Minimum plays to confirm I can hit 70%

[brxbmbers42]

I started this last week. You want the documentation?

http://forum.sbrforum.com/nfl-handicapping/821553-what-your-best-plays-week-13-a.html#post7776667

Every team I mentioned, and I'm bet them all, again, believe what you want, was a winner.

Not you so much Deluxe, but everyone else, if you wanted the goods, what are you gonna say now?

I can somehow manipulate timestamps and SBR FORUM?

'Nuff said.

congrats. you were hot last week. which means like 99% of the other human beings walking the planet that gamble. you will be 2-12 next week.

[StackinGreen]

a minimum 500 plays is needed.

you're welcome.

If you can tell me on what mathematical model / confidence interval you base this, I would appreciate it.

Something tells me that's not the case though.

[StackinGreen]

congrats. you were hot last week. which means like 99% of the other human beings walking the planet that gamble. you will be 2-12 next week.

So I'm not a liar.

Again, bx I never picked anything with anyone here. All I wanted to know was a stat to help me find out if I'm bu--sh-t

Do you see why this is so frustrating?

IT'S NEVER BEEN ABOUT TOUTING. How many times do I have to say it?

[BeatTheJerk]

Your pathetic name calling can't change the fact that my week 13 picks were 4-0

Sorry to disappoint you.

But hey, I like winning. Obviously you don't.

Again, this thread hasn't been about touting. Just analysis.

Still, I exposed all the trolls. Now I expect more name calling, because I delivered the goods and they feel stupid. It's too predictable.

This is what I do on here when I smell a big pile of shit. I'll stop now.

[StackinGreen]

bronx

Bronx, that's the whole point: I might not even bet next week.

Just saying. I've been around a lot longer than you realize. Just because I'm a new SBR poster doesn't mean I don't know anything.

I'm every bit as savvy as the next. All I'm looking for is some helpful advice.

[brxbmbers42]

So I'm not a liar.

Again, bx I never picked anything with anyone here. All I wanted to know was a stat to help me find out if I'm bu--sh-t

Do you see why this is so frustrating?

IT'S NEVER BEEN ABOUT TOUTING. How many times do I have to say it?

i never said you were or werent a liar. i didnt click you're link with you're past picks. just believing what you say. it means nothing pal. 70% last week. 22% this week. that is a certainty. and dont ever question rudy. guy is 10000 times sharper than you.

[StackinGreen]

OK

This is what I do on here when I smell a big pile of shit. I'll stop now.

Thank you. Just to show you that I'm class, I take no offense. I understand that most posts are of the variety of madness / just to F with other people and waste their time.

I only wanted to know a mathematical method, ironically enough, that would expose me for BS if that's in fact what I was.

That's all.

[BeatTheJerk]

Thank you. Just to show you that I'm class, I take no offense. I understand that most posts are of the variety of madness / just to F with other people and waste their time.

I only wanted to know a mathematical method, ironically enough, that would expose me for BS if that's in fact what I was.

That's all.

[StackinGreen]

i never said you were or werent a liar. i didnt click you're link with you're past picks. just believing what you say. it means nothing pal. 70% last week. 22% this week. that is a certainty. and dont ever question rudy. guy is 10000 times sharper than you.

You don't read my posts. Why bother posting on my threads then?

Do you realize the point of this thread? I have made it clear over 5 times in responses.

I may not bet next week. Even if Rudy is sharper than I am, you wouldn't know.

You don't know anything about certainty, or uncertainty, bronx, or you would have answered my stats question.

[StackinGreen]

My man.

Trust me, I'm unlike the rest.

Does that mean I can keep picking winners? That remains to be seen.

Cheers.

[brxbmbers42]

You don't read my posts. Why bother posting on my threads then?

Do you realize the point of this thread? I have made it clear over 5 times in responses.

I may not bet next week. Even if Rudy is sharper than I am, you wouldn't know.

You don't know anything about certainty, or uncertainty, bronx, or you would have answered my stats question.

guey the geeks hang out in the think tank like i previously stated. head over there. this is the forum where real men shoo the sh1t.

[StackinGreen]

guey the geeks hang out in the think tank like i previously stated. head over there. this is the forum where real men shoo the sh1t.

