Kelly Re-Visited

[ChuckyTheGoat]

I haven't read beyond BSims' original post. Some basic thoughts on the topic:

* Kelly is a sound betting practice.
* Most experienced players will opt for a reduced Kelly bet-size, like 1/2-Kelly, to not be overly aggressive.
* The real catch-22 is that the HARDEST calculation for the sports-bettor is PROSPECTIVE win-rate. This isn't like dice or cards, which have fixed odds. The NEXT bet you make is speculative, and therefore it's extremely hard to pin-point your expected future win-rate.

BSims, per usual, your content provides good food-for-thought. Have a nice wkend, and Good Luck.

[turbobets]

I use turbo-kelly

Calculate the full kelly for your past wagers and average them. Make sure you have a decent sample size. Divide the full kelly for each new wager by the average kelly from your previous wagers. This normalizes every bet so that your average bet size will be 1% of your bankroll. Update after every bet.

[ChuckyTheGoat]

Bsims, I had a chance to digest this a bit more. My thoughts:

1) Posts in this thread are pretty good. Guys understand the concept + get the Risk vs Reward.
2) I'm 99% sure that link is a bit dated. JR Miller was a pretty big name in the sports-betting world about 25 years ago.
3) I DON'T think he was a winning player. He liked to post a lot of info, but I don't think he won. And he surely didn't win after lines got sharper.

[ChuckyTheGoat]

I'll echo some points that have been repeated by me or other posters in this thread:

1) Calculating your PROSPECTIVE edge in sports is extremely difficult. It's not the same as looking backwards and doing a calculation. And it's not the same as cards or dice.
2) One certain truth is that you can't OVER-BET your edge. Variance is a monster. Over-betting your edge means your ROR (Risk of Ruin) goes thru the roof. In some instances, over-betting virtually guarantees you're going to bust-out (to a 99% clip). Ed Thorp talks about Bankroll Management and advocates bet-sizing slightly less than your true edge.
3) Partial Kelly is surely a logical middle-ground. Bet a good chunk but allow for volatility.

[ChuckyTheGoat]

Bsims, I wanted to compliment you again on some of the topics you present in Think Tank. Very thought-provoking. Always good to be thinking + trying to carve out an edge.

I've got a potential thread for you. What about monitoring Expert Opinion? If we track respected bettors and note their results over time, can we build a profitable portfolio when sharp-minds line up. Take it one step further. If we note a game Sharps line up + we see Squares on other side...time to go to the window!

[tsty]

kellys risk of ruin is 0

it doesn't matter how small you make your bet if you cant beat the market

people also need to stop this half kelly nonsense since the only people who advocate for it dont really understand kelly well enough

bet size has no relation to variance and thinking it can lessen it is laughable

half kelly is for nits who hate money

[Bsims]

First off, thanks for the continued feedback. Most of it is quite helpful. I’ve completed my first simulation runs and have documented these on my last two posts on ole44bill.blogspot.com. I elected to simulate 1,000 seasons of 100 wagers each season. I’ve put the results in a spreadsheet that you can access at bit.ly/2MhtXbh. I’ve also started to look at the results and find something for everyone. There are no definitive results. If you like Kelly, you will find some supportive numbers. If not, there are numbers there to support that position. Following is a table of some raw counts.

Compare Ret/$ Overall	Kelly	Flat
Best Using Ret/$	38.6%	61.4%
Average Ret/$ Per Season	$1.027	$1.044
All Seasons Totals	Kelly	Flat
Average Kelly Pct	5.49%
Average Bet Size	$64.29	$100.00
Average Net Per Bet	$3.58	$4.44
Total Ret/$ All Seasons	$1.056	$1.044
Maximum Season Gain	$7,215.89	$3,179.00
Maximum Season Loss	-$855.20	-$2,169.00
Repeat Above for 55 Win Seasons	Kelly	Flat
Best Using Ret/$	29.5%	70.5%
Average Ret/$ Per Season	$1.027	$1.051
Average Kelly Pct	4.88%
Average Bet Size	$62.03	$100.00
Average Net Per Bet	$1.93	$5.05
Total Ret/$ All Seasons	$1.031	$1.051

From what I’ve seen, I’m not ready to draw a personal conclusion. But I do have some further investigations to do and more detailed analysis in the works. I’ll keep blogging and posting until I do make a personal choice.

[Bsims]

I've got a potential thread for you. What about monitoring Expert Opinion? If we track respected bettors and note their results over time, can we build a profitable portfolio when sharp-minds line up. Take it one step further. If we note a game Sharps line up + we see Squares on other side...time to go to the window!

