Arbusers

kinda long post ahead

Hi guys, im new to this site, and im really impressed by the community, so i wanted to join it and learn a few things. So heres my story:

Im new to valuebetting, and overall sportsbetting, and ive started this value strategy, you already know the drill. Im still doing tests, and im currently just finding valuebets, and tracking their outcomes. Now there are a few things to consider:

1. I know that betting in theory and betting in real life arent the same. Currently, my goal is to create a big database, from which i will draw some conclusions, and eventualy start the actual betting with real money.

2. I know all those things about account limitations and other bad things, and ive got all of that covered in more ways than one. I have close people that do this for a living (arbitrage mainly), and im building this system so we can work together. My part is the actual system and strategy, theirs is the excecution, so dont take those things into consideration, because im in a really specific situation, and actually got that covered.

Okay, now lets get into the betting part, the fundamental ideas for my system are these:

Odds that has a value of 10%+, compared to sharp bookies, and are in the range of 1 and 3.

I have noticed that some people use different methods with the juice and all that, but im using a different one (or maybe the same idk really). I use reference odds from this site, and then compare to actual odds from the soft bookies i play.

The thing that bugs me the most, currently, is the stake model. I dont know which one should i follow. With a constant stake (1), these are my overall results after 360-ish bets:

https://ibb.co/3THpMFM

Thats a profit of 19 bets on 360. Is that good? Bad? Thats 5.26% per game. By the way, my current odd average is 52.23% and my wins are 52.05%.

Currently, my best interval, is 15%+ (1.5 , 2.5), with a return of 13.41% on 100 games :

https://ibb.co/frzn8xY

The odd average in this range is 50.45% and my wins are 50.52%.

I think that currently, i should improve my stake model, i dont really know what do to with. What are your tips? I get that variance plays a big role currently, but anyways i want to try new things now rather than later. My plan is to get to 1000 games, and then draw conclusions. Feel free to ask further questions, i will be glad to answer them.

How should i even track my bets? What numbers should i look into? How do i calculate my profits when the stake isnt the same? Should i use Kelly for stakes? Thanks in advance, im excited to hear your opinions. I would really like to have a chat with someone that has built strategies before. Have a nice day

@Marko Kostic,
Welcome to the Smart community. Are you willing to accept some material from Arbusers? Maybe these ones will help you if you read them.

Now regarding your post above:
Large postings are extremely helpful for thinking upon several ideas, at the same time and therefore they're short expressions on complex ideas.
On the other hand, large postings indirectly share lots of hints (nobody will reveal their system, nor sold it if that one is pretty OK), which may be helpful for users, willing to evolute in the business. The Knowledge is a Power.

I'm glad you're creating a database. The bigger the sample, the better one. Some theories work upon the hypothesis, that you have an endless database, and the given advantage if spotted by you, is applied upon the endless time frame. But in the reality, that's not possible that way and therefore the differences between theory and practice arise.

I'm glad to hear about your value of circa 10 percent and above. The graphics you've applied are very good in my eyes if they're applicable in a very long time, a very big amount of your account, and a proven value.
Try to process these graphics via EMA, SMA-like approach, like Forex, and you may see some things.

Regarding the value, caught by you - you have your way. I personally determine my own odds, and further try to estimate the value. I personally don't use other odds for that estimation, neither sharp nor soft agencies' odds. There are many reasons for that, the value traps and variable margins are among them. They give a negative value to us in terms of maths. Do you remember Pinny's margins 10 years ago? Nowadays? In the liquidity markets? In the non-liquidity ones?

Regarding the first graphic, attached by you, you'll be pretty OK in the long run if the value over the period is much over 1,93 EU odds / -108 American odds AT THE SAME CONDITIONS which produced these results - if these conditions are unchanged in the future - staking plan, money flow, the market conditions, the hidden conditions, liquidity, drawdowns - frequency and magnitude, also these for the winning peaks, fundamental basis, conditions on/out of the pitch, lack of "black swans" (read the genius Mr. N.N.Taleb) and many, many more quality and quantity factors.
Regarding the second graphic, attached by you, you'll be pretty fine if much over 1.98 EU/ -102 American, respectively at the same unchanged conditions, for the future period.

