[danshan11]

and just so everyone knows it and hears it, kelly sucks, use kelly to figure out your grocery bill or something but NEVER in sports betting, where you never truly know your edge.

[TommieGunshot]

and just so everyone knows it and hears it, kelly sucks, use kelly to figure out your grocery bill or something but NEVER in sports betting, where you never truly know your edge.

Kelly is considered maximally aggressive. It's great, because the formula 0.5 x Kelly Criterion mitigates both the effect of going bust and missing an exact estimate of an advantage. In general all that needs to be known is that the edge is positive.

[danshan11]

Kelly is considered maximally aggressive. It's great, because the formula 0.5 x Kelly Criterion mitigates both the effect of going bust and missing an exact estimate of an advantage. In general all that needs to be known is that the edge is positive.

lets agree to disagree!

[Larkman]

and just so everyone knows it and hears it, kelly sucks, use kelly to figure out your grocery bill or something but NEVER in sports betting, where you never truly know your edge.

Why do you think no-one can know their edge?

[danshan11]

Why do you think no-one can know their edge?

if we use the bet line to closing line, we can assume that is our edge but as we have heard many times, the closing line is not ALWAYS truly efficient.
if we use our edge over the last 10 zillion results, that is giving us our average edge but does not pertain to that particular bet, so this I would say is not reliable either on each particular game.

overall our edge is X but that does not make each individual game X

[Larkman]

if we use the bet line to closing line, we can assume that is our edge but as we have heard many times, the closing line is not ALWAYS truly efficient.
if we use our edge over the last 10 zillion results, that is giving us our average edge but does not pertain to that particular bet, so this I would say is not reliable either on each particular game.

overall our edge is X but that does not make each individual game X

Lets step away from gambling for a moment. When meteorologists create a model to predict the chances of it raining in a given time period, how do they know how accurate the output probabilities are? After all, there's no closing line for the weather. Are they left hopelessly shrugging their shoulders, entirely unable to judge whether one model's predictions are more accurate than another's? Or do they have rigorous statistical techniques and metrics that allow for the accuracy of a given set of predictions to be measured? And if fields studying something as complex and chaotic as the weather can manage it, what makes you think its impossible for sport?

It seems like where you go wrong, if you don't mind me saying, is that you are wedded to this axiom of closing lines being the most accurate possible representation of the true probabilities, and they aren't, they are entirely beatable.

[danshan11]

because you and your friends believe that you are smarter than the entire betting market and can create models more efficient than the entire betting world is hilarious.

People might drink your kool-aid but I never have.

I stated clearly that the closing line is not efficient in every particular game. I truly believe the line is not efficient in every game and this again causes a problem with kelly because if the line is not ALWAYS efficient than what can you use to determine your edge.

if you use your average edge over a million bets that is as flawed as the closing line dream, avg edge is exactly that its an average edge but it is not indicative of each INDIVIDUAL game.

therefore my opinion is if you dont know your actual edge on that specific game how can you use kelly correctly to determine the actual edge you have on that particular game. at best kelly is an estimate of your edge and adds an unnecessary dimension of randomness to your betting experience. there are enough random things in betting as it is to add another one just for fun or just cause some asshole said it makes sense 49 years ago before excel proves it is useless.

kelly backtesting is impossible unless you got the actual line you bet at and even if you did variance could make the results not accurate.

I have not heard an argument yet that can state why Kelly is a good thing

people scream the line is efficient and they have data to support it, nonefficient line people dont have any data to support that, but maybe ????

people say kelly is great but they dont have any data anywhere to support that opinion.

I understand why so many people believe they can "pick winners" and line does not matter cause they seen a guy do it but I seen a guy make a lady disappear at circus circus and I still dont believe that people can make people disappear, GET CLEAR! get wise!

[Vyasports]

I never use it, instead of "helping" you it will only confuse you or add more tasks on your plate. The only way lines can be efficient is when you already know the results....

[KVB]

Fully Kelly is dangerous in sports betting and I seriously doubt anyone is successfully applying the math over a longer term to most sports markets.

