ladders than it is to pick baseball ladders. I don't know how many of you played on the baseball site, but we've only had one week out of 22 where an overwhelming majority of the players have gotten at least 3 consecutive picks. The only reason that happened was because all the favorites actually won that day. Many weeks a score of 10 (4 correct picks) wins, and that's usually only one or two people!

It's simple math. It's much harder for baseball teams to win every time out on a particular day of the week (Sunday)...the best teams only have a .600 win percentage - and that's only if you win over 100 games! . There are only about 2-3 teams that do that each year. So, even though there are a SMALLER number of games to pick each week (15 instead of 16 in football), the chances of someone picking all the correct winners is significantly lower in baseball.

In football, the best teams normally have a win percentage of .625 to .814, and that counts every team that goes 10-6 to 13-3. There are normally about 8 teams that finish 10-6 and above.

Therefore, there is just a higher probability of certain teams winning week in and week out. As long as you put the "sure thing" teams down at the bottom, you should get at least 3-4 correct picks each week barring an upset. It just depends on the matchups for that day. If there are a lot of good teams playing each other, the scores normally trend lower. If it's good against bad as a majority, we have weeks where over 50% of the players get 6-10 picks in a row.

[Arlo]

Interesting idea!

I'd love to see some numbers here; I'm assuming that you could indeed come up with some interesting conclusions. The polls we'll be running on the front page for Monday night games (there's one up now) might also be worth analyzing since they ask for point differentials instead of just the winner (and also receive a large number of replies).

For example, take a look at the poll we ran on the season opener, which turned out to be very accurate.

Then again, maybe we should just mine MadScott's ladder. For most other folks, week 1 was pretty rough...

[pawnking]

I would suggest that you could contact Aaron Shatz over at http://www.FootballOutsiders.com mailto:aaron@footballoutsiders.com

He's sort of the Bill James of football, and might be interested in such a project. I just hate to see a good database go to waste

Myself, I understand statistics just enough to look foolish trying to talk about it But I do know that data mining has lead to surprising results frequently, in sports as well as in business.

Just throwing things to the wall here. My ladder didn't do too well when San Fran upset St. Louis. It looks like my Sweet 16 should get 10 teams to week 2, which is better than I did last year