But they aren't the same in other periods. Things always change. That's for sure... That may be the first threat to you.
Other threats are "percent win per game" and "percent return for a given interval" you've mentioned. I hope, you treat these for statistics processing purposes only and you seriously estimate the conditions for these achievements. The same future achievements are not guaranteed, don't pursue these levels at any cost. If you do that, may worsen the risk/reward ratio.

Regarding Kelly/RalphVince, these are very good starting points for building your model but have to optimize these according to your strategies. Profit geometric optimization has some risks, such as above the average risk of decapitalization for a given period, if wrong estimations are done, such as "advantage" parameter in Kelly formula/ Vince's "Optimal F" parameter, such as wrong "mean" parameter in statistical distribution, or if many losing streaks happen before winning ones although the "advantage" parameter computing is correct, and so forth.
As an initial point, if you have no very big experience managing the staking, use a pure flat strategy, safely very low stakes. A fixed amount (not fixed percent of the bank!) every time, the bank amount no matter here.
On one hand, you'll avoid bigger losses if Kelly/Vince is improperly implemented. So you'll see clearly the real advantage, not the computed, and on the other hand, the sequence of winning/losing streaks has no influence here. So, if you have a real advantage, you'll be rewarded. But if the "advance" is real, the winning amount over the long haul is much lesser than if achieved by Geometric Optimisation methods.
Later, you may compare several money management strategies upon given circumstances, and serious backtests. You'll see which fits you. You even may modify the given MM to your own MM.

Regarding your question about mean/variance, read Harry Markowitz, but work on your own implementation, maybe the pure theory is not 100 percent applicable.

And some words for the statistical distribution. Only you may apply some statistical distributions for the data processing regarding the data samples, collected by you. For small samples of data, and when the "sigma" parameter is not clear, Student's distribution is recommended (if applicable). Later Normal distribution etc. Naturally, at given circumstances may apply a given kind of statistical distribution, determined by your model and so forth.
Good luck!

arb12 wrote: ↑

Wed Feb 09, 2022 4:38 pm

@Marko Kostic,
Welcome to the Smart community. Are you willing to accept some material from Arbusers? Maybe these ones will help you if you read them.

I feel very welcomed by this reply on my post

Everyday I learn something new, and the idea is constantly evolving. Yesterday i read about the compounding stake method, and it seems very good i must say. Tho it kinda solves some problems, that really arent the ones i have. If your bank is a small one, indeed, the compounding method is phenomenal, but yeah, besides that, it wouldnt serve the purpose i need.

For now, the plan is to just build a database, and hopefully make some conclusions. Im actually going to start betting in practice, but with smaller stakes. Theory is very important, but practice makes perfect . There are really a lot of methods, and ive got a feeling in my guts that my method is possible, and would make great results. But obviously ive got a long way ahead of me. Anyways, here are some conclusions i came up with. Im pretty sure im not the first person to come upon these ideas, and im not sure whether theyre even true. All of them are from a math standpoint.

Variance

-The thing you hate the most if youre valuebetting. Obviously you overcome it by placing more bets, but that isnt efficient past some point, in time, and both profit terms. The first solution is to just play less games, that are more specific, so they will even out more often, and finally, make you positive profit, which is the ultimate goal. Okay, but now we have a new problem emerging, how the hell can you make profit, if the fundamental ideas of this method are actually the opposite, playing more. Thats the basic problem, and ive came up with a hypotethical solution, which isnt actually a solution, just an idea i will work on.

Lets start with an ideal situation. The good old infinite amount of games, we would all love to have, right. If you played stakes on only single value bets, lets say 2.00, with a value of 10%, in theory, every two games you wouldve gotten a win and a loss, which would increase your profit 10% per two bets. Now, in practice, in my spreadsheet, there are 400 odds, and there is a very small number of odds with the same value, so im not gonna get the maximum profit until i reach a stupidly high number of plays. Well, what if, in the same way we created an interval of play (1.5-2.5, or whatever), that obviously works, we created even more intervals, and basically treat different odds as the same valued ones, and destroy variance, in a profit generating manner. Lets actually investigate the two extreme sides of this problem.

