Broken RNG

@janushr

That would be great, I will add you on Skype tomorrow and we can talk about it.

@abnaxis

I did calculate what my real ROI would be based on the number of BIs I roughly estimated I'm running under EV. 1.800 BIs over 32.000 SnGs is around 5.6% under EV. Currently the ROI I achieved is -0.2% which would mean my real ROI would be around 5.4%.

I just took 5.13% because it seemed close enough but you got a great point there. If the numbers were different that would mean more or less HUs and a bit different simulations. I guess I got an edge over those guys because I table select so I should be a bit bigger favorite HU. On the other hand, I more often end up HU with less chips so I have to put that under consideration as well. I will experiment with those simulation and get back to you.

Anyway, great comment, or shall I say 14th post. I see you're not wasting yours by spamming and trolling like some people. Winky, wink... vmarqui.

Originally posted by vmarqui
i'm not gonna look at the thread again (mostly because i won't spend much time trying to tell people they are not running bad), but try to take advantage of the advice people who actually make money from this tell to you.

It's people like you that are tilting me more then the downswing itself. Making idiotic assumptions based on absolutely nothing.
You didn't read the thread? Don't comment please.
You don't know anything about math? Don't comment please.
You like to troll? Don't comment please.

I am a winning player despite running 1800 BIs under EV (or 5.6% ROI under EV) and you concluded what? What did you say about my logic again? Do you even know what logic means?

I played against many bad players who got their good ROI by being lucky. I bet you're one of them.

Just to fill up my 'claim' about 0% ROI appearing as I increase number of simulations:

Everything is completely the same except first is 10k and the second one is 10m. There are so many new numbers appearing in each column, because the more simulations, the better chance more extreme values will appear.

9% ROI and 1% ROI suddenly appeared. Largest drop was 269 before and now is 436. Worst OOM streak was 39, now is 54. These changes are huge and I bet I can make 0% appear just so I can have some better picture of a chance my downswing can happen.

Originally posted by w34z3l
......I know exactly how you can use your ROI Simulator (that only calculates to 1dp) to still quickly establish a reasonably accurate probability of your ROI.

Unless I chased you away, I'd love to hear about this.

Trained statistician here. I'm pretty sure either

a) the assumptions of your simulations are wrong

or:

b) the assumptions of your stats compiler is wrong

Unless you actually know how your compiler of your poker statistics is computing EV, your arguments have 0 credibility. It is very possible that it is computing your stats incorrectly. For example, according to my pokertracker I am running like... 200 buyins above EV. It's because it's calculating $EV wrong.

So just consider all the facts - question everything, don't merely accept that the RNG is fair or biased. In my professional opinion your personal connection to the results is leading you to draw conclusions when there are none (as of yet).

Can you explain how your simulations work?

Originally posted by VladimirN
2) By saying that it cannot appear, you actually took my side that my downswing is impossible

The reason it doesn't appear is not because there were 0 occurrences of your swing, but because your swing happened less than 0.049% of the time in the sample (of trials) and therefore is not recognised by your ROI calculator that displays it as 0.

Originally posted by VladimirN
Now, take a downswing which happens just once in a 100 Million. We will call it '0% ROI'.
Will it appear in 5 Million simulations? Once every 20 bundles of 5 Mil simulations (maybe not and maybe more then once).
Will it appear in 200 Million simulations? Most probably twice.

Do you see how increasing the number of trials can change the result or is it just me?

I understand what you believe. Try and understand why you are wrong.

When you run your trial of 5M times your swing is probably appearing hundreds of times, it just isn't displayed by your ROI calculator because it appears less than 0.049% of the time. However big the number of trials, that swing will appear the same percentage of the time in that sample of 32k.

Your delusion began when you started thinking that 0.0 meant your downswing didn't occur even once in the given number of trials.

