Kelly betting live test - is it a madman's goal as some quants around here suggest?

[Duritz]

Overbetting or under betting are not the issue here. It's the Kelly edge/odds betting method that I'm attempting to successfully employ, with my own twist, is the static br.

If my bank gets to 15k then I'll be chronically under betting, and so on, but that's the point. The ROI using Kelly's unique staking plan that's really the emphasis here.

 

[Duritz]

carl -15.5, $700 at 2.75 betstar
gws 40+, $190 at $41 bet365
rich line -13.5 $800 at 1.935 pinnacle
gcst over 15.5 $575 at $4.70 flemsb
wcst h2h $1099 at $2.36 pinnacle
wcst h2h $930 at $2.35 tatts
hawks over 15.5 $2500 at $2.45 sportingbet

This total = $6794

 

[Nothing]

wcst h2h $1099 at $2.36 pinnacle
wcst h2h $930 at $2.35 tatts

Has NICNATNUI stolen your account? I don't know what you have seen in WC in the last few weeks but there midfield is the weakest it's been in years.

 

[Duritz]

Has NICNATNUI stolen your account? I don't know what you have seen in WC in the last few weeks but there midfield is the weakest it's been in years.

Yeah, I know... I was not agreeing with the bet either. But it's me devoting this 10k to the method, I can't add a "Duritz discretionary factor" (DDF) no matter how much I've been tempted.

 

[Duritz]

Thank GOD west coast won...

 

[Duritz]

Gee, Bob Terwilligers, it's been a baaaaad weekend for the old index so far, but there's BIG betting action to be had yet!!

syd line - 59.5, 1648 at 2.02 pinnacle
syd line - 59.5, 352 at 1.98 pinnacle
port over 15.5, $150 at 8.00 flemsb
port h2h $1645 at $3.42 pinnacle

 

[Lensen]

jesus christ mrs duritz is gonna be pissed at you

 

[Duritz]

jesus christ mrs duritz is gonna be pissed at you

Flowers?

 

[Lensen]

Flowers?

you might want to sneak out and pick them from someone's garden, though

 

[Thedeftouch]

Some Kelly research for you: "If the gambler has zero edge, i.e. if b = q / p, then the criterion recommends the gambler bets nothing. If the edge is negative (b < q / p) the formula gives a negative result, indicating that the gambler should take the other side of the bet."

 

[d3ft]

Cruel.

 

[Duritz]

Some Kelly research for you: "If the gambler has zero edge, i.e. if b = q / p, then the criterion recommends the gambler bets nothing. If the edge is negative (b < q / p) the formula gives a negative result, indicating that the gambler should take the other side of the bet."

And what if, like Richard Douglas in "Falling Down", you're on the edge?

 

[Rourke]

FWIW the AFL Predictor has been using something like* a half-Kelly this year based on underlying macro KPIs geared towards luck reduction. Bank is up from initial $1,000 to $2,070 currently. I advise concentrating on the 50/50 line and head to head because they have the smallest rakes, and the AFL Predictor calculates the optimal mix of bet sizes using this paper.

*By "something like", I mean that it uses the market as an extra source of information that it takes into account using Bayes, weighted against its confidence in its own knowledge. The result has some similarities to half-Kelly. We tend to bet around 25-35% of the bank each week.

 

[somethingrandom]

And what if, like Richard Douglas in "Falling Down", you're on the edge?

 

[Dominion]

Except that the ups and downs in punting in real life (not simulations) do tend to be cyclical. When this happens, Kelly is disastrous (without a static BR figure).

I hope I'm misunderstanding the first sentence, or at least that you don't truly believe it, because that is just ridiculous (wins and losses obv come in completely random streaks, you're never 'due' for a win/loss). But more importantly, the second sentence is also false. Using Kelly, you'll be in the EXACT same spot after 100 bets if you consistently win exactly 3 out of every 5 bets as if you lose your first 40 and then win the next 60

 

[Duritz]

I hope I'm misunderstanding the first sentence, or at least that you don't truly believe it, because that is just ridiculous (wins and losses obv come in completely random streaks, you're never 'due' for a win/loss). But more importantly, the second sentence is also false. Using Kelly, you'll be in the EXACT same spot after 100 bets if you consistently win exactly 3 out of every 5 bets as if you lose your first 40 and then win the next 60

No it ain't, there's cycles. Just because you THINK it's purely random, doesn't mean it is. It's cyclical because the things we can't control tend to happen in batches, like batches of unpredictable weather, that kind of thing. Of course you're never "due" for a win, etc, but runs of ins and outs DO occur. Take your head out of the book and bet in hundreds of thousands of events, then get back to me.

 

[Jan Juc]

Surely if things were truly cyclical in nature as you're suggesting, you'd be able to identify the start of a bad streak and take up knitting for the month, then when you know you're out of the woods (cyclical = seasonal, predictable, recurring, repeating) load up on a good stretch of "cyclical" conditions...

 

[Duritz]

Surely if things were truly cyclical in nature as you're suggesting, you'd be able to identify the start of a bad streak and take up knitting for the month, then when you know you're out of the woods (cyclical = seasonal, predictable, recurring, repeating) load up on a good stretch of "cyclical" conditions...