If you have not made an interval, your system would rely mostly on luck, and the number of games.

On the other hand, if you try and make an infinite amount of intervals, you would never reach a number when you actually even out (because its infinity).

So, the solution is certainly somewhere between these two points, and maybe its simpler than it seems. I have a few ideas on how to solve this problem. Not actually, because this isnt really a solvable problem, just an ideal to chase after.

The idea is simple, and i will try to explain it. Basically, you are chasing odds in little intervals, not really chasing, but betting that way. The hypothetical intervals are not to small, but not too big, because of varience as i explained before. This seems like an upgraded flat stake method kinda. You play the same bets on, lets say 1.20-1.40 , 2.00-2.20, and so on. In theory (or just my mind) the numbers have bigger room to grow than if you played a flat stake, because you make these fake intervals, but also than if you played flat gains, beacuse it lowers the number of games problem.

I really do believe that this is something you can chase, and most certanly will try to do so. The whole idea revolves also around the flat stake or flat gain method, so yeah, theres that.

Im really mad now because the thing we can agree on that is true, is this:

Large postings are extremely helpful for thinking upon several ideas, at the same time and therefore they're short expressions on complex ideas.

I kinda let my mind flow with this reply, and that im too mad about, maybe ill make an another post, cheers!

arb12 wrote: ↑

Wed Feb 09, 2022 4:38 pm

But if the "advance" is real, the winning amount over the long haul is much lesser than if achieved by Geometric Optimisation methods.

I dont really get the message, can you please explain it to me in simpler terms? Im not really familiar with math terms in english . Youre saying that alternative methods to the flat stakes are more worthy and profitable, is that correct? Anyways thats my general opinion on the subject of stakes.

Do you remember Pinny's margins 10 years ago? Nowadays? In the liquidity markets? In the non-liquidity ones?

Im really new to this, and even if i werent, I was in middle school at the time anyways .

Other threats are "percent win per game" and "percent return for a given interval" you've mentioned. I hope, you treat these for statistics processing purposes only and you seriously estimate the conditions for these achievements. The same future achievements are not guaranteed, don't pursue these levels at any cost.

Im glad you told me this, I kinda was lost with the results, and kinda didnt know what to chase. Only now i get that chasing profit isnt reallly the way, its more about chasing the value of your bets and actually maximising it.

Anyways, for statistic and reference purposes, what are good returns per bet? Only so i know if im doing things the right way or not.

I get that my methods of finding valuebets arent maybe the best, obviosly, but im sure that this can work in some way, because in a way, the wrongs will surely cancel eachother out. At least i can say, after some games, that they in fact do in their own way.

Im going to start placing actual bets really soon, and im going to post results hopefully.

Nevertheless thanks for your in depht reply.

marko kostic wrote: What numbers should i look into? How do i calculate my profits when the stake isnt the same? Should i use Kelly for stakes?

Congratulations on your dedicated work so far.

It is no surprise that the more data you have, the better. But then, enough data is never enough, so you will have to make some probabilistic assumptions.

I would ask myself the following question: What are the chances that this is either:
A: a 10% EV sequence (reference is the only truth) which unluckily performed only 5.26%
B: a 0% EV sequence (reference is no better than soft) which luckily performed 5.26%
C: a minus 10% EV sequence (soft is right, reference is nonsense) which luckily performed 5.26%

How? run the the entire population of bets (360ish) a 1000+ times through a simulation (Monte Carlo) with the EVs above. Then you will get a probability distribution of outcomes. If the distribution is sufficiently narrow, you should be able to make some conclusions. At the end of the day, you will never be convinced 100%, but it will probably guide you to a certain degree of faith that your system works.

At a later stage you might try to split the population according to: market liquidity, sport type, time before game start, day of the week, league, type of bet (ML or other), etc.

Register as much data as possible, it will become useful one day.

With regards to stake sizes, you can use theoretical stake sizes just for analysis, getting a feel for it. That goes with Kelly as well. For instance, when I applied the optimal Kelly fraction on my entire history of bets, I found that I would have had substantial bigger profits, but the swings would have been intolerable.

@Marko Kostic,
I'll reply here to cases, mentioned by you in your previous two postings.

