why the followers are more likely to lose

[Thremp]

wow. Data just OWNED Thremp. Data is the sheriff of the Think Tank!!! Laying down law!

obvioustrollisobvious.jpg

Did he just make a claim, ask other people to test it, cite anecdotal evidence, and claim it was untestable without offering any evidence of his own?

This is Players Talk material.

[Data]

the capper thinks that the odds of his second play being 60%+ are x.

I do not know what you mean by x here. Only once the capper sees the line he knows if he has an edge or not. Literally, you are asking what are his estimate of the future line. I don't know.

[tomcowley]

Yes, the estimate of what his edge will be on the second line he thinks he has an edge on.

[Data]

Yes, the estimate of what his edge will be on the second line he thinks he has an edge on.

So, he made his estimate of the true line, he does not know what the market line is and you want to know his estimate of his edge over the market line that does not exist? He does not have that estimate. I don't know why does he need to have one.

[tomcowley]

Ok, how about after the first play he makes. Does he have an estimate for the second play after he makes the first play? (and before he sees any more lines)

[Data]

Ok, how about after the first play he makes. Does he have an estimate for the second play after he makes the first play? (and before he sees any more lines)

No, he never makes these estimates.

[tomcowley]

Ok, so if these estimates cannot be made, how are you making them? You've just said that the first game doesn't allow you to estimate the second (or thirds, or fourth, etc) games, but in earlier posts, you've said that earlier games do affect the estimates of later games.

[Data]

I have never talked about this type of estimates. If you provide a quote I will try to explain what I meant. Generally, the capper estimates a true line for a given contest, that's all.

[tomcowley]

Post 63.

[Data]

He expects that there will be certain contests where his +EV is higher than in the others. It's relative, not absolute. If he already played out a number of his higher +EV contests he can say that going forward on average his EV/play is going to be lower comparing to average EV/play before those contests.

I don't know, looks similar to what I already said. Unsure how to clarify this. Perhaps, there could be some benefit if I mention that he is going to bet only if he has a certain predefined edge or better. So, the lesser higher EV games ahead, his EV will be converging to preset bottom line EV, the borderline betable contests.

[tomcowley]

He expects that there will be certain contests where his +EV is higher than in the others. It's relative, not absolute. If he already played out a number of his higher +EV contests he can say that going forward on average his EV/play is going to be lower comparing to average EV/play before those contests.

Not that I agree with the third sentence at all, but can you write an equation for EV/play before the season and EV/play for the rest of the season after one game that allows them to be calculated or compared?

[Data]

Not that I agree with the third sentence at all, but can you write an equation for EV/play before the season and EV/play for the rest of the season after one game that allows them to be calculated or compared?

Let the handicapper to have a list of conditions that historically were not correctly reflected in the lines. Then,
H - a known variable, a historical average edge he had when certain conditions were met
x - an unknown variable, a factor close to 1 that shows the extent of how much the lines are sharper this season; likely, x<1; for simplicity of estimates, can be assumed to be a constant.
x(i)*H(i) - the EV for a single future contest
m - a constant, a minimum edge that calls for a play
N - a total number of contests meeting one or more conditions and where x(i)*H(i)>=m

The season average EV/play is the sum of x(i)*H(i) divided by N, where 1<=i<=N. Since H(i) is a variable, the average EV/play for the rest of the season varies after each contest completion.

[tomcowley]

How do you know any individual H(i) before the season, or the sum of H(i)? Or even N. Preseason, H(i) is still a random variable. It isn't a known until the situation and line become known.

[Data]

How do you know any individual H(i) before the season, or the sum of H(i)? Or even N.

H(i) is known as it shows what happened in the past. We do not know the exact N because we do not know if and how many times the certain conditions will occur and also because we do not know the x(i) and the N depends on x. The effect of a single contest on N is marginal.

[Tomahawk]

Definitely gotta believe in what you're doing and hang in there.

That's a good point. If you hang in you can't miss any winning or losing streak.

[luegofuego]

data,

try to apply your argument for a coin flipping 50% tout and think about what happens when this flipper goes 2-8, or 8-2, to start the season. in this case, we know to 100% perfection what his expected winrate for each and every game is. if you don't see the fallacy with convergence after this experiment, there is no hope for you.

[Salamander]

Even if OP thesis is correct (doubtful IMO), that early in the season (or whatever timeframe OP chooses to use) +EV handicappers EV per play is likely higher than when the followers jump aboard, and then inevitably and speedily converges to average EV thus causing agonized followers, the best you could hope for is changing the title of this thread from

"Why the followers are more likely to lose"

to

"why a tiny tiny tiny minuscule virtually imperceptible fraction of a percentage of followers are more likely to lose"


On paper, as an exercise, yeah maybe you can play with numbers enough to show there is some significance, but in reality....come on now, the effect, such as it were, is negligible. And the original post, if I remember correctly, does talk about accounting for real world observations/illusions, not theoretical possibilities.