I'm not sure where we would get the "Expert Opinions". I've done two similar projects in the past. One was at Covers.com where they had contests and you could see what the public picked. They also took the top 10% of the guys picking winners for each team and considered them team experts. They listed their picks. Later I subscribed to a Bettorsworld offer of basketball picks from professional touts. I faithfully tracked them all season.

The bottom line of both was that I didn't feel there were any experts. Sure some performed better than others. But I attributed that to normal variance. (Take a room full of monkeys and let them make picks. After a few weeks some will have picked better than others. Some indeed would be quite profitable. But I wouldn't bet on any of them.)

Chucky, if ypu have a better definition of the project let me know. Might be able to do something. Also, if you ever make it to Turning Stone this summer, let me know. I'll be making at least weekly trips there through September.

[Rich Boy]

Kelly works fine, as long as your predicted edge is roughly on target. You can have errors, but the average must be about correct.

Most people fail at this stage, drastically over-estimate their edge, and end up past the peak of the EG curve .... which has the added negative of absolutely horrendous bankroll volatility, which few can handle.

See below .....

Nice chart.

Ive always used kelly to maximize my expected growth. The website you reffered to Bsims is clearly an amateur writing it. I closed it after reading 1 paragraph. Anyone who doubts Kelly has no fundamental knowledge of calculus.

Full Kelly is not for the faint of heart but I find as long as you do have an edge you will make the most return. A more psychologically pleasing kelly ratio is 0.75 and you can capture most of the returns 1.0 kelly would achieve

I am actually using a kelly model for my stock trading system after years of use for parlays. Ganch had a few good threads back in the day about kelly. and his Kelly excel spreadsheet is fabulous.

[KVB]

... From what I’ve seen, I’m not ready to draw a personal conclusion. But I do have some further investigations to do and more detailed analysis in the works. I’ll keep blogging and posting until I do make a personal choice.

That's just it. While we can all agree on awful money management for sports betting, when it comes to reasonable management, it is a personal choice.

Everyone is different, even with different goals, as we are all the same in trying to win.

Some just can't handle the risk tolerance, and while some may say they shouldn't be betting (and I agree in some cases), it's not necessarily true.

Everything from your bet size to how many plays you make end up being a personal choice, even if guided by the numbers.

[KVB]

As far as compiling data, or "expert opinion" or ways to handicap, I have done something similar.

Instead of just looking at some posted plays, I'll set up different lines based on different levels of handicapping "sophistication" from what makes the opener to sharper lines that predict the movement an overwhelmingly majority of the time.

I can say this, when things line up among these "ways" to a high degree, things tend to go south for that bet.

Some of you modelers see this effect in games that are outlier predictions, showing huge blowouts or Total predictions. Often times, you'll find those outliers aren't as successful as the predictions say, they fail more often than, say, predictions closer to the mean or median, for lack of a better way of saying it.

Once again this relies heavily on the current market environment when the play was made and result obtained, something that is rarely tracked by bettors.

[Rich Boy]

kellys risk of ruin is 0

it doesn't matter how small you make your bet if you cant beat the market

people also need to stop this half kelly nonsense since the only people who advocate for it dont really understand kelly well enough

bet size has no relation to variance and thinking it can lessen it is laughable

half kelly is for nits who hate money

Ive been a successful Kelly player for years and I completely disagree.
While full Kelly will maximize return it is far from applicable.
We are humans, not robots, and while in theory "maximum bankroll growth" sounds nice. At what cost does it come?
Full kelly doesnt factor in stress, quality of life and human emotions one is to feel when on the roller coaster ride that is full kelly.

I am a gambler and risk taker at heart and I love the action of laying large amounts on plays that warrant it, but to recommend that to every gambler I wouldn't do.

[tsty]

bsim no offence but dont quit your day job

[KVB]

...Full kelly doesnt factor in stress, quality of life and human emotions one is to feel when on the roller coaster ride that is full kelly...

It's one thing to run the numbers in a vacuum, and assume we as numbers based players shouldn't have emotion. It's a good way to move the discussion along for many of us.

But not everyone is like us. Many have enough troubles with all the many things that can kill a bettor that don't involve play selection.

I just want readers to keep that in mind.

[tsty]

It's one thing to run the numbers in a vacuum, and assume we as numbers based players shouldn't have emotion. It's a good way to move the discussion along for many of us.

But not everyone is like us. Many have enough troubles with all the many things that can kill a bettor that don't involve play selection.