“Compound interest is the most powerful force in the universe.” - Albert Einstein. No matter the bank amount. But if work in that manner, due to practical reasons, always withdraw some of your profits before continuing to compound the amount.

If you choose value betting approach, the dispersion and the variance aren't avoidable. In fact, they may add experience to your evolution, don't you think? But have to distinguish when to change value approach to trading, to hedging, and later to take positions in a value betting manner, or somethin' mixed.

Regardin' your ideas about "interval breakdown of the odds and further processing", I suppose if I get you right, that you have in mind proportional added value in every odds interval and how to handle it, provided that methods, similar to the flat betting approach are in use by you. You can handle the value if exist in every odds range, but the variance and dispersion in larger data streaks always appear regarding the probability theory and the Law of the Big Numbers, and have to be handled generally by the methods, applied in the statistics. I don't tell you not to try your ideas! You may invent some unbelievable achievements in the future. Just use the evolution in the "Applied statistics" discipline so far and contribute later, if you may.

In your question regarding "avance", that is my mistake,I've meant "advantage". I don't need to hurry up when typing and to make mistakes, excuse me.
That's an "advantage (real value)" in Geometric Optimisation Methods. Did you check your e-mail several days ago? Maybe there are some 20 materials.
It is possible to use the arithmetic "mean" parameter, when computing Geometric optimization and that's wonderful!
Please, read Ralph Vince's literature, Kelly's literature and you will figure it out much clearer. In general, when you use geometric optimization of the winnings, in the very long haul, you'll be much happier, but the cost in the short run is run-out of a much bigger amount of the capital than to the flat betting approach!
For instance, did you remember the match Crvena Zvezda vs Napredak (1:1 at the end), held in March 2018, priced in the range of odds 1.1 - 1,15 for the home win? If you calculate advantage 7 percent, or 15 percent, or another advantage, how much capital you will lose, compared to the flat approach? Or Rad Beograd vs Crvena Zvezda, 2:2, May 2021? Or, if you achieve most of the cases much more Winning bets, vs Lost ones, these matches will be only small Operative costs, giving you experience, and due to Geometric Optimisation, you will win much more, than if you apply a flat approach? You see how complex is that case, and some wisdom is needed...Some experience gives us some wisdom...



Similar are VidaBlue's observations. That user has published postings, among the smartest. He hints, that every event may have a different Expected value, negative or positive, while you treat all the events in your spreadsheet as they have equal EV. Follow his advice. Suppose, you know that Monte Carlo simulations are used in the Manhattan project? I'll only add to his advice toward you, much more Monte Carlo trials, such as 10 000, 100 000 or something.
Monte Carlo trials may be executed via Excel or other software.



Regardin' your question about Pinnacle sports/Pin88/Pinbet/PS3838 margins, please read carefully how the agencies make their profits, no matter the event outcome - bookmakers margin/juice/vigorish. After you read that, log-in to your portfolio of the bookies, and see what is the liquidity of the English Premier soccer, and what is this one, let's see on the Romanian Basketball. Further, compare the margins and market behavior upon these circumstances in vast situations.
For your info, several years ago, circa 2014-2015, that major bookie upon new ownership worsened the margins...
And last, but not least, after you read about Kelly Criterion, Ralph Vince's "Optimal F" theory, please read "Long term secrets to short term trading" by Larry Williams, and "The trading game" by Ryan Jones.
Until you read carefully these, restrain to bet for real!
By the way, there is a piece of software, called TSM (The Staking Machine), when you can try many staking methods, but that software can't give you real value on every odds and market! That's your task only!

"Staking plans comparisons", type that in google search images, you will see some things. But have in mind, any bet has a different Expected value, while these images are built under the same-expected-value method. Do you understand what are the traps in front of you???

marko kostic wrote: ↑

Fri Feb 11, 2022 2:42 am

Variance

-The thing you hate the most if youre valuebetting. Obviously you overcome it by placing more bets, but that isnt efficient past some point, in time, and both profit terms. The first solution is to just play less games, that are more specific, so they will even out more often, and finally, make you positive profit, which is the ultimate goal. Okay, but now we have a new problem emerging, how the hell can you make profit, if the fundamental ideas of this method are actually the opposite, playing more. Thats the basic problem, and ive came up with a hypotethical solution, which isnt actually a solution, just an idea i will work on.