Just stop dude, please.

I'll answer any question you want, then you can judge my acumen. But don't judge me unless I've analyzed or taken a stand and you think it's crap ... AND you can tell me why.

That's how real cappers do it.

[gryfyn1]

Your pathetic name calling can't change the fact that my week 13 picks were 4-0

Sorry to disappoint you.

But hey, I like winning. Obviously you don't.

Again, this thread hasn't been about touting. Just analysis.

Still, I exposed all the trolls. Now I expect more name calling, because I delivered the goods and they feel stupid. It's too predictable.

of course you didn't point out your week 12 picks went 1-3, but whatever --

As far as how many picks will it take to be statistically significant -- the answer is much more than you think --

but its doubtful that number can be maintained over more than a couple hundred plays.

[StackinGreen]

I never made a claim about priors, gryf. Like I said, though, that isn't the point of this post. Here is the point, as Jim from atomicfootball so nicely sent to me:

---
A good question...

I've done a fair amount of significance testing with my work, but I must confess that I always have to look up the equations every time. In the meantime, here's a pretty good rule of thumb though...


Assuming that you're never so good as to achieve probabilities too terribly far from 50% (even 70% is close enough for this to be a good estimate), a 3-sigma significance (you can look up the confidence intervals) at 70% would equate to approximately...


N = (0.5 * 3 / (0.7 - 0.5))^2 or about 56 games for a 39-17 record ATS.


I'm not too serious about my pro predictions (actually this year's endeavor is purely experiment to see if the lines undercorrect for certain information -- seems to maybe be the case since my ATS percentage was quite low). For college predictions, I believe numbers as high as 65% are possible, but only with a pitifully few number of games (maybe one out of 50, at most).

---

So, it looks like 39-17 would be a minimum starting point.

You guys would at least agree that if you did that over 2 years, it would be VERY profitable. Ok, I can retire this thread now.

If anyone is interested in NFL thoughts, though, you can PM me or ask here. I wish you all the best of luck.

[RudyRuetigger]

I never made a claim about priors, gryf.

You showed how you went 4-0 and could easily hit 70%. but the 1-3 week doesnt matter?

Like I said, though, that isn't the point of this post. Here is the point, as Jim from atomicfootball so nicely sent to me:

---
A good question...

I've done a fair amount of significance testing with my work, but I must confess that I always have to look up the equations every time. In the meantime, here's a pretty good rule of thumb though...


Assuming that you're never so good as to achieve probabilities too terribly far from 50% (even 70% is close enough for this to be a good estimate), a 3-sigma significance (you can look up the confidence intervals) at 70% would equate to approximately...


N = (0.5 * 3 / (0.7 - 0.5))^2 or about 56 games for a 39-17 record ATS.


I'm not too serious about my pro predictions (actually this year's endeavor is purely experiment to see if the lines undercorrect for certain information -- seems to maybe be the case since my ATS percentage was quite low). For college predictions, I believe numbers as high as 65% are possible, but only with a pitifully few number of games (maybe one out of 50, at most).

---

So, it looks like 39-17 would be a minimum starting point.

You guys would at least agree that if you did that over 2 years, it would be VERY profitable. Ok, I can retire this thread now.

If anyone is interested in NFL thoughts, though, you can PM me or ask here. I wish you all the best of luck.

None of this matters. 39-17 isn't a starting point. +EV wagers are.

http://forum.sbrforum.com/handicapper-think-tank/779237-acceptable-strike-rates-sample-sizes.html#post7458171

Here is how I look at this.

First I don't start by saying I want to "shoot" for any particular win rate ATS. I collect various data, compare that data to each other in various combinations and "see" if there are any combinations that creates a bell curve of subsets. I rarely go back more than one season. If I go back more than one season it is because there were not enough events to get a valid sample size. Once I find a combination that gives me what I am looking for, I then test it forward (and always with real money although relatively small money in the testing phase. Paper trading is not as valid). The win rate is determined by one of the subsets that the market will give up, not by what I want.

Going forward, it will either be profitable or it won't. As I go forward I keep checking for validity (profitable or not) using the following method. For ATS wagers, I assume that the random win rate is 50%. I then calculate the SD for the number of wagers based on a 50% win rate (sq rt of (# wagers*.5*.5)). I then compare the # of winners to how many SDs it is away from 50%.

