[Data]

Are we discussing...?

I have some doubts that we are actually discussing anything.

...cappers who create their own estimate of a true line, or people who will bet the line at any price?

Cliff notes:
Data: The follower will lose due to juice even while following a wining capper/system if he jumps in and out.
tomcowley: If he folows a winning capper his results will be no worse as the follower makes the plays with the same EV as the capper
Data: The jumping follower is likely to miss the higher EV games and stay with lower EV games.

I suggest we model a winning/capper system, make sure the model is reasonable, does not contradict with reality, and then see if the theory is plausible or not. The "cappers who create their own estimate of a true line" seem to be a reasonable but by far not the only subset.

[tomcowley]

OK, so would you agree with the following version of your assertion: Preseason, an actual stat/news/weather/whatever capper who makes his own lines can make an estimate of what his EV/play will be at various points in the season. You believe that a follower who hops in and out for periods of time will have a lower EV/play, for the plays he plays, than the capper expects to have in those time periods before the season starts.

[JOHON8]

My advice to those who tail:

-follow big plays, especially if you like the reasoning, going with it blind will make you confused when you lose
-tail plays that you originally liked but didn't have the balls to go through with
-Always remember that you are not making any money in the future unless you learn from other cappers rather than tailing blindly.

[Tomahawk]

Followers are not likely to lose, couse they also have the edge, no matter when they bet.

[Data]

OK, so would you agree with the following version of your assertion: Preseason, an actual stat/news/weather/whatever capper who makes his own lines can make an estimate of what his EV/play will be at various points in the season. You believe that a follower who hops in and out for periods of time will have a lower EV/play, for the plays he plays, than the capper expects to have in those time periods before the season starts.

Following your definition of the winning capper, the capper does not know his EV going forward. For instance, the upcoming line for a single future contest is a range of numbers. The capper does not know the probability for any single number within that range while he must assign the probabilities for every number in that range.

[Jeffrey2010]

ha ha this whole argument sounds like one guy telling the other, hey you can't get from one side of the room to the other because in order to get to the other side, first you have to go halfway there. From this new halfway point, you have to go halfway. From this 2nd new point point you still have to go halfway...etc since you always have a point halfway, you never get there. And the other guy says, no you can get to the other side of the room. Then the first guy says, no, on paper you can't get to the other side of the room, then the first guy says no in the real world you actually can get to the other side of the room. And the first guy says but clearly in theory you can't get to the other side of the room.

Should I go on?

[tomcowley]

Following your definition of the winning capper, the capper does not know his EV going forward. For instance, the upcoming line for a single future contest is a range of numbers. The capper does not know the probability for any single number within that range while he must assign the probabilities for every number in that range.

His estimate of his EV of play n is E(EV(n)), not EV(n). I see no problem here.

[Data]

His estimate of his EV of play n is E(EV(n)), not EV(n). I see no problem here.

"Estimate" means "guess", means "we do not know it" going forward and I see this as a big problem.

The funny thing that even if we could assign P()s for all the "what if"s we would get nearly endless possibilities which would threw this model out as unreasonable due to requirement of a fantastic computing power.

[tomcowley]

"Estimate" means "guess", means "we do not know it" going forward and I see this as a big problem.

The funny thing that even if we could assign P()s for all the "what if"s we would get nearly endless possibilities which would threw this model out as unreasonable due to requirement of a fantastic computing power.

Nobody can KNOW the EV of one sports bet, ever, because that would require complete, perfect information about everything, applied perfectly. Beating this dead horse while simultaneously talking about knowing the EV of a follower's plays is hypocritical.

Every statement anybody makes about the EV of a sports bet, going forward, looking backward, or anything else, is just their best estimate based on the information they know and their ability to analyze it. This is simply not a problem.

Your claim, in the simplest terms, is that winning some games lowers E(EV(n)) in the near future, and losing some games raises it. Given sports market dynamics, I find it far, far more likely that the opposite is true. Using the simple example of a 4-game series, is game 4 going to be a higher EV play if the market mispriced games 1-3 by a lot or mispriced games 1-3 by a little? I can't imagine a winning capper who would expect the EV in game 4 to be higher in the latter case than in the former.

[Data]

We assume the capper makes +EV bets, this is not about his perception but a fact within this model. We did NOT assume he can guess the line with god's precision, so, please, just leave this out. If you do not require him to be precise then his imprecision only confirms my point that EV varies.