My opinion on that - A quality filter you must apply in every position on the market. Experience is needed. Real, proven value, added to positions, is OK only. Therefore the positions aren't so many like in theory. On the other hand, the theory teaches us, that more and more positions, in the long run, give us the probability to overcome the losses, provided that they're containing a real value. But when participating in the market by proven value positions only (filtered by your criteria), you may have a profit in the beginning, and you're not required to add more entry points, except proven value positions. No need to hurry up, and therefore avoid other dangers - overtrading and mind-burning are among them.

When re-assessment of the position like probably loser one, beyond predefined stop loss, causing big harm to you, immediately cancel that position (hedge) for a small loss. Keep only the winning ones in the time-line to grow. In the end, systematic approach will give you profit, equal to the real achieved value (not the computed one!) minus the operative expences.
Warren Buffett once said, “The first rule of an investment is don’t lose [money]. And the second rule of an investment is don’t forget the first rule. And that’s all the rules there are.”

Also, at the end of the trading day, willing the stressless transition from working day to the personal life moments, for mind-health purposes, need to stop these and do not add positions, always will be tomorrow, always will be more games to profit etc.

In short - be cautious!

VidaBlue wrote: ↑

Sun Feb 13, 2022 12:34 pm

marko kostic wrote: What numbers should i look into? How do i calculate my profits when the stake isnt the same? Should i use Kelly for stakes?

Congratulations on your dedicated work so far.

So heres a little update. Im currently at 450, yesterday i worked as if it was my last day. Also the big news is that ive actually have spoken to someone whos professional work is building and managing databases, and he is willing to help me out on my journey! Im very excited about that, because, no matter what, we will draw conclusions from the results, and i will constantly update the data. That is the plan, and im afraid, even when i start actually playing, i will have to continue with this work.

On how many different parameters should he build the database? I mean he could obviosly build it on how many he wants, but what should i chase when i choose bets? My idea goes like this:
SOFT BOOKIE ODD
SHARP ODD
VALUE
SPORT
COUNTRY
LEAGUE
MEN/WOMEN

I dont really know if that is even worth it, but i have to try it, regardless. Tho if i do my part of the work properly, i will probably in a way be rewarded tomorrow, so it is what it is, and thats for sure what i want. Ive actually stopped looking at my returns and all that garbage, because my data is still too small, and if i could change anything when starting this, i would just not track the profit . Nevertheless, ive checked with a flat stake, and it is a decent profit i guess.

Let me get this straight: if i take the average reference odd, should that be my reference point for EVs? I mean, is that the number i should be chasing (not me, my odds)? I dont know if that is the term, and i kinda forgot about it, would really not like mixing terms, or even making some up. And if you arent near the EV, it means that there is luck (or bad luck) in the mix?

Im really interested in the monte carlo simulation you mentioned, and will do that, for sure. Thanks for the reply!

That table for you as a rookie, is a good start. In fact, the base data for pro punters contains hundreds, if not thousands variable positions. Read on the Internet data about one major sport - NBA METRICS. Now imagine that database for 10 or 20 more kinds of sports, that you've participating-in.
You probably fill in these data pre-match. OK, but later, when you become much experienced, you have to fill in these data in-play mode. Several monitors will be required for you.
Good news - all these data have unequal weight. Some of them are much helpful, others lesser, due to Pareto's Law. Your future experience will help you to reduce them.
Backtesting is a strongly advised operation and extremely helpful to adjust your model.
Monte Carlo and the backtesting will help you to achieve several tens of thousands and more trials, for the future and for the past, respectively.
And don't forget - after every round/match/championship, the data must be updated relevantly, and after that operation, these are prepared to input in your operative model.
Regardin' your question about EV, for reference I personally use odds, computed by me. Many users use sharp bookies/exchanges as basic bookies and later compute the EV.
By the way, did you notice the literature, I've recommended to you in the previous postings, addressed to you?

Based on the fact that I do not have a whole team of players, I will probably just focus on football. I expect that the data sometimes wont be useful, but it will surely have a purpose for improving future bets.