No it's not "truly" cyclical, as in it's not a perfect wave pattern, but it comes in clumps.

 

[Jan Juc]

So we've gone from cyclical to clumpy?


Seems plausible.

 

[Duritz]

So we've gone from cyclical to clumpy?


Seems plausible.

You're getting caught up in words, definitions. For reasons beyond our control, and usually predictive powers, winning periods happen and losing periods happen.

Consider this: a bad weather pattern comes through for a couple of weeks. This throws up surprise results, and you're on the losing end of them. Then, things level up weather wise, however the form is still haphazard because of horses (lets use horse racing) who ran in the strange weather period, and their form is less clear. Until that period clears through, you might struggle to win. This could lead to a bad month on the punt.

Consider this: your method likes certain types of horses. At the same time, a few horses who fit that bill find form. You have them shorter than the market generally, because they fit your method. If they find form and hold it, you win on them over and over again, while the market under-estimates them over and over again. So, you have a purple patch for a month or two, because some horses who are in form happen to fall into the criteria of horses you like.

Two examples only of how clumps of good and bad form can happen.

To suggest it's purely variance is to over simplify the true nature of what's happening out there, ie what the real life stuff is that we're betting with.

 

[brett128]

You're getting caught up in words, definitions. For reasons beyond our control, and usually predictive powers, winning periods happen and losing periods happen.

Consider this: a bad weather pattern comes through for a couple of weeks. This throws up surprise results, and you're on the losing end of them. Then, things level up weather wise, however the form is still haphazard because of horses (lets use horse racing) who ran in the strange weather period, and their form is less clear. Until that period clears through, you might struggle to win. This could lead to a bad month on the punt.

Consider this: your method likes certain types of horses. At the same time, a few horses who fit that bill find form. You have them shorter than the market generally, because they fit your method. If they find form and hold it, you win on them over and over again, while the market under-estimates them over and over again. So, you have a purple patch for a month or two, because some horses who are in form happen to fall into the criteria of horses you like.

Two examples only of how clumps of good and bad form can happen.

To suggest it's purely variance is to over simplify the true nature of what's happening out there, ie what the real life stuff is that we're betting with.

I agree with you, but I don't agree with you.

You concept of reality expressed as through streaks is blatantly wrong, as some people have already expressed. One thing I will say though is that I think you have a Bayesian logic without even knowing it (look it up, not a bad thing at all).

You're getting caught up in words, definitions. For reasons beyond our control, and usually predictive powers, winning periods happen and losing periods happen.

Here is where I sort of agree with you.

I'm overseas at the moment getting some data science/mining tuition from some very smart dudes to hone my sports modelling abilities. It may sound somewhat esoteric, but modeling certain sports with these guys make reality seem like a funny thing. Because that's what we're all essentially doing, we're trying to predict the course of a little piece of reality.

No matter how deep you go, or whatever method you use to drill into the solution space, there always seems to be something else you can do. One layer of numbers leads to another layer of numbers. Even if we don't 'drill into' but move across the solution space using various orthogonal methods, one finds other methods that also help to fit the bill in uncovering another piece of our little reality puzzle.

The way in which various methods work so readily in expressing this ability of reality to predict the future course of itself is somewhat eerie/haunting to me.

 

[d3ft]

I agree with you, but I don't agree with you.

You concept of reality expressed as through streaks is blatantly wrong, as some people have already expressed. One thing I will say though is that I think you have a Bayesian logic without even knowing it (look it up, not a bad thing at all).



Here is where I sort of agree with you.

I'm overseas at the moment getting some data science/mining tuition from some very smart dudes to hone my sports modelling abilities. It may sound somewhat esoteric, but modeling certain sports with these guys make reality seem like a funny thing. Because that's what we're all essentially doing, we're trying to predict the course of a little piece of reality.

No matter how deep you go, or whatever method you use to drill into the solution space, there always seems to be something else you can do. One layer of numbers leads to another layer of numbers. Even if we don't 'drill into' but move across the solution space using various orthogonal methods, one finds other methods that also help to fit the bill in uncovering another piece of our little reality puzzle.

The way in which various methods work so readily in expressing this ability of reality to predict the future course of itself is somewhat eerie/haunting to me.

Dude, you're guessing which bunch of 22 blokes is gonna beat the other blokes - this isn't astrophysics.

 

[Duritz]

A 200 game moving window of Mean Absolute Error (MAE) of the AFL line is 29.5 points.

brett128, a query here, because I've been delving deeper into my methods, etc... researching... re-inventing the wheel...

Given your comment above, if I develop an out of sample method with a MAE of 29.0 points... just under the market, would that be considered a profitable model?

 

[iluvparis]

Dude, you're guessing which bunch of 22 blokes is gonna beat the other blokes - this isn't astrophysics.

One of the funnier calls on this thread- gold

 

[brett128]

Dude, you're guessing which bunch of 22 blokes is gonna beat the other blokes - this isn't astrophysics.

You can delve as deep as you want to. If you want to limit yourself to scratching the surface, that's your decision. When the utility of that method becomes incorporated into the market and your win % drops off, you're back to square one. I'd rather prefer to be so far ahead in my methodology(ies), that we won't have to worry about this for a long time. Hence the reason why we devoted some winnings to getting me overseas for some cutting edge data mining tuition.