Let's say you ran 1M trial and your downswings happened 100 times. What percentage do think your ROI calculator is going to show you? 0.0001%? Heh. It doesn't go to that many decimal places. It'll just show you 0.0.

You may see fluctuations over smaller number of trials, but this is a result of variance. There is hardly going to be any variance when you are running an insane amount of trials.

I don't think I'll help you calculate the probability of your downswing after all, you don't deserve it imo. Go and pay for a coach.

/

Originally posted by VladimirN
w34z3l I love having some meaningful discussions, mainly because I can learn something new. I'm happy admitting being wrong.

You're wrong lmfao. Waste your money failing to get a penny from a lawsuit against Pokerstars.

Please close this thread.

Do you even lift??

Originally posted by w34z3l
I don't think I'll help you calculate the probability of your downswing after all, you don't deserve it imo. Go and pay for a coach.

But if it turns out that I was right all this time, would I deserve it then?

I'll just mention Two crucial points to my argument that you're skipping for some reason.

1) I know that 0.0 means it just happened less then 0.0499% and I said it already.

Look at these red numbers on the next picture:

My downswing is at -0.5% ROI
If 0% ROI happened, it would be on the list and the chance would say 0.0 but it is not. That means it happened ZERO times.

2) Now, how can it appear if I run bigger number of simulations:

Something happens 1 in 100m simulations. If I run 3m simulations, it most probably won't appear. If I run 300m it most probably will appear and it will say it's a 0.0 chance that it happens.

Can you now please comment on these two points without veering off the course which you do in your every comment? Please no 'But your swing probably happened hundreds of times...' and then continuing to talk about things we are not arguing about.

This is just for thought, don't answer this because I'm sure you will veer off again:

Originally posted by VladimirN
If 0% ROI happened, it would be on the list and the chance would say 0.0 but it is not. That means it happened ZERO times.

Statement is not true. It means it happens 0% of the time when rounded down, not that it never happened. It could have happened several hundred times in that sample. In fact I guarantee it happened more than once in that sample size, just looking at the output you've given.

Originally posted by VladimirN
2) Now, how can it appear if I run bigger number of simulations:

As I've explained, it will not appear. Of course it will appear more times over a larger sample, but the overall frequency with which it appears is unaffected. The only thing increasing the number of trials will do is increase the accuracy of that frequency.

I would have already explained to how you could use your ROI simulator to calculate the probability of your ROI, but tbh I find your attitude irritating, and we never even managed to get past basic probability.

Anyway, gl with your lawsuit.

I do maths for a living, and I can tell you for sure that shit happens, even if it doesn't happen very often.

Originally posted by jbpatzer
I do maths for a living, and I can tell you for sure that shit happens, even if it doesn't happen very often.

QFT , actually it happens to me a lot ♦

imo when a result looks impossibru

review:

[ ] are you positive you understand the result properly?
[ ] are you positive you are not over-estimating you're ROI?
[ ] are you sure you are understanding the parameters in the equation properly that produce that result?
[ ] do you understand the -ev in SNG, MTT generated by tracking software (ie it is done on chip level meaning that you can be several BI under ev within one tourney?
[ ] do you fully understand variance?

I think there are a lot of people interested in your POV at first but changed their mind later on after seen the evidence. IMO don't even posting hands as it is completely pointless. The worst chance of winning a pot is 1 backdoor out on the flop (example an overpair vs quads otf and hitting runner runner quads or something like that) and it is not even near impossible.

IMO stop wasting your time on this and spend the time on improving your game. If you don't like stars, change site.

Just realise that it is too easy to blame poor results on the tools we are using.

Over and out

Best of luck

Thazar

jbpatzer, you seem like an intelligent human being. Can you please assist me with my two points I made above which are blindly ignored once again? Can you please take a second and tell me where am I wrong?

1) I know that 0.0 means it just happened less then 0.0499% and I said it already.

Look at these red numbers on the next picture:

Spoiler

My downswing is at -0.5% ROI
If 0% ROI happened, it would be on the list and the chance would say 0.0 but it is not. That means it happened ZERO times.