[Salamander]

OK here's a hyper-extreme example.

A capper with a long distinguished career has historically achieved 54%. Amazingly, his model indentifies only two quantum +EV states: a game can be 53.5% or 60%.

Therefore, over a 100 play sample going forward, one would expect 90 of these 53.5% games, and 10 of the 60% games (since overall he is at 54%), with the caveat that you can only actually get what the market gives you. Just taking the liberty of saying the future will stay in line with the long, long past.

So, the more 60% games clumped into a series of 10 consecutive plays, the greater the probability of that 10 game clump having a grouping of more wins is greater than if all 10 games in the grouping were 53.5%.

OP seems to be on the right track.

But...who can tell me why this is still gambler's fallacy?

[tomcowley]

OK here's a hyper-extreme example.

A capper with a long distinguished career has historically achieved 54%. Amazingly, his model indentifies only two quantum +EV states: a game can be 53.5% or 60%.

Therefore, over a 100 play sample going forward, one would expect 90 of these 53.5% games, and 10 of the 60% games (since overall he is at 54%), with the caveat that you can only actually get what the market gives you. Just taking the liberty of saying the future will stay in line with the long, long past.

So, the more 60% games clumped into a series of 10 consecutive plays, the greater the probability of that 10 game clump having a grouping of more wins is greater than if all 10 games in the grouping were 53.5%.

OP seems to be on the right track.

But...who can tell me why this is still gambler's fallacy?

Given this model, is there any follower strategy you can articulate, that doesn't involve time machines, that can change the EV/play? (hint: no) If not, do the follower and the capper converge to the same long-term ev/play? (hint: yes)

[tomcowley]

We assume the capper makes +EV bets, this is not about his perception but a fact within this model. We did NOT assume he can guess the line with god's precision, so, please, just leave this out. If you do not require him to be precise then his imprecision only confirms my point that EV varies.

You assume the capper makes +EV bets and conclude that followers manage to be -EV playing them.

[Sawyer]

My followers don't lose..

[Data]

You assume the capper makes +EV bets and conclude that followers manage to be -EV playing them.

The assume the capper makes more +EV bets than he makes 0EV and -EV bets.

[Data]

Given this model, is there any follower strategy you can articulate, that doesn't involve time machines, that can change the EV/play? (hint: no)

True, if the games were random, but, again, the sampling is without replacement. My example with a ref illustrates this.

[Data]

But...who can tell me why this is still gambler's fallacy?

Do not listen if they do. Whoever makes an argument referring to gamblers fallacy does not understand that it is applicable only to a random process and the winning capper cannot be modeled as a random (Bernoulli) process. I already explained why.

[tomcowley]

True, if the games were random, but, again, the sampling is without replacement. My example with a ref illustrates this.

Modeling a ref who will rig an exact number of games for you is about the worst model of a capper there can be.

Here are a couple of simulations I wrote:

In all of these, bets are in chunks of 100. The follower starts out not following. If the capper wins 60+, the follower jumps on. If the cappers wins <=50, the follower jumps off. The initial capper win% is .57. It drifts (in ways explained below), but can't go above .6 or below .54 (if the drift would move it out of that range, it moves it to .6 or .54). Obviously the win% in reality could go below .54, but that just corresponds to a period of no plays, so there's no point in considering it until the win% comes back to that.

1) After every 100 bets, the win% changes randomly by uniform (-.005,+.005)

The follower does better than the capper in terms of win%. Capper is 57% (obviously), follower is a little better.

2) After every 100 bets, the win% changes randomly on (-.006,+.004) or (-.0055,+.0045).

The follower does better than the capper in terms of win%, though both are in the 54's.

3) Same as 2, but there is also a 2% chance that the win% jumps to 60% (the capper finds something)

The follower does better than the capper.

4) Same as 2, except EVERY time the win% is below .55, AND the follower is not following, the win% jumps to .6.

Every time this happens, it gives the capper a free set of bets at .6 where the follower is not following. Even with the capper only finding angles while the follower isn't following, the follower still does marginally better per play.

I fail to see how this general idea of drifting edge with a tendency to decay, with occasional spikes (something like 3) isn't a fairly reasonable model. It's clear that the timing strategy isn't bad at all in terms of win% (although it's still bad for overall $ because it passes on +EV plays).

