Kelly Re-Visited

[u21c3f6]

yes but it has nothing to do with previous wins and losses, go back and read Joe thinks that his edge has something to do with his previous wins and losses and that is not a very good way to estimate your future edge.

Danshan11, I really wished that you did not use my name again. This is exactly what I meant by having to write too much but here goes.

When you wrote above that you got a line of -170 on the Yankees, that means to me that you think the Yankees will win this game approx 63% of the time. Do you agree or disagree with this?

Joe.

[danshan11]

Danshan11, I really wished that you did not use my name again. This is exactly what I meant by having to write too much but here goes.

When you wrote above that you got a line of -170 on the Yankees, that means to me that you think the Yankees will win this game approx 63% of the time. Do you agree or disagree with this?

Joe.

disagree roughly considering varying margins but I would say assumign -104 -104s as pinny baseball I would say there is a 62% chance the Yankees win that game BY my numbers ONLY (which do not mean shanoodle), the true chance of the yankees winning that game will be dictated by the closed line at pinnacle minus the margin and that would be the true chance of the Yankees winning that game.

[danshan11]

and this exact misstep is the exact place people go rogue on kelly and many other spots overestimating edge and more. Simple question is this why add more calculations and more variables to a variable super mix already, Our job as bettors is to value each variable and now we need to account for this variable to add more deviation and variance to the mix. you could easily because of variance alone get smoked by kelly and not because you are overestimating or making a mistake just pure good old fashion variance.

[u21c3f6]

disagree roughly considering varying margins but I would say assumign -104 -104s as pinny baseball I would say there is a 62% chance the Yankees win that game BY my numbers ONLY (which do not mean shanoodle), the true chance of the yankees winning that game will be dictated by the closed line at pinnacle minus the margin and that would be the true chance of the Yankees winning that game.

If I understand what you wrote, it is your belief that the closing line is the actual probability for that event. If that is your belief, then you and I will never be on the same page. I do not have that view of the closing line and my record supports my view. This is not to say that you cannot be successful doing it your way, it's just that you cannot say that I cannot be successful doing it my way (and I have been very successful doing this for a very long time).
Joe.

[HeeeHAWWWW]

and this exact misstep is the exact place people go rogue on kelly and many other spots overestimating edge and more. Simple question is this why add more calculations and more variables ..........

Because it makes more money, simple as that - if you have a well-calibrated model, it's simply irrational not to. It also provides you with a method of tweaking growth and volatility to your particular financial situation.

Not sure why you think edge is an additional variable. It's one you already have if you have a probability estimate.

[danshan11]

If I understand what you wrote, it is your belief that the closing line is the actual probability for that event. If that is your belief, then you and I will never be on the same page. I do not have that view of the closing line and my record supports my view. This is not to say that you cannot be successful doing it your way, it's just that you cannot say that I cannot be successful doing it my way (and I have been very successful doing this for a very long time).
Joe.

so you are now saying out loud in public that the closing line is not the closest thing we have to actual probability of an event, what you do all by yourself is more effecient?
You can do it your way but to be irresponsible and tell other people to base their kelly wagers on their past winning percentage is a bad bad move!

[danshan11]

Because it makes more money, simple as that - if you have a well-calibrated model, it's simply irrational not to. It also provides you with a method of tweaking growth and volatility to your particular financial situation.

Not sure why you think edge is an additional variable. It's one you already have if you have a probability estimate.

I respect your view but your response is vague at best. how many people have a well calibrated model, that is like telling someone its ok to go 100mph and keep the car in front only 6 car lengths because in your ferrari with all its advanced braking suspension and tire systems it is safe but in my 1984 olds with drum brakes all the way around not so much.
?1 how many people have a well calibrated model? that is long term proven and accurate? shit pinnacle doesnt even have one
?2 how can you have a probability estimate that is accurate, pinnacle dont?
?3 how is it not an additional variable, what do you even mean?

[danshan11]

If I understand what you wrote, it is your belief that the closing line is the actual probability for that event. If that is your belief, then you and I will never be on the same page. I do not have that view of the closing line and my record supports my view. This is not to say that you cannot be successful doing it your way, it's just that you cannot say that I cannot be successful doing it my way (and I have been very successful doing this for a very long time).
Joe.

Joe Please go find one other successful bettor who will say they know of something more effecient than the closing line, please do that, Joe you might know too much that might be your problem, Joe my friend the closing line is the most effecient thing we have on earth to estimate probability of sports events. Come on brother, wake up!

[HeeeHAWWWW]

?1 how many people have a well calibrated model? that is long term proven and accurate? shit pinnacle doesnt even have one

Source needed.

I'd also suggest a read of this:

https://en.wikipedia.org/wiki/Accuracy_and_precision

?3 how is it not an additional variable, what do you even mean?

As before: you have a probability estimate and odds. Multiply them together, this is your edge estimate. By definition, it's not an additional variable, it's a function of existing ones.