It's that nature of the sports markets in particular that make this true.

That said, in my opinion, a fractional Kelly system can be applicable depending on what your handicapping uncovers.

Good Luck everyone.

[tsty]

Fully Kelly is dangerous in sports betting and I seriously doubt anyone is successfully applying the math over a longer term to most sports markets.

It's that nature of the sports markets in particular that make this true.

That said, in my opinion, a fractional Kelly system can be applicable depending on what your handicapping uncovers.

Good Luck everyone.

You would be wrong. There are no arguments for using half kelly instead of full kelly besides emotion.

Making less money while also still having the same risk of ruin is absolutely retarded

[Vyasports]

emotions...feelings....intuitions....bet ting....

[KVB]

I understand the argument, but very few if any are using Full Kelly when betting and never neglect the potential emotional aspect.

It's not just emotion, there's a reason and it has to do with sports betting in particular.

If you have been for an extended period, that is going full Kelly, then good for you. My doubts that anyone is doing it can be wrong.

But I'm not having a discussion of argument or even theory here, I'm talking about applied math in sports betting and probably should qualify with having some sort of plan draw as some income.

A long term account that is never touched would even be unlikely, but could exist.

Regular, long term Full Kelly users sign in and be honest. Are you drawing down as a plan, as a busness, or even at certain thresholds?

Who's out there?

[danshan11]

my "opinion" is kelly adds a dimension of risk that is not necessary. I dont think Kelly is designed for estimated edges.
Closing line is the closest to true probability but its only an estimate of probability.

to use kelly effectively I think you need to know the true edge.

I dont know if a person has a model that is more accurate than the closing line but I think that would be really white rhino rare. and even if they did have a more accurate model than the closer it is still an estimate and is not true probability.

and if you are using your average edge from history that definitely is not a good estimate of the exact game you are calculating using kelly.

Just vote no on Kelly!

[HeeeHAWWWW]

my "opinion" is kelly adds a dimension of risk that is not necessary.

Kelly optimises expected growth ... hardly unnecessary.

It's also entirely robust against errors, as long as events are independent and your mean error is zero. Breach those and it can be damaging, but most formulae have preconditions, that's nothing unusual

[danshan11]

Kelly optimises expected growth ... hardly unnecessary.

It's also entirely robust against errors, as long as events are independent and your mean error is zero. Breach those and it can be damaging, but most formulae have preconditions, that's nothing unusual

sounds like a guy like you can handle Kelly and use it to his benefit, now a dumb guy like me who doesnt even know how to calculate his edge or just uses some random edge based on an article I read from a "smart tout" might struggle with it.

you give kelly to a dumb person like me and that is like giving a scalpel to a 4-year-old, very good important tool perfect the best nothing better in a surgeon's hands but in the 4-year-olds hands, not so much.

Vote NO on Kelly!

[danshan11]

I run an api 24 7 placing bets and I tried both flat and fixed profit staking methods and they came out darn near identical in actual ROI.

flat betting disregard the 936,1936, 1.07% and 1.21% those are api parameters

W	936	1963	L	P	1.21%	Net
486	2.71%	8.392	435	15	52.77%	4609.35

and here is fixed profit betting

W	915	1963	L	P	1.07%	Net
479	2.47%	8.867	423	13	53.10%	6876.09

these are actual bets not backtested
you remove the slightly higher win% and you get darn near same result avg line was 49.85% on fixed profit and 49.29% on flat

[Vyasports]

thanks OP

[TommieGunshot]

I dont think Kelly is designed for estimated edges.

It sounds like the real issue is poorly estimating edges. As the estimates become better, Kelly Criterion becomes better at maximizing profits

[Gaze73]

What do you think about Jeff and Davidgoliath in the tennis subforum? Both seem to be tennis sharps and they make decent profits despite the lines barely ever moving in their favor. Is it just survivorship bias in both cases or perhaps they have an actual edge in picking winners because they're tennis fanatics?