2) Now, how can it appear if I run bigger number of simulations:

Something happens 1 in 100m simulations. If I run 3m simulations, it most probably won't appear. If I run 300m it most probably will appear and it will say it's a 0.0 chance that it happens.

If you didn't read the discussion, it all comes up to this:
If the chance says 0.0 it means it anywhere between 0.049% and 0.0000000000000000001%
This guy says that 0% ROI maybe happened hundreds of times even though it's not even on the list.
I say that if it happened, it would be on the list, listed with 0.0 probability.

Also, please comment on my example in 2). What am I doing wrong?

For the love of god, I feel like Louis CK in this clip:

Please just run the same simulation for 8k games and find the probability of your specific ROI.

Probability of your ROI over 32k hands is x^4 where x is the probability of your ROI over an 8k sample.

That's of course assuming the stuff in your OP is all correct. I didn't actually bother to read it.

Originally posted by VladimirN
jbpatzer, you seem like an intelligent human being.

Appearances can be deceptive. Tbh, I haven't read all the details in this thread, but I would say that the most likely error is in the calculation of your EV line. I've been deeply suspicious of it since I played a HUSnG where I got sucked out on so often that it gave me an expected $ won that was greater than the first place prize. It's hard to take it seriously for SnGs, and I would just use it as an indication that you ran incredibly badly. Unlucky. Join the club. I wouldn't even start to try to quantify it. I refer you to my previous post.

Lets say you min raise every hand and opponent folds. When he reaches 10 chips you get AA every hand and you shove every time you are in sb. You called every time and you loose every time. Then your EV would be greater than the prize pool which is correct

I made an extreme sample to understand what i mean. Having EV greater than prize pool is right and normal. I didnt read all other posts but i think w34z3l thoughts are correct. You need to make your sample 20k at first imo or 15k. In 4m sample 0.004 has huge difference than 0.0. I think you dont fully understand how simulations work

Originally posted by belthazorrrr
Lets say you min raise every hand and opponent folds. When he reaches 10 chips you get AA every hand and you shove every time you are in sb. You called every time and you loose every time. Then your EV would be greater than the prize pool which is correct

I made an extreme sample to understand what i mean. Having EV greater than prize pool is right and normal. I didnt read all other posts but i think w34z3l thoughts are correct. You need to make your sample 20k at first imo or 15k. In 4m sample 0.004 has huge difference than 0.0. I think you dont fully understand how simulations work

I think I know what you mean, but it's still wrong. Your expected winnings in any given SnG should always be capped at the 1st place prize, however you choose to calculate it.

EDIT: And ofc the flip side of this is that I assume that the calculation can potentially give someone an expected loss of more than a buy in.

I think it would be wrong because you cant calculate the EV of one hand for the entire tournament. The ev of each hand adds up in the end for the final EV you won/lost

The ev cant be calculated the way you say. It is unrealistic

^^
Just on the point of calculating EV for a single tourney where the result is that you should have won or lost more than 1 BI:

$10 sng. Suppose we get in 99bb vs 101bb with 72o on a A3T vs guy with AA. Turn 4; Riv 5. And we bink our 1.6% chance of winning. At this point we have got lucky to the turn of about $9.84 and EV redline records it as such.

Villain still has 2bb left and somehow grinds us back down to 99bb again with no all ins. By some miracle we get the same situation as before with our 72o. Again, we get amazingly lucky and win our 1.6%, and go on to win the sng in the next hand.

The value of the good luck to us is again $9.84. In my opinion it's reasonable of the software to state we ran nearly $20 over EV (which is close to $0 for this game) even though our maximum possible profit or loss is $10.

Otherwise, if it's capped at 1 BI, the consequences of this second AI would basically be ignored and that doesn't seem right if the software is trying to give us an idea of how well or badly we actually ran.