[u21c3f6]

What Data is saying is a long known phenomenon that occurs in the financial markets. He is just applying it to sports wagering.

You either get it or you don't, just like +EV.

Joe.

[Data]

data,

try to apply your argument for a coin flipping 50% tout

Nobody expects that a follower of a coin flipper will win. The point is that a follower of a winning capper will win much less than a winning capper, not enough to overcome the juice, meaning he will lose too. I talked at length in this thread why it is wrong to apply coin flipping approach to a winning capper, it is a wrong model leading to wrong conclusions.

[tomcowley]

H(i) is known as it shows what happened in the past. We do not know the exact N because we do not know if and how many times the certain conditions will occur and also because we do not know the x(i) and the N depends on x. The effect of a single contest on N is marginal.

H(particular situation, particular line) is known. But you don't know the situation and line that will correspond to pick i (or game i of the season) until it happens. Those are still random variables, so the function H of those random variables is still a random variable.

[Thremp]

What Data is saying is a long known phenomenon that occurs in the financial markets. He is just applying it to sports wagering. You either get it or you don't, just like +EV. Joe.

lol

[JustinBieber]

Data fools morons once again

[bztips]

lol

Agreed. No clue.

[Data]

H(particular situation, particular line) is known.

No, it is H(particular situation) and it is known. The unknown future line is in x(i).

[tomcowley]

Whatever- you can't know what the situation is going to be in the 50th game of the season before the season starts. It's still a random variable.

[Data]

Whatever- you can't know what the situation is going to be in the 50th game of the season before the season starts. It's still a random variable.

Simplyfying, that variable is either 0 or H(condition met), it did not change after the first game and has no affect on the fact that the EV/play has decreased.

[tomcowley]

Simplyfying, that variable is either 0 or H(condition met), it did not change after the first game and has no affect on the fact that the EV/play has decreased.

Ok,, the second part is a non sequitur. But the first part, before the season, it is a RANDOM variable. It might turn out to be 0. It might turn out to be some positive number. You don't know. That's the whole point. You can't stipulate that the distribution of every H(i) is a point-mass of 1 on either 0 or some other number before the season starts

[u21c3f6]

Is it really necessary to use derogatory words?

Here is how I see it. People tend to gravitate to those things that appear to be working better than others. For example, they see Mutual Fund A earning double digits while other Mutual Funds are earning single digits. What happens? More people begin to invest in Mutual Fund A. At this point, unless Mutual Fund A continues to produce at that higher rate or better, the new investor will earn less. Worse, once Mutual Fund A gets back to a more normal rate of return and the new investor is only earning single digits (or even losing money), the investor now sees another Mutual Fund that is earning double digits and the whole process starts over again. It is a well documented fact that the average investor does not get anywhere near the actual rate of returns of their investments.

We can agree to disagree whether or not this scenario can be applied to a Handicapper A. I personally see the similarities. The choice to agree or disagree is yours.

Joe.

[Data]

Ok,, the second part is a non sequitur.

OK, I do not see it that way, perhaps we reached the point where the argument no longer develops. Not for you, but for a hypothetical interested reader I will provide an illustration.

The capper knows a ref who tells him before the season that he is going to officiate 10-15 games in the upcoming season and will give him some tips. The capper is pretty average otherwise. At some point in the season he goes 10-0. That gets him a follower. The tomcowley's argument that the capper's expected winrate till the end of the season is the same as his expected winrate before the season and I say it is likely lower which is tomcowley cannot agree with.

[tomcowley]

That's an entirely different argument. What is the distribution of H(50) before the season? A point mass of 1 on 0 or some positive number, or an actual distribution in the colloquial sense?

[djiddish98]

That's kind of a bizarre example - wouldn't it be more accurate (or at least more representative) to assume that the inputs into the capper's decision were fixed throughout the year (he used his same method for devising his picks - not mixing inside information with mediocre information at random times)?

Also, he could have gone 10-0 due to random chance, in which case, I don't think we can say his 10-15 games of inside info have been used up, per se.

[Data]

he could have gone 10-0 due to random chance

yes, this is why I keep saying "likely", even in the thread title

[djiddish98]

Fair enough - I came in without parsing every word in the message.

is Part 3 re: known edges an assumption to your theory? I'm having trouble figuring out how we as observers can determine the difference between a high EV sample and a lucky sample.

[Data]

is Part 3 re: known edges an assumption to your theory?

Come again?

I'm having trouble figuring out how we as observers can determine the difference between a high EV sample and a lucky sample.

We cannot, why even try? We can only guess what is more likely based on our assesment of that capper's skill set. But then, how confident we are in our assessment?

[djiddish98]

Part 3. While there is no such a thing as a winrate, the edge on a given bet is real and known to a capper before he makes the bet. Therefore, there are "subseries of those bets (chosen before the results of the bets being chosen are known)" that have more or less than average edge.

Sorry, this is what I was referencing. Is this an assumption for your theory (that's how I'm interpreting it).