As for knowing your edge on individual games, you can get close enough when you get into specific categories.
E.g. if you have 1000 NBA games under your belt at 5% roi, you really don't know if taking Bulls +6 is a 5% roi play. BUT if you have a sample size of 50 6-point home dogs running at 5% roi and Bulls are a 6 point home dog tonight, you can safely assume they're in the 5% ballpark. In practice you don't need to be so hyper specific but you get the point. Similar teams at similar odds in similar situations produce similar results.

[danshan11]

What do you think about Jeff and Davidgoliath in the tennis subforum? Both seem to be tennis sharps and they make decent profits despite the lines barely ever moving in their favor. Is it just survivorship bias in both cases or perhaps they have an actual edge in picking winners because they're tennis fanatics?
As for knowing your edge on individual games, you can get close enough when you get into specific categories.
E.g. if you have 1000 NBA games under your belt at 5% roi, you really don't know if taking Bulls +6 is a 5% roi play. BUT if you have a sample size of 50 6-point home dogs running at 5% roi and Bulls are a 6 point home dog tonight, you can safely assume they're in the 5% ballpark. In practice you don't need to be so hyper specific but you get the point. Similar teams at similar odds in similar situations produce similar results.

Why do you think the line is wrong in Tennis? How are they able to win but not beat the line? what is your opinion on what causes their success? How do you think books can survive if their tennis lines are wrong?

you use the term close enough for estimating an edge, I searched through kelly research articles and I never found the exact or any similar variation of the term close enough.

it kinda sounds like you think average efficiency is a good guide in staking but not a good guide in closing line efficiency? which way is it ?

[danshan11]

It sounds like the real issue is poorly estimating edges. As the estimates become better, Kelly Criterion becomes better at maximizing profits

very true, I dont see the necessity of adding an extra variable of what if to an already very what if situation. I state the same thing if someone can create a line more efficient than the closing line, I would say that person MIGHT want to use kelly. I personally have never met anyone that can do that. you check the prediction tracking and you can see none of the popular models even come close to the accuracy of the closing line but yes it is possible someone could have a more predictive model than the closing line I just have never seen it.

even the closing line which is very accurate we still can agree that it is not always efficient.

does someone here have a model that is more efficient than the closing line and if you do please chime in, I would love to know this exists

[danshan11]

It sounds like the real issue is poorly estimating edges. When the estimates get better, Kelly Criterion becomes better at maximizing profits

I also think staking methods, in general, is just another excuse to try and find profit when a person does not genuinely have an edge.

bettor A flat bets but he can pick winners, but he ends up losing over a large sample size because he had no edge
bettor A than says well maybe it was my staking method that caused my failure still loses
bettor A than says well I did not lose 10 in a row so I can use martingale and eventually he loses
the problem with bettor A is not the staking method or bad run it's he never had an edge to begin with.

a guy with a true edge could flat bet, fixed profit bet , fractional kelly or whatever normal method he wants would still win long term.

people design staking methods not to maximize winning but to try and solve losing and if you dont have an edge the staking method wont matter long term.

and back to the winning tennis cappers, man I would hate and love to be them
love to be them so I can win!!!!!
hate to be them because they have no danger gauge and no one knows when they will lose, AKA "CLV I am good gauge". I mean how can someone see an edge if they are not getting confirmation from overall beating the line?

[danshan11]

oh I forgot VOTE NO on Kelly!

[Gaze73]

Tennis lines are wrong all the time, apparently both the bookies and the market are clueless. E.g. A player opens at -110 and the market pushes him all day up into +160. Why? Did the player fall down a staircase during the day and the market reacted step by step? Conversely, sometimes the market gives strong support to outclassed dogs for some dumb reason and they lose 1:6 2:6 as expected.

Close enough is close enough, the world won't end if you don't bet 100.00% according to kelly. Sometimes you overbet by 1% and sometimes you underbet by 1% but it doesn't matter in the long run.

[danshan11]

Can you show me an actual example of a major market tennis match where the line moved from -110 to +160. I think you are thinking of 250 limit challengers.

If close enough is close enough and you really don't know your edge and it will all even out, why not just flat bet then? I mean Kelly is meant to get more money on your higher edge plays and less on the lower edge plays, right?

