Software simulator and unexpected results

sagefr0g · Oct 15, 2009

snappers percentages

when the number of blackjacks a player receives are considered number wise and percentage wise and the number of pushes a player receives are considered number wise and percentage wise....... how are pushes and blackjacks treated when both the player and the dealer has a blackjack hence a push?

i mean for the case where both the dealer and player have a blackjack, is that considered statistically a push for the player and not a blackjack since it didn't pay anything?
i mean that statistical number of blackjacks a player can expect (circa 4.75%)
according to this link for six deck games http://www.blackjackincolor.com/blackjackfirstcards1.htm
does that sort of number exclude pushed blackjacks?

sorry for the confusion, i guess i'm wondering what the conventional way to consider the stats would be.

:whip:

QFIT · Oct 15, 2009

If both player and dealer get a BJ, that is counted as a push and a BJ.

sagefr0g · Oct 15, 2009

QFIT said:

ok thank you, hmm, i didn't expect that answer, lol.
i'm definitely not trying to be argumentative here, just this kind of perplexes me as i was leaning towards thinking of it as just a push.

:whip:
what would be the reason behind considering or counting it as both a push and a BJ? maybe the fact that it's not a winning BJ hand for the player?

QFIT · Oct 15, 2009

A BJ is defined as two original cards adding to 21. How or if they are paid depends on other rules.

Sonny · Oct 15, 2009

The two tallies are independent. A win/loss/push is based on the payment of the hand. A BJ is based on what cards make up that hand. Think of the win/loss/push as the main count and the number of BJs as a side count. They measure different things and will sometimes overlap, but they can be useful on their own.

-Sonny-

sagefr0g · Oct 15, 2009

quoting QFIT

Sonny said:

right Sonny i'm beginning to get the point, slowly but surely, lol.

one thing i learned is that for a basic strategy only player (not making an insurance bet) is that the conventionally quoted circa 4.75% player blackjack statistics for say a six deck game isn't exactly what i thought it was in my naivety, lol.
i mean yeah that percentage of blackjacks is a sweet thing, it being what it is and all but that in the cold hard real world doesn't really mean you got payed 3:2 on that entire percentage.

:cry::whip:

Kasi · Oct 16, 2009

sagefr0g said:

To re-hash what you already understand lol.

Take single-deck. Number of BJ's received is based on initial 2 cards dealt 2*(16/52*4/51)=4.83%.

Given that you have a BJ, the prob of the dealer not having one is 1-2(3/50*15/49)=96.3%. (4 cards off the top now) so the player's chances of a winning BJ are 4.65%.

So, it's good to know, like you say, the number or % of BJ's that actually pay 3-2. Like for 6-5 BJ, the HA is increased by 1.39% due to that rule because you lose an extra 0.3 units 4.65% of the time.

Can you imagine how immensely more profitable BJ would be should ALL player BJ's be paid 3-2 whether they push the dealer or not? ! lmao.

\So, yeah, if you counted up only number of winning BJ's over a few thousand hands and thought that'd be the higher number, you'd be spending time thinking how unlucky you were that you received so few "BJ's" lol.

In my early internet days before I knew my a** from 3rd base, not that I do now, but I'd make my sheet only count my winning BJ's while playing and 10000 hands later wonder why I was short so many BJ's thinking they would be the higher % and especially while also counting total dealer BJ's dealt and comparing my winning BJ's to his total BJ's. He always had way more than I did that way - it was so unfair lol.

That's how I learned that one lol.

sagefr0g · Oct 16, 2009

Kasi said:

thanks for all that good stuff dotted out above, some day i'm gonna try and learn that stuff, lol.

i stumbled upon the question when i was fooling around setting up the sort of stats in the image below. the hi-lited stuff.

by the way does that figure in the image look correct for total units wagered and total units won, all that stuff, lol?

:whip:

this thing has been scaring me cause at 9768 hands it's still ahead on money won. i even used your sheet (for a similar game) to figure how much standard deviation that represented.

heck it's up 100.5 units if i'm figuring that stuff right about one standard deviation.