The ball is in your court to provide a model of EV drift that

1) supports your conclusion
2) makes any sense at all applied to a sports capper
3) isn't "The capper will lose his edge and go -EV eventually, so that person's followers will lose in the long run"

[Data]

I never said "exact number of games", I gave a range.

I already presented my model, it calls for rare market inefficiency that occur under certain conditions that are exploited by the capper/system. It is very reasonable due to a simple thought that if there is nothing special about a contest it is unreasonable to think that the line is going to be off enough to create an +EV betting opportunity. Your system, on the other hand gives the capper/system a magical ability to bet at 54%+ while playing random games. A model there the capper/system is assumed to be god-like fails reasonability test.

[tomcowley]

I never said "exact number of games", I gave a range.

I already presented my model, it calls for rare market inefficiency that occur under certain conditions that are exploited by the capper/system. It is very reasonable due to a simple thought that if there is nothing special about a contest it is unreasonable to think that the line is going to be off enough to create an +EV betting opportunity.

Then all of his (low volume of) plays are like this, and jumping on/off is irrelevant.

If you argue that these appear in a frontloaded way, and he plays other random -EV games to become -EV going forward, you fail criterion 3.

And, yeah, plucking +EV games all year based largely on stats is what a lot of cappers do. Why do books have microlimits on openers (or wait to post at all) if they aren't afraid of +EV bets? More than enough games, in any sport, in any season, have lines that move enough to be arbable with themselves at a different point in time, meaning +EV bets were available (and bet). How do people make a living betting sports?

Your system, on the other hand gives the capper/system a magical ability to bet at 54%+ while playing random games. A model there the capper/system is assumed to be god-like fails reasonability test.

Why would the games be random? What does this even mean? it simply ignores games with no edge, which is what actual winning cappers do.

[Data]

Why would the games be random? What does this even mean?

I mean that the contests where the capper has +EV bets appear randomly as in a sampling with replacement. You either do not realize that your model requires randomness or how unreasonable to model the winning capper as a random process. I already explained both of these points. I have no interest repeating this over and over.

On a side note, the low limits are due to uncertainty which is due to the lack of info about the contest. Once more info is available the limits grow.

[donjuan]

On a side note, the low limits are due to uncertainty which is due to the lack of info about the contest. Once more info is available the limits grow.

I lolled.

[tomcowley]

I mean that the contests where the capper has +EV bets appear randomly as in a sampling with replacement. You either do not realize that your model requires randomness or how unreasonable to model the winning capper as a random process.

Do you know, before the season, that play 50 will be a +EV bet or not? Of course not. Therefore it has some chance of being one and some chance of not being one. Therefore the proper treatment of it is as a random process (every bet doesn't have to be the same EV, but you need to describe how the E(EV) behaves to be able to say anything about how a follower will do going forward). What the hell alternative are you suggesting? I've given you multiple opportunities to describe such a process that will give rise to the behavior you expect (subject to the constraints in the earlier post), and you haven't.

[tomcowley]

On a side note, the low limits are due to uncertainty which is due to the lack of info about the contest. Once more info is available the limits grow.

Lol, no. Do you actually bet? It's because the people who set openers suck balls at it and want to get their mistakes corrected (to a degree) relatively cheaply. Lines that are mistakes. +EV bets. Appear all the time. They'd set almost equally shitty lines at gametime even if they had all relevant stat/player/weather/etc information (but no information on the line, incoming bets, etc).

[sideloaded]

On a side note, the low limits are due to uncertainty which is due to the lack of info about the contest. Once more info is available the limits grow.

[Data]

Lol, no. Do you actually bet? It's because the people who set openers suck balls at it and want to get their mistakes corrected (to a degree) relatively cheaply. Lines that are mistakes. +EV bets. Appear all the time. They'd set almost equally shitty lines at gametime even if they had all relevant stat/player/weather/etc information (but no information on the line, incoming bets, etc).

Lol, yes. Do you bet? Ever heard about circled games? Ever thought why the same book has different limits for different leagues? In all instances of low limits the book realizes there is a good chance that there is something that they do not know about the contest while there are people who do. The more uncertain they are the lower their limits.

[Data]

I've given you multiple opportunities to describe such a process that will give rise to the behavior you expect (subject to the constraints in the earlier post), and you haven't.

So, at this point we are trying to model the winning capper. For the follower the capper is a black box but we need that box to be plausible and transparent so that its behavior can be easily understood and predicted. I hope it is evident that if I can demonstrate a model based on a non-random process then I prove my point while if you demonstrate a model based on a random process you prove yours.