What is exactly the purpose of the Monte Carlo by the way? Is it so I can calculate the odds of my system results?

I will try to calculate the EV, actually I will only predict the value, so I can play based on that value. Yes i will update the database after every round, and I do use sharp odds as a reference point (sportmarket.com). Anyways I think that I will get descent results, because, after all, I only chase the overvalue.

I have notices the literature, tho there is a lot of it obviously, but today i wanted to try actually reading some of it.

And last, but not least, after you read about Kelly Criterion, Ralph Vince's "Optimal F" theory, please read "Long term secrets to short term trading" by Larry Williams, and "The trading game" by Ryan Jones.

Are you referencing this? I will start from there i guess. What literature should I start with from the files you sent me?

The good news I forgot to mention is that I will probably have the access API for the soft bookies I bet on. My friend will automatize the whole process and I will waste less time on the parameter inputs, Im really excited! For now, I think that I want to start actual playing when I get to 1000 odds. I get that my way wont be at all optimised and smart, but I think that i can evolve it from that point forwards. After all, besides my studies, I want to make this my primary work in the future. At least thats the feeling Ive got right now.

By the way, are you from Serbia? Ive noticed the games you referenced above are Serbian first league, and im curious now, tho its all good if you dont want to answer the question. Cheers

Hello again
Please, read the opinion written by the user VidaBlue above, he is one of the smartest here and hinted to you some of many Monte Carlo possibilities. Search on the Internet. Many of us here are so busy, not only in the sporting manner but also in other businesses and when the users have enough time, they'll write here. But their lessons, they've learned for decades, are not possible to be written here in short. Please, search a little on the Internet, before asking something, and we will gladly help further, when we're free of our activities during the working days and nights.

Kelly Criterion, Ralph Vince's "Optimal F" theory, "Long term secrets to short term trading" by Larry Williams, and "The trading game" by Ryan Jones are among must-to-read. But only you may implement the ideas by the Authors in the right way to your purposes. That's one of the giants' tasks in front of you. The theory is not equal to practice. Read these and later practice real positions. Try The Staking Machine software. You're so dedicated rookie, but several years from now, if you don't fail, you'll see that from the very beginnin' you are been among the sharks in the business and so many dangers are there around...

I don't come from that Country, but the nature of the business requires me to be better than the agencies. I know very well Serbian leagues, Taiwanese ones, Primera Nacional Argentinian, Belgian and probably many dozens of Championships and teams over the World historically, not only in the Soccer, in a vast range of the time. That helps me to construct the model and to estimate the odds. While the agencies rely on many departments of statisticians, risk managers, infinite capital, and some hidden factors, I personally rely only upon myself in the business.
Good luck and best profits!

Hello again, @ Marco Kostic,
Is there any news from your side?
May we salute any of your good-building strategies in the business, or maybe you have any questions to help you in the evolution regardin' the betting business on your side?

arb12 wrote: ↑

Sat Mar 12, 2022 4:46 pm

Hello again, @ Marco Kostic,
Is there any news from your side?
May we salute any of your good-building strategies in the business, or maybe you have any questions to help you in the evolution regardin' the betting business on your side?

Hi, old friend!

I haven't seen the tag because it was off, but I came across this post accidentaly so here I am.

I'm doing great, I tested a system, changed a lot of things, but I feel really confident right now. My tests were about 8.5% yield, 10% EV, 40% ROI on flat stakes, and 70% on a geometric staking model. Three days ago I figured out it wasn't worth waiting anymore, and cashed out $2000 for valuebetting. The first day I made the single biggest loss in my, short I should say , betting career, but nevertheless I still feel very confident, and have recovered about 50% of the loss in only the second day. After all, this is a long game, and I'm not concerned about day to day balances. If anything, I'm glad that variance hit me in the face. I wrote a really long post, but the site just crashed on me when I tried to submit it, but I will post numbers and graphs and scientific stuff about my progress in a few days time. These days I'm trying to get into live valuebetting, and prepare for the future with way bigger stakes.

Right now I don't have any questions, but fore sure will sometime in the near future.

How are you doing? Link me a interesting post or discussion from recent times. Thanks for checking in, hope you're doing great as well!