The sample size for validity will be dictated by the win %. I don't call anything valid until I am at least 2 SDs away from 50%. If it is valid, it shouldn't be long before the results fall outside 2.5 SDs and with more events, a valid "method" will usually fall outside 3 SDs.

Hope that made sense.

Joe.

[JohnGalt2341]

If you can hit 70 out 100 of -110 plays I will consider you God. I'm an Atheist by the way.

[pokernut9999]

I hit 70% every damn day !!!!!!!!!!!






One day 70% winners , another day 70% losers

[milwaukee mike]

i guess a short answer to your question is that it a number can never be placed on it because it depends on who you ask

some people think it takes billions of trials to validate, others think you can run it like polling and assume a 4% error rate or some such b.s.

i for one think the lines are way too good to allow even ONE pick a year to hit 70% winners over a billion years, so even though i'm usually significantly better than 50% in contest plays/announced picks, i think the real way to be profitable is through bonuses and scalps

[brxbmbers42]

I hit 70% every damn day !!!!!!!!!!!






One day 70% winners , another day 70% losers

bingo. no way this guy could do any better than 47% over 1 years time. i fukkin guarantee it. maybe he won 70% one week. big fukkin deal. next week you will be 2-12. how hard is that to understand

[StackinGreen]

brx, you don't deserve the winners

I will post them at a different site.

Would you like me to come back when I hit the number so you can see?

Seeing is believing even for you.

Anyone can PM me if they want to see where the posts will be.

There's no way I won't get 60% minimum. But you have to give me time, that's all I ask.

That's why people like you lose, Bronx: You bet every week (day?).

PM if interested, otherwise wait until the data is out to say "I told you so."

[Rod1010]

PM me. hahah hey sportspick gtfo phaggot

[philswin]

How many trials would depend on what confidence level you are looking for. If you consistantly hit 70% percent you should be a pretty wealthy person. At least a millionaire. You should be in Vegas unloading 10-20K per game. Is that what you are doing?

[StackinGreen]

How many trials would depend on what confidence level you are looking for. If you consistantly hit 70% percent you should be a pretty wealthy person. At least a millionaire. You should be in Vegas unloading 10-20K per game. Is that what you are doing?

Do you know Kelly wagering?

Do you realize that 70% plays, almost by definition, can't come along very often?

If you recognized either of these two principles you wouldn't say something like you did.

Here, I'll condense it for you: Variance even with 70% winners can CRUSH you.

Get in the know, then make statements. Think before you write.

[brxbmbers42]

Do you know Kelly wagering?

Do you realize that 70% plays, almost by definition, can't come along very often?

If you recognized either of these two principles you wouldn't say something like you did.

Here, I'll condense it for you: Variance even with 70% winners can CRUSH you.

Get in the know, then make statements. Think before you write.

for the fifth time. the geeks hang out in the think tank. this is players talk where real men come to shoot the sh1t. beat it kook. no one is interested in you winning 42% of you're plays.

[RudyRuetigger]

So I'm not a liar.

Again, bx I never picked anything with anyone here. All I wanted to know was a stat to help me find out if I'm bu--sh-t

Do you see why this is so frustrating?

IT'S NEVER BEEN ABOUT TOUTING. How many times do I have to say it?

seemed reasonable, but then i reread the first post and then this one below:

Do you know Kelly wagering?

Do you realize that 70% plays, almost by definition, can't come along very often?

If you recognized either of these two principles you wouldn't say something like you did.

Here, I'll condense it for you: Variance even with 70% winners can CRUSH you.

Get in the know, then make statements. Think before you write.

And oh yea, clearly you don'tknow Kelly wagering. If you did, you'd realize it has nothing to do with this thread.

[dice]

I'd say hitting 70% (more than 30 picks a year) over 3 years would give you enough data points.

[ssk13809]

Don't listen to these haters that say it can't be done.

I met one of them, he bet me 200 points I couldn't hit 70+% over a min of 24 plays.


I'm 13-5-1 right now, closing up on that 200 points.

[MonkeyF0cker]

This again? Seriously? Is this the same dude creating ghost after ghost?