I just want readers to keep that in mind.

if your emotions get in the way then this isnt for you

[tsty]

also im fairly sure this a troll thread

its fairly evident from ur sim where u put a 1k bankroll and are betting 100 flat lol which one is superior

[Bsims]

also im fairly sure this a troll thread

its fairly evident from ur sim where u put a 1k bankroll and are betting 100 flat lol which one is superior

That's the reason I use the metric "Ret/$". It removes the issue of bet size.

[KVB]

if your emotions get in the way then this isnt for you

I completely understand this viewpoint and get the power of it.

But not everyone is cut from the same cloth and, again, they have different goals and time frames.

[ChuckyTheGoat]

I'm not sure where we would get the "Expert Opinions". I've done two similar projects in the past. One was at Covers.com where they had contests and you could see what the public picked. They also took the top 10% of the guys picking winners for each team and considered them team experts. They listed their picks. Later I subscribed to a Bettorsworld offer of basketball picks from professional touts. I faithfully tracked them all season.

The bottom line of both was that I didn't feel there were any experts. Sure some performed better than others. But I attributed that to normal variance. (Take a room full of monkeys and let them make picks. After a few weeks some will have picked better than others. Some indeed would be quite profitable. But I wouldn't bet on any of them.)

Chucky, if ypu have a better definition of the project let me know. Might be able to do something. Also, if you ever make it to Turning Stone this summer, let me know. I'll be making at least weekly trips there through September.

BSims, you're RIGHT. Very hard to find expert opinions, especially opinions that are publicly available. In some ways, it's easier to find Square opinions that you can feel confident fading.

Yeah, will keep you in mind for a meet-up. Probably no Turning Stone this summer. I do get to Florida sometimes. Will let you know when we might be able to meet up. Food + drinks on me.

[Bsims]

I’ve run across something a bit perplexing. I decided to modify the simulator and replace the random edge selection with a rotating one. This removes one random variable in each season’s wager simulations and only leaves one, the random decision on whether the wager won or lost (with a 55% chance of winning). I’ve uploaded the Excel log of the 1,000 seasons in a shareable file. You can access it at bit.ly/2twM3zp.

The log is sorted by the number of games won in a season. Notice in particular cases 276 and 682. In these season simulations, 61 wagers were won. That should make for a good season. In season 276, Kelly made a profit of $2,513.38 (Ret/$ of $1.201). But in season 682, Kelly only made $2.09 (Ret/$ of $1.000). So, what is the difference? It almost has to be in the sequence of the wins and losses. I’m going to look at the individual wagers in detail to see if I can learn anything. Hopefully, I’ll have an answer by tomorrow morning. Meanwhile, does anyone have an explanation?

[KVB]

Each time you change your bet size, you change the breakeven and increase vig costs.

While the order generally doesn't matter when betting a set percentage of the current bankroll (an example that is a fluctuating risk situation), there is a cost in vigorish that may serve to erode returns when using Kelly style betting. This may be different with different sequences.

I believe the system that is market knows this and it is one reason why I say the recent environment when the bets are made make a difference. I mentioned before it made a difference with Kelly.

[tsty]

Each time you change your bet size, you change the breakeven and increase vig costs.

While the order generally doesn't matter when betting a set percentage of the current bankroll (an example that is a fluctuating risk situation), there is a cost in vigorish that may serve to erode returns when using Kelly style betting. This may be different with different sequences.

I believe the system that is market knows this and it is one reason why I say the recent environment when the bets are made make a difference. I mentioned before it made a difference with Kelly.

why do you keep just posting utter dribble? I would say you're a troll but 20k posts

you have no penetrating idea what you are talking about yet type with so much confidence

you are a penetrating retard

[tsty]

That's the reason I use the metric "Ret/$". It removes the issue of bet size.

What are you talking about?

You have the kelly bet averaged at 5% where it seems to be around $60 but the flat bettor is betting 100?

Do you even understand what you're doing?

lets not even get into the other horseshit in that sim

[ChuckyTheGoat]

Bsims, I think tsty (above) is trying to say that he enjoys your work.

[Bsims]

Bsims, I think tsty (above) is trying to say that he enjoys your work.

Hopefully he is reading everyone's posts with an open mind and learning from them.

[oilcountry99]

tsty is a self certified kelly grandmaster

[tsty]

learning from people who cant even run a basic sim? lol

you guys are like grade scoolers questioning a university professor

[oilcountry99]

learning from people who cant even run a basic sim? lol

you guys are like grade scoolers questioning a university professor

At least I can spell.

[oilcountry99]

Sorry Bsims....I'm done cluttering up your thread.

[tsty]

http://www.aussportsbetting.com/guide/the-kelly-criterion/

cant get more reductive than this

it even has a time value section for all the dopey future bettors

[ChuckyTheGoat]

Hopefully he is reading everyone's posts with an open mind and learning from them.