I am going to stick with the "helping ref" model. I realize you hate it. Of course, that model is plausible and very simple. It is also easy to understand and its behavior will prove my point.

What we got from you is this. The capper will make +EV bets because he has a superior methodology. And that methodology is modeled as a random process. Congratulations, you just substituted a black box with another black box, before we had a "winning capper" and now we have "his methodology" that we know nothing about and can be anything. I lol'ed at the part there you assigned that "methodology" a winrate declaring it as a random process and came to the conclusion that the capper is too a random process. I was expecting something much better from you.

Please tell me what is wrong with my model and come up with something better. No, you know what, take that back, just stop instead. I am pretty sure that I am stopping myself here.

[tomcowley]

Lol, yes. Do you bet? Ever heard about circled games? Ever thought why the same book has different limits for different leagues? In all instances of low limits the book realizes there is a good chance that there is something that they do not know about the contest while there are people who do. The more uncertain they are the lower their limits.

No. They don't open at low limits because of uncertainty about information material to the game. It's not like they spend the week accumulating information and by noon on gameday know (or even think they know) as much as Billy Walters.

Minor markets have low limits because they don't attract sufficient donk betting volume to offset the EV they give up to sharps. If there were as much donk betting volume on 10th division Albanian youth soccer as there is on the NFL, the limits would be comparable. Pinnacle frequently opens games at low limits, and once some volume is bet, increases the limit to "normal", and once an even higher volume is bet, blue-circles it to higher and higher limits.

[tomcowley]

I am going to stick with the "helping ref" model. I realize you hate it. Of course, that model is plausible and very simple. It is also easy to understand and its behavior will prove my point.

This model is absolutely goddamn stupid. It requires
1) somebody rigging games
2) somebody specifically aware of it making these picks public for people to tail
3) mixing in a bunch of -EV picks
4) a high enough percentage of rigged games going forward to offset the -ev picks going forward
5) future rigged games expected to appear in highly uneven distributions
6) and some other implausible crap too

Seriously, the number of followers who have EVER lost to this mechanism is likely zero. It's a completely worthless argument. It is not why actual followers actually lose, ever.

What we got from you is this. The capper will make +EV bets because he has a superior methodology. And that methodology is modeled as a random process. Congratulations, you just substituted a black box with another black box, before we had a "winning capper" and now we have "his methodology" that we know nothing about and can be anything.

That's the point. Come up with ANY sensible mechanism that gives rise to the behavior you claim will exist. You have complete freedom. Do it already.

[Data]

Minor markets have low limits because they don't attract sufficient donk betting volume to offset the EV they give up to sharps.

Q: Why this game is circled and I cannot bet as much as on the other games?

A by Data: This is because the status of the star player is uncertain.

A by tomcowley: This is because the books could not attract sufficient donk betting volume on that game.

[Data]

This model is absolutely goddamn stupid.

The model serves its purpose to a t. It clearly demonstrates how this may work.

It requires
1) somebody rigging games
2) somebody specifically aware of it making these picks public for people to tail
3) mixing in a bunch of -EV picks
4) a high enough percentage of rigged games going forward to offset the -ev picks going forward
5) future rigged games expected to appear in highly uneven distributions
6) and some other implausible crap too

I see no problem with 1 - 3. Don't know what 4 is for. 5, the ref will substitute another ref who is going to be out on medical leave. 6, there is nothing implausible about this model.

Seriously, the number of followers who have EVER lost to this mechanism is likely zero. It's a completely worthless argument. It is not why actual followers actually lose, ever.

Right, and the system above is completely worthless and stupid too because no way in reality that red ball can produce enough heat and light to warm up the smaller balls flowing around it in the absolute zero degree environment.

Come up with ANY sensible mechanism that gives rise to the behavior you claim will exist. You have complete freedom. Do it already.

Not that I feel I have to because I have one and you have none but I'll oblige. I say that the capper must have one or more winning angle, the angle that is not a common knowledge. Getting tips from a ref is an example of this type of angle. Here is another angle, I call it "The Mice and The Elephants Angle" as the elephants are said to be afraid of mice. The angle is in the contests where the team that has one of the three smallest starters plays against a team that has one of the three tallest starters because in the past those contests were mispriced. We assume that this angle is a big +EV angle. Throw in a number of plays that produce 0EV and we have a winning system. The system which is a non-random process. So, it's two for me, none for you.