TRC · Oct 16, 2009

Counting in the sim

Fabio,

Great start on the sim. I had been thinking about writing my own and it seems you beat me to it. A few concerns though.

1) In your counting algorithms you seem to have the options to count using Red 7, KO, and HiLo. Red7 and KO are unbalanced counts whereas HiLo is balanced. You seem to be doing a conversion to the true count the same way regardless of system.

2) The calculation for the KO count is off as the 7 should be counted as a low card (+1) not a high card (-1) as you currently have it.

3) The Dealer::hasBeenDealtBlackjack() method will only return true if the first card is an Ace, but not when the Ace is the second card. You commented out that check though, so the logic is there just not used.

4) Your handling of checking for pairs can be done much cleaner by comparing their value directly rather than checking if they're repeatedly. (hard to verbalize, see below) There are several other areas that can be cleaned up. If you're interested, PM me and I can send you the code changes.

Code:

if ((( A == 1 ) && ( B == 1 )) || (( A == 2 ) && ( B == 2 )) || ... || (( A == n ) && ( B == n )))

vs.

Code:

if ( A == B )

That's all for now, but if I notice anything else glaring I'll reply to this thread again. Like I said, I've made some changes to clean up the code and handle some parts more efficiently. If you'd like to talk more please PM me since I don't want to cover the forum with C++.

k_c · Oct 17, 2009

Interesting site

Here is a site www.agner.org with links to information about random number generation as well as open source code to rng packages.

Kasi · Oct 19, 2009

sagefr0g said:

Hi Wise One,

I just spent the last several hours thinking in my mind the Steelers had 10 points and was counting my money after the 13 point win. Man was I disappointed to re-read what I did and discover it was actually a 14 point spread lol.

Long way of saying, don't trust what I say lol.

Whatever, I see where all your numbers are coming from.

Big Picture

- why would it scare you have decided your results are just a little over 1SD to the good? There are white witches out there too that fly around the "right" side of the curve visiting their white magic just as often as the black witches do on the "left" side of the curve from EV.

So, let me get this straight, while many perhaps might curse their slightly -1+ SD bad luck yet say to themelves - no problem - just a little bad luck - you get a little +1+SD* luck and decide I'm doing something wrong lmao?

When things get + or -3+ SD, then begin to question what may be goin' on lol.

Really Small Picture

and how your sheet results may or may not compare to a "sim"'s results.

A sim would not say your win rate is 100.5/10617=0.00938. It would say your win rate is 100.5/9768=0.001029. Likewise with those other w/l/t/ %'s on that line.

The 1.03% would better compare to the "HA" in a sim, based not on total units wagered but initial bet/rd.

Going one level farther down, would your sheet treat a split hand as 2 "hands" or just 1? As far as w/l/t%'s go, a sim would trat a round where one split "hd" won and one lost as one round and a push.

I suspect maybe your sheet might be treating a split hand "round" as 2 or more " hds?

In other words, ideally, your 9768 "hds" should also mean you've played against 9768 dealer upcards. What I call "rounds".

In other words, it could be, depending how your sheet treats "splits", you have only actually played against 9668 or fewer dealer upcards ( a wild-ass guess btw).

If your sheet is a 6D game, then "2*(24/312)*(96/311)*(1-2*(95/310)*(23/309))=4.5323% of total rounds dealt should be a winning BJ.

Like, the way I imagine things, CVCX doesn't have all that number of rounds win% stuff but CVDATA might.

Anyway, the BIG pictiure of this SMALL picture is that I wouldn't worry that much about how you treat "splits" anywy for a while longer.

Also, I have no way of knowing how many splits and doubles would be expected to occur under games with various rule sets anyway. I just know, empirically, from internet play what I actually did get lol.