[ouman101]

I don't even know why I waste my time reading this shit, other than watching the OP make an ass of himself. Post picks or gtfo. Nobody wants to hear how "you always win", posting winners is all that gets respect.

[MartinBlank]

I am convinced these "I can hit 70% of my plays" posters are all the same guy.

[Richards]

Not that I believe the original poster, but to go a little further on some basic statistics, there is no way to "minimally confirm" that your lifetime win rate is > 70% without your sample size approaching infinity.

You can however narrow the probability down to a small enough number that it is negligible with a large enough sample size.

You would do this:

a) pick your confidence interval, let's say 99%, in laymans terms we want to be 99% sure that the probability answer we get (p-value) is correct.

b) Name your null hypothesis: "Does my lifetime loss rate exceed 30%?"

c) Take your sample. Test assumes these are independent samples. (Another conversation). MINIMUM sample size for a valid Z-test is 30 but your going to want many more, to have ANY accuracy here.**

d) Do a Z-test. Consult the Z-table. We'll be doing an "estimator of true probability." Really the best we can do is be have a certain confidence that the error, or the difference between our sample and the true probability is smaller than a certain size.

You'll end up with a results something like (made up numbers) with 55 wins out of 100 in your sample you'll get an answer like "Your true loss rate is 55% with a margin of error of +/- 15%. Not a very good estimator.

If you post an actual record, I'll volunteer to do the real estimator or true probability in this thread

** Win rate at beating the spread should be a similar test to "is a coin fair" which requires the following sample sizes for probability estimation: (from Wikipedia)

If a maximum error of 0.01 is desired, how many times should the coin be tossed?

at 68.27% level of confidence (Z=1)
at 95.45% level of confidence (Z=2)
at 99.90% level of confidence (Z=3.3)

So you're gonna need 10,000 plays to have a 95% confidence you are within 1% of your true win/loss pct.

A little tired when I typed this, so any other math geeks please throw in some corrections for me

[StackinGreen]

I'd say hitting 70% (more than 30 picks a year) over 3 years would give you enough data points.

Thank you for giving input that answers my original question. Best,

SG

[StackinGreen]

Not that I believe the original poster, but to go a little further on some basic statistics, there is no way to "minimally confirm" that your lifetime win rate is > 70% without your sample size approaching infinity.

You can however narrow the probability down to a small enough number that it is negligible with a large enough sample size.

You would do this:

a) pick your confidence interval, let's say 99%, in laymans terms we want to be 99% sure that the probability answer we get (p-value) is correct.

b) Name your null hypothesis: "Does my lifetime loss rate exceed 30%?"

c) Take your sample. Test assumes these are independent samples. (Another conversation). MINIMUM sample size for a valid Z-test is 30 but your going to want many more, to have ANY accuracy here.**

d) Do a Z-test. Consult the Z-table. We'll be doing an "estimator of true probability." Really the best we can do is be have a certain confidence that the error, or the difference between our sample and the true probability is smaller than a certain size.

You'll end up with a results something like (made up numbers) with 55 wins out of 100 in your sample you'll get an answer like "Your true loss rate is 55% with a margin of error of +/- 15%. Not a very good estimator.

If you post an actual record, I'll volunteer to do the real estimator or true probability in this thread

** Win rate at beating the spread should be a similar test to "is a coin fair" which requires the following sample sizes for probability estimation: (from Wikipedia)

If a maximum error of 0.01 is desired, how many times should the coin be tossed?

at 68.27% level of confidence (Z=1)
at 95.45% level of confidence (Z=2)
at 99.90% level of confidence (Z=3.3)

So you're gonna need 10,000 plays to have a 95% confidence you are within 1% of your true win/loss pct.

A little tired when I typed this, so any other math geeks please throw in some corrections for me

Things are very roughly received here. I have no problem posting plays on an unbiased and timed thread but doubt I'll do it on Players Talk.

Most are disrespectful and they don't deserve it. If they think I'm BS anyway they won't care. Good. I'll save it for the people who have contacted me.

That said, thanks for the response, Richards. Best,

SG

[RudyRuetigger]

i dont understand why sbr doesnt ban these people trying to get customers

[brxbmbers42]

i dont understand why sbr doesnt ban these people trying to get customers