Exactly. Have to respect everyone's opinion. We're here to learn.

[ChuckyTheGoat]

BSims, I'll take one step back + give my two cents. Compound interest is a monster. People don't quite grasp the power of compound-interest.

Kelly Criterion betting is really an application of Compound-interest. If one can harness the power of compound-interest, they will see their $$ grow at a very fast rate.

It is not easy to quantify the true probability of your NEXT bet. If you can, you're on your way. I view Kelly as a guide to making proper bet-size. Half-Kelly should be viewed as a way to not over-bet your edge. The REAL problem w/ all of this is that it's REALLY HARD to build a portfolio that's filled with a series of +EV bets.

[KVB]

why do you keep just posting utter dribble? I would say you're a troll but 20k posts

you have no penetrating idea what you are talking about yet type with so much confidence

you are a penetrating retard

I know, you might know math.

You just lack in the ability to apply it and it's frustrating to you.

Here's what I'm talking about.

At -110 you break even is about 52.4%.

That's winning 11 bets and losing 10. Because I'm not sure you get it tsty, that's 11 divided by 21.

It's actually 52.38% but we are doing what's called "rounding up." This break even is that far above 50% because of the vig. That's where the bet winner is shorted odds that become the books's commission.

Now, if you bet 5.5% of your bankroll, and re-adjust to 5.5% after every bet is settled, you can reach your break even of 52.4% (winning 11 bets and losing 10).

But no matter what order you put your wins and losses, after giving yourself 11 wins and 10 losses, the supposed break even, you will be down nearly 3% of your bankroll.

Every time, no matter the order.

After 110 wins and 100 losses betting the same way, you will be down nearly 25% of your bankroll, even though you hit your break even.

The erosion is cause by the vig and a change in your break even percentage.

Oh, and tsty, thanks for the link that tells us what we already know. Now if you're not going to contribute to the topic, then leave the think tank.

You should review the above dribble, because you show you don't understand the math when applied when you reveal you frustrations, even if you can use Google to look up the words Kelly Criterion.

[Waterstpub87]

I’ve run across something a bit perplexing. I decided to modify the simulator and replace the random edge selection with a rotating one. This removes one random variable in each season’s wager simulations and only leaves one, the random decision on whether the wager won or lost (with a 55% chance of winning). I’ve uploaded the Excel log of the 1,000 seasons in a shareable file. You can access it at bit.ly/2twM3zp.

The log is sorted by the number of games won in a season. Notice in particular cases 276 and 682. In these season simulations, 61 wagers were won. That should make for a good season. In season 276, Kelly made a profit of $2,513.38 (Ret/$ of $1.201). But in season 682, Kelly only made $2.09 (Ret/$ of $1.000). So, what is the difference? It almost has to be in the sequence of the wins and losses. I’m going to look at the individual wagers in detail to see if I can learn anything. Hopefully, I’ll have an answer by tomorrow morning. Meanwhile, does anyone have an explanation?

Haven't looked at your file, but it is a common thing when looking a series. What you are really measuring that the end is cumulative growth.

If you have early loses, it screws up the compound.

Say that you start with 1,000, betting between 1-5% of your roll. If you lose 5 straight at 5%, you end up at 750. To just get back up to to 1000, you have to generate a gain of 33%, which is going to be 6 straight wins or so with the compounding. If you won those 5 at the beginning instead, you are now betting 66.5 on your 5%, which allows you to grow quicker.

You see it alot in the financial world when people compare cumulative growth on mutual funds and stuff. If you look at a five year window, the people who did worse in year one never catch up, because you have to gain so much more to even break even, let alone catch the peer group.

I think this is similar to the point that KVB makes about risk bouncing around. If you consistently bet 1% of your roll per play, you are not going to be risking much different across the season. Maybe 100 becomes 90 or 110, maybe 120 or 80, but you are still in a relatively consistent range. Having more variation, between 1 and 5% dependent on where you are, risking much more, can cause you to bounce, so instead of a tight range, you could be in a range of $50 to $500, dependent much more of the path your bankroll has taken.

[KVB]

It may be hard to see when you are looking at the forest, but when you start to look at the trees of a given time span, especially surrounding higher volume events, you will see that the sports markets exploit what WaterSt. saying.

It's one reason I'm trying to lay out the difference in applying the math to the sports markets. They behave differently when you break down the mechanics of the market and follow the money.

Money and handicapping methods are constantly built up and taken down. One way to avoid the overconfidence that a "bull" market will instill is to keep that Kelly number in check.