And there was a big difference in % of doubled and split hands in Micro SD with D10, no DAS, vs multi-ceck Crypto or CON with DOA, DAS rules.

sagefr0g · Oct 19, 2009

Kasi said:

hey, thank you a lot for running all that past me, it's good i'll be able to refer back to this post about that sort of stuff.

the one Bengals game i watch and they lost it today, after quite a winning streak i hear. go figure.

i guess i wasn't all that scared by the circa one standard deviation, but i dunno it just seemed like the results kept coming in where the player was consistently ahead, not like how it goes for me when i play in real life sort of thing where it's such a wild up and down ride, sort of thing.
it did get me to thinking about what QFIT says about excel's rng and the story you told about the lotus thing with repeating results, sort of thing.
lmao, but i'm also not forgetting what ExhibitCAA said about let's be sure not to blame our programing mistakes on the rng, sort of thing as well. and i'm fully aware the program is making mistakes at this point, with splits and some other glitches.

but yes, where i had number of hands played, well that's counting hands created by splits, where those hands are played against a single dealer up card, so if you define rounds by how many dealer up cards are played against then the number of rounds played would be less than the number of hands i denoted.
i guess maybe it's good to consider rounds by dealer up cards, so that way maybe you have the initial bet information to make calculations with, sort of thing.
i guess too, like when a double down occurs and a successful snapper occurs
that those factors have implications when considering initial bets as well.
anyway you gave me some food for thought stuff, thank you.
edit: and oh yeah don't give me that crap "don't trust what i say" stuff,
heck i'm totally banking on what you say, lol.

Kasi · Oct 20, 2009

sagefr0g said:

Don't give me that cr*p lol - isn't it you who, in real like, turned a sh*t starting roll into $6K plus? Basically, "Kissed by the Hand of God and Shot in the Ass with Luck" initially :grin: :laugh: :whip:

Despite use of your vaunted "fuzzy counting" system :laugh:

Were you not similarly blessed in the Weekend Warrior thread? (Very possible my memory fails me on that one lol). Maybe you were similarly cursed lol.

Anyway, I suppose if your sheet's results get out there at some point to 6-10 SD's from expected, I suppose you'd have some problem for some reason at that point whether it be a faulty RNG or something else.

I think, most sims would technically count a "round" with split "hands" as 1 "round" that your sheet calls one "hand".

All my internet stats did exactly the same thing too until one day things just didn't seem that reasonable anymore but after adjusting, in a broad fashion, they made more reasonable sense. Just like with "winning" Bj's vs total BJ's. So at least now I realize they are off a little.

It seems to me like maybe CVDATA comes up with these W/L/T%'s and/or IBA vs TBA %'s. (like the 100.5/10617 whatever - TotalBetAd vs 100.5/9768=1+% (InitialBet Adv).

To everyone else - I consider The Wise One a friend and that is the only reason I'm taking the liberty of initially giving him a little "grief" lol.

Alot of days, I wish I could meet more of you that post here.

sagefr0g · Oct 20, 2009

Kasi said:

by that do you mean, errhh like i was saying and i think you was saying that what is called a round is counted by how many dealer up cards one is seeing?
like for instance say the rules say you can split three times then you could have as many as four hands against one dealer up card, so that would be considered just one round?
is that what a round is considered in your sheet?

if you'd drag your self to the bj-bash you would, geesh.

:whip:

Kasi · Oct 22, 2009

sagefr0g said:

Well, I think any splits would be counted as 1 "round" in most sims. It's just, it seems, most sims, for some reason, never even use the word "round" lol.

I don't know why they don't use the word "round" lol. Maybe "hand" is defined in the fine print somewhere?

Like in the Wiz's Appendix 4 where he gives "net wins" and "net losses" there with a range from +8 units to -8 units. How else do you get a -7 or -8 units unless you'd count a "round" with 4 split "hands " as 1 "round"? Then, below, he includes in "win%" any net win.

So, if my sheet is based on a sim, since (if!) the sim is based on spllits being 1 "round", then I guess my sheet is too lol.

The more I think about it, I don't know why, but I have a vague feeling QFIT may have said at some point CVDATA and CVCX might actually treat what is included in w/l/t% differently - basically treating differently the definition of "hands" in both softwares?! Maybe something like CVCX when it says a "200,000,000" hand sim, it actually might be counting a split "round" as 2 "hands". I'd have to dig around for that lol or maybe he might comment on ramblings lol.

