[Grind-It-Out]

I have access to all the baseball picks from 'cappers on one sports monitor.

I wrote a quick and dirty script to try to get consensus plays, but I'd really like your help on how to make it better.

I have...
Step 1: Query for all handicappers who are hitting at the equivalent of 57% or better on a -110 line (I calculate this by units of profit per bet).
Step 2: Take each capper's (units of profit / #games) and set that as the initial rating for that capper. If a capper has picked less than 100 games, multiply the initial rating by (#games / 100) so we take into account the confidence level.
Step 3: Have an outer loop with all games being played today and an inner loop fetching all picks from each qualified capper for any particular game. If a capper picked a team, add his rating to the total value of that team. If a team ends up with a value of >.25 + the value of its opponent, pick that team as a consensus play.

Like I said, it's quick and dirty, and not very good.

I can provide code if that would help. Thanks so much!

[FreeFall]

What would you like to know? It sounds like you have it working as planned? Has it lost in the past year or something?

[Grind-It-Out]

What would you like to know? It sounds like you have it working as planned? Has it lost in the past year or something?

I have a situation where I have a group of cappers like:
Capper A: 238 - 157, +36.75
Capper B: 37 - 18, +22.84
Capper C: 72 - 41, +20.02
.
.
.
Capper X: 9 - 1, +8.1
..., etc.

I'm basically asking if anybody has an idea of a more optimum way to get a consensus play when some of these players disagree, because you'll never get a consensus from ALL of the handicappers.

[mathdotcom]

Your first step already sucks.

What if I've bet one game and won? I am hitting 100%, but who cares? You need to use a metric that accounts for statistical significance.

[Grind-It-Out]

Your first step already sucks.

What if I've bet one game and won? I am hitting 100%, but who cares? You need to use a metric that accounts for statistical significance.

Read step 2.

Also, I agree I need to filter further. Thus why I'm asking for help.

[durito]

There are no profitable long term baseball players listed on any consensus site.

[Maverick22]

Are there that many cappers that are hitting 57% over lots of games?

[Grind-It-Out]

There are no profitable long term baseball players listed on any consensus site.

This isn't a consensus site. This is a handicapper monitor. I just happen to have access to all of their picks, so I can create my own consensus.

[Grind-It-Out]

Are there that many cappers that are hitting 57% over lots of games?

There are currently two guys hitting over 57% after 100+ games. But there are also guys that are 8-0 after 8 games, 14-2 after 16 games, etc. I'm wondering what value, if any, I should give those players.

Maybe it would be better to simply play all of the games from the two 100+ game cappers, and ignore any picks on which they disagree.

[mathdotcom]

Your 2nd step is not correctly bringing in a confidence interval.

I would take all the cappers, run them through a filter to see who is significantly doing better than some % (57% in your case), and drop everyone else. I imagine you will be dropping a shitload of people though.

[durito]

This isn't a consensus site. This is a handicapper monitor. I just happen to have access to all of their picks, so I can create my own consensus.

Either way if you think these people are gonna hit the equivalent of 57% v -110 on WA MLB lines you are wrong.

[Grind-It-Out]

Your 2nd step is not correctly bringing in a confidence interval.

I would take all the cappers, run them through a filter to see who is significantly doing better than some % (57% in your case), and drop everyone else. I imagine you will be dropping a shitload of people though.

Very good point. What about something like this for determining each capper's rating...

1: Determine adjusted win % by taking units of profit per pick (units of profit / # bets) and setting it equal to .91*w - (1 - w).
2: Figure out adjusted # wins by taking adjusted win% * num bets.
3: Use binomial distribution and solve for the probability such that the Cumulative Probability: P(X < #wins) ~= .95.

Example 1: Handicapper A is 252 - 182, +42.74 units.
1: 42.47/(252+182) = 1.91w -1
w% = 57.5%
2. #wins = .575 * (252+182) = 249.55
3. In our binomial distribution formula, set #trials to (252+182) and #successes to 249.55. Find the probability such that Cumulative Probability: P(X < 249.55) ~= .95. Use the resulting probability as our final adjusted win% for this capper. Capper A is a 53.5% capper.

Example 2: Handicapper B is 6-0, +6.05
1: 6.05/6 = 1.91w - 1
w% = 105% -> adjust to 100%, since a greater % is not possible.
2: #wins = 6
3: Capper B is a 59% capper. Hmmm, I was hoping it didn't work out like that.

[Grind-It-Out]

Either way if you think these people are gonna hit the equivalent of 57% v -110 on WA MLB lines you are wrong.

Very true. Perhaps I should and will lower the number.

[Maverick22]

And this all assumes, each capper comes with his own unique picks...

If a capper "borrows" from another capper your "consensus" becomes just that much weaker...

Not sure how much this happens, but its something to think about.

[sharpcat]

Judging ones success comes down to a lot more than win loss records and profit over small samples.

Are these cappers beating the closer? not beating the closer? have you actually kept tabs on their W/L record or are you just taking their word for it? if you are personally tracking their record, are they profiting against the opener or they profiting against the closer?

[Grind-It-Out]

Judging ones success comes down to a lot more than win loss records and profit over small samples.

Are these cappers beating the closer? not beating the closer? have you actually kept tabs on their W/L record or are you just taking their word for it? if you are personally tracking their record, are they profiting against the opener or they profiting against the closer?

I can verify that all the records are accurate.

I have access to:
Time of pick, team picked, line at time of pick, closing line

Maybe I need to find a way to use all of that information.

[Grind-It-Out]

And this all assumes, each capper comes with his own unique picks...

If a capper "borrows" from another capper your "consensus" becomes just that much weaker...

Not sure how much this happens, but its something to think about.

That's actually a really good point. I know they don't borrow from each other (because they can't see each other's picks), but they may borrow from some third party. Or, multiple cappers may use a very similar handicapping method, which would be almost as bad.