[Gaze73]

Can you show me an actual example of a major market tennis match where the line moved from -110 to +160. I think you are thinking of 250 limit challengers.

If close enough is close enough and you really don't know your edge and it will all even out, why not just flat bet then? I mean Kelly is meant to get more money on your higher edge plays and less on the lower edge plays, right?

https://www.oddsportal.com/tennis/us...illy-KSBm3Eb3/
https://www.oddsportal.com/tennis/br...erto-bsVD6eJn/

I do have high confidence plays which historically do much better than medium confidence ones so I can stake more on those.

[danshan11]

curious what exactly is a high confidence play? can you give me an example of what a high confidence play looks like, I am very curious

[Gaze73]

curious what exactly is a high confidence play? can you give me an example of what a high confidence play looks like, I am very curious

Depends on the sport. I haven't had a high confidence tennis play in 5 days but I have 2 basketball dogs today at 66% to cover the spread.

[danshan11]

you have a team that is 66% to cover the spread? so the world has 50% to cover the spread and you have 66%? do you find that as odd?

[danshan11]

lets say the line was -5 on every book at even money and one book had it at -1.5 what do you think would happen, that is about the same thing as a 66% chance to cover the spread in basketball

[danshan11]

Gaze when you say crazy things like that. anyone with an inch of skin in the game knows you are talking crazy and do not have a clue how to model or cap. no one whoever does this for a living or seriously ever has 66% on a 50-50 coin flip game, they might have 55% chance and that is probably rare. this is a game of inches

[Gaze73]

lets say the line was -5 on every book at even money and one book had it at -1.5 what do you think would happen, that is about the same thing as a 66% chance to cover the spread in basketball

Won't happen because every book copies pinnacle. Everyone but me is clueless.

[danshan11]

yes wont happen correct cause it never is that far off but just for a conversation say it did, what would happen at that book that had it at -1.5 if it was a reputable real book?

[Gaze73]

One of my high confidence plays Darussafaka ML just won at +283 by 5 points - clueless clown market, nobody believed them like the sheeple they are. Nobody with a clue bets on -300 favs.

[danshan11]

One of my high confidence plays Darussafaka ML just won at +283 by 5 points - clueless clown market, nobody believed them like the sheeple they are. Nobody with a clue bets on -300 favs.

why would only fools bet on -300 faves?

it seems -300 faves win at about the exact rate of +100

line < -299 SU: 218-74 (2.39, 74.7%) avg line: -329.3 / 278.5 on / against: -$2,375 / -$1,470 ROI: -2.5% / -5.0% RL: 118-85 (1.14, 58.1%) avg line: -150.7 / 135.5 on / against: -$1,201 / -$152 ROI: -3.9% / -0.7% OU: 131-151-10 (0.42, 46.5%) avg total: 8.5 over / under: -$3,386 / +$867 ROI: -10.5% / +2.7

line < 101 and line > -101
SU: 1215-1233 (-0.13, 49.6%) avg line: -100.0 / -110.2 on / against: -$1,800 / -$10,508 ROI: -0.7% / -3.9% RL: 1149-749 (1.04, 60.5%) avg line: -166.5 / 148.3 on / against: -$4,635 / -$7,067 ROI: -1.4% / -3.3% OU: 1139-1191-112 (0.40, 48.9%) avg total: 8.6 over / under: -$16,526 / -$6,049 ROI: -6.2% / -2.2%

nothing here seems to indicate that there is any difference for either bettor

now people who bet super big dogs are subjects to fave longshot bias and get hit bigtime

line > 280
SU: 19-85 (-3.22, 18.3%) avg line: 311.6 / -360.7 on / against: -$2,555 / +$1,575 ROI: -24.6% / +4.2% RL: 34-47 (-1.65, 42.0%) avg line: 145.3 / -162.6 on / against: +$227 / -$829 ROI: +2.8% / -6.3% OU: 46-55-3 (0.99, 45.5%) avg total: 8.2 over / under: -$1,424 / +$510 ROI: -12.4% / +4.5%