[danshan11]

where does your probability estimate come from?

[danshan11]

Source needed.

I'd also suggest a read of this:

https://en.wikipedia.org/wiki/Accuracy_and_precision

As before: you have a probability estimate and odds. Multiply them together, this is your edge estimate. By definition, it's not an additional variable, it's a function of existing ones.

Pinnacle sets lines before games, do those lines move? so there model is not as effecient as the closing line. if their model was the lines would not move and they would not take bets, sure they adjust to sucker in people because they need to get action but their model is not as accurate as the closing line, it cant be just because of last minute injuries, weather and many more factors

[HeeeHAWWWW]

where does your probability estimate come from?

My (well-calibrated) model.

[danshan11]

My (well-calibrated) model.

yeah but what are you calling a probability estimate, is that the line you are getting for the game from your model?

[Derailer]

Pinnacle sets lines before games, do those lines move? so there model is not as effecient as the closing line. if their model was the lines would not move and they would not take bets, sure they adjust to sucker in people because they need to get action but their model is not as accurate as the closing line, it cant be just because of last minute injuries, weather and many more factors

If you do not believe your model is better than the market, why would you bet the games? The closing line is the best estimate available of the expected outcome of a game. It is not the best possible.

[danshan11]

If you do not believe your model is better than the market, why would you bet the games? The closing line is the best estimate available of the expected outcome of a game. It is not the best possible.

No I do not believe my model is better than the closing line of course not. I bet the games because I know there will be gaps between a settled market and a current market example
I like Yankees -170 but currently its at -140, I think that line will come my way and in turn give me line value, maybe to -170 maybe not but normally enough movement gives me enough to beat the line. no it is not the best possible, LOL is this english class or a betting forum, if its english class I am screwed, if its betting I might have a chance. The closing line is the most effecient method we currently have.

[Derailer]

No I do not believe my model is better than the closing line of course not. I bet the games because I know there will be gaps between a settled market and a current market example
I like Yankees -170 but currently its at -140, I think that line will come my way and in turn give me line value, maybe to -170 maybe not but normally enough movement gives me enough to beat the line. no it is not the best possible, LOL is this english class or a betting forum, if its english class I am screwed, if its betting I might have a chance. The closing line is the most effecient method we currently have.

You have a belief in infallibility of markets. The markets are the best publicly available tool for a predicted outcome and a solid basis for measurement. However, a model can be better than the market.

So if a person is winning at a rate better than CLV might predict, they have to do a deep dive into the specific games that are causing the difference. By doing so, they may be able to determine if it was simply random luck or if the model is indeed better than the market.

[danshan11]

I dont think we can prove you right or me wrong in the variance argument. I did not say the market cant be wrong, the market is wrong occassionally look at the super bowls, I said the market is the most effecient tool we have. I think if you think your model is sharper than the market, I think you believe you are better at flipping coins than others. wins and losses mean nothing, if the modeler does not beat the line they will eventually lose except in very rare instances where the market is small or not effecient. some people play markets that are not effecient. I am talking about effeciency in major sports, not prop bets for president elect.

[danshan11]

winners who dont beat the line long term are just deviations and their future plays will come back to 50-50 minus margin at some point

[Derailer]

I dont think we can prove you right or me wrong in the variance argument. I did not say the market cant be wrong, the market is wrong occassionally look at the super bowls, I said the market is the most effecient tool we have. I think if you think your model is sharper than the market, I think you believe you are better at flipping coins than others. wins and losses mean nothing, if the modeler does not beat the line they will eventually lose except in very rare instances where the market is small or not effecient. some people play markets that are not effecient. I am talking about effeciency in major sports, not prop bets for president elect.

I understand. But if a person has a winning model, there is a basis for going by wins and losses over the long term. Presumably these things equal out over time. But a person with a winning model is going to go with it for as long as it works because it beats the market. In the longest of terms, it is the wins that matter and not the CLV.

[danshan11]

coin flip again
flip 1000 times
you get 500 H and 500 T NO you get 547 heads and 453 tails
that means heads has an edge of roughly 3.7% over tails, is that true?
wins and losses mean nothing without line value.
yes there are some very rare rare rare instances of people succeeding and not beating the line and that is like saying Warren Buffet is just lucky at the market or that 1 guy lived jumping off a skyscraper so that makes it true, Wins and losses to the general 99.99% of us MEAN NOTHING!
If I create 1000 SBR accounts and post picks on all of them a few them will be the best cappers in the world over 1000 plays and a few will be the worst.

[danshan11]

I understand. But if a person has a winning model, there is a basis for going by wins and losses over the long term. Presumably these things equal out over time. But a person with a winning model is going to go with it for as long as it works because it beats the market. In the longest of terms, it is the wins that matter and not the CLV.

if you play roulette and win you keep playing when do you stop as soon as you run out of money? You got to know your edge to keep betting and if you dont know or maintain a real edge you are recreational betting. But this is getting off topic, the topic is kelly and even though I use a 3rd cousin of kelly. I stand firmly on the belief Kelly is for no one but especially not for the people who really dont know their edge, win % based on current line is a horrible method, lines move a ton and that can kill you with kelly

[Derailer]

coin flip again
flip 1000 times
you get 500 H and 500 T NO you get 547 heads and 453 tails
that means heads has an edge of roughly 3.7% over tails, is that true?
wins and losses mean nothing without line value.
yes there are some very rare rare rare instances of people succeeding and not beating the line and that is like saying Warren Buffet is just lucky at the market or that 1 guy lived jumping off a skyscraper so that makes it true, Wins and losses to the general 99.99% of us MEAN NOTHING!
If I create 1000 SBR accounts and post picks on all of them a few them will be the best cappers in the world over 1000 plays and a few will be the worst.

All winning models have line value. This is a matter of degrees. There are no professional models that exist that win without line value.

[danshan11]

All winning models have line value. This is a matter of degrees. There are no professional models that exist that win without line value.

I dont know that as fact but I can agree a winning anything in sports betting has line value above the margin for sure!

[danshan11]

@derailer what is your opinion on Kelly?

[Derailer]

I dont know that as fact but I can agree a winning anything in sports betting has line value above the margin for sure!

So you are too caught up in the theoretical and not focusing on the practical. The people discussing these issues are interested in the practical usage of a model. They do not believe the closing line is necessarily the right number. They believe they may have the right number. In any case, you should focus your efforts on the practical applications of a model and issues you are dealing with rather than arguing over theoreticals.

[danshan11]

brother you can plainly see the line is accurate, how can that even be argued

MLB

O/U 16743-17440-1682 (0.42, 49.0%)

19414-16554 (0.15, 54.0%)

avg line: -128.1 / 115.5

on / against: -$82,673 / -$97,260

NBA

13849-14158-550 (-0.09, 49.4%)

NFL

3526-3561-197 (0.19, 49.8%)

avg line: -2.6

3538-3606-140 (0.75, 49.5%

case closed brother!

[danshan11]

So you are too caught up in the theoretical and not focusing on the practical. The people discussing these issues are interested in the practical usage of a model. They do not believe the closing line is necessarily the right number. They believe they may have the right number. In any case, you should focus your efforts on the practical applications of a model and issues you are dealing with rather than arguing over theoreticals.

I am 100% focused on winning and I got in the Kelly debate trying to learn what I am missing, I think my method I picked up from BT is flawless, X.XX to win 1 on faves and .XX to win 1 on dogs, flawless and simple, less risk on losers and more risk on more likely winners sounds good to me. Only issue i have is the weighted effect on CLV value

[Derailer]

@derailer what is your opinion on Kelly?

I would agree with your assessment more or less because your edge is always unknown and subject to change so betting using it is likely a road to ruin. Most use a modified version that is far less aggressive and those methods are probably solid.

You are never up until you've made your last bet or taken money off the table.

[HeeeHAWWWW]

Most use a modified version that is far less aggressive and those methods are probably solid.

Yes, that's a given. Nobody uses full Kelly - see my simulations in earlier posts!

[danshan11]

heehaw
when you place a bet using kelly
you factor your probability right? how do you do that using your win probability right?
so you make a line of -170 and bet kelly based on that right and that is flawed cause thaat is not where it closes either right?
so you get a kelly number using current line and that is flawed as well right cause that is not the closing line right?

so how do you accurately factor kelly not knowing the closing line? I ask because I have the same problem, using my method I bet at say -150 so I bet 1.5 units to win 1 and if it closes at -120 I over bet technically and if it closes at -190 I underbet technically

[danshan11]

Kelly back testing simulations are very very accurate because you know the closing line and you cant beat that but in future bets you dont have that luxury. and if you use an avg you are basically flat betting, LOL

[HeeeHAWWWW]

heehaw
when you place a bet using kelly
you factor your probability right? how do you do that using your win probability right?
so you make a line of -170 and bet kelly based on that right and that is flawed cause thaat is not where it closes either right?

I don't consider the closing line at all, it's not relevant to the subject tbh.

[HeeeHAWWWW]

Kelly back testing simulations are very very accurate because you know the closing line and you cant beat that but in future bets you dont have that luxury. and if you use an avg you are basically flat betting, LOL

The simulations use a variety of edges, odds and include errors.

The closing line has nothing to do with these sims.

[danshan11]

The simulations use a variety of edges, odds and include errors.

I think that right there says it all, it is a big ol crap shoot, OK thanks!

[danshan11]

thanks dude, I appreciate the back and forth and anytime any questions or feel like dusting it up, let me know I am always here for a good ol argument about things I literally know nothing about!