Maybe it was, the more I think about it lol, more like something like CVBJ counts "split hands" as 2 hands but CVCX and CVDATA don't lol. SO a guy practicing alot with CVBJ could have slightly skewed results when comparing to a CVCX or CVD sim lol. I think that was it lol.

Does CVCX even give W/L/T %'s somehwere behing the usual main screen?

Basically, in Don's book, I guess I've always assumed, rightly or wrongly, that the first 3 columns of Freq, win% and adv has already taken into account the
assumption that splits are 1 "round". But it could be CVCX doesn't do this.

It just goes to the point, even when one might actually be measuring actaul results vs a sim, one should, ideally, understand the assumptions behind the sim.

Or, otherwise, like in Don's tables, if back-counting, one may interpret the "W/100 ($)" column" and "SD/100($)" column as meaning per 100 hands physically played and make an egregious error. Or, perhaps worse, interpret the "N0 (hands)" column as physical hands needing to be played to achieve N0.

I guess it all goes to expressing everything (EV etc) in terms of one's initial bet. - like an EV when doubling is often twice that of when only hitting.

Likewise, EV's when splitting are expressed in terms of initial bet too.

I will hope forever to join a BJ Bash

sagefr0g · Oct 22, 2009

Kasi said:

like you'd think maybe i'd try and find the answer about hands and rounds sorta thing in cvcx documentation, doh

:whip:

well, like i just read this one snippet,
Hands per Hour - Number of rounds observed at the table. If you are back-counting, this includes skipped rounds.

then i know if your going to run a cvcx sim, it'll want to know how many rounds you want to run, sort of thing.

well like in your sheet, i think the sheet gives some number of rounds to N0, that will give you N0, well then ok, take that number of rounds and put it in your sheet where you stipulate number of hands and the sheet will come up and give you your ev = one standard deviation.
edit: probably maybe if you put the actual number of hands represented by a N0's worth of rounds maybe you'd get the same answer, cause maybe the number of hands wouldn't be all that much greater? just guessing here.

ok but like we was discussing, like you said this program thing i'm making well i had it stating some number of total hands played. and like you said you figured those number of hands played would be a higher number than the actual number of rounds played, essentially because of splits sorta thing.

lmao, well, what ever, i'm sorta confused but at least maybe every thing is what it is and maybe at least that can be known if needed, lol.

edit: but anyway i think i'm convinced the proper way to think of rounds is really by how many dealer up cards you see, sort of like i think you was saying.

Kasi · Oct 24, 2009

sagefr0g said:

It's no big deal Wise One. It's more about when you tell a sim to run 200,000,000 something, are you asking it to run against 200,000,000 dealer upcards or are you asking it to run 200,000,000 "hands" if one counts split hands as 2-4 hands lol?

Sure. my sheet gives N0 after some number of something but is it "hands" or "rounds"?

Since N0 is based on EV/"hd or rd"/ and SD/"hd or rd", it behooves one to know which it is lol.

Just so, when one later comes back after a session to measure one's results, is one going to estimate as "hands" played including splits or is one going to estimate the number of dealer upcards one played against?

And, no, my sheet would not come up the same N0 if one chose hands vs rounds unless one adjusted the sims results of freq of TC's, win%/TC and SD/TC for "hands" vs "rounds". I imagine each might slightly change in that case since one may start out at +2 but by the 4th split "hand" that "round" not be +2 anymore and therefore not be included in the freq of +2 count kind of thing.

Basically, it all goes back, to me anyway, how BS itself, and "EV" is only based on initial 2 cards dealt vs 1 dealer upcard and always expressed in terms of initial bet before one doubles or splits.

Like there's no way your 0.94% would equate to "W/L(%)" col in Don's tables.

Your .94% is 100.5/10716 - that would be a "TBA" in CVDATA (total bet advantage"). 100.5/9768 would be more like an "IBA" (initial bet advantage) in CVDATA. The way I understand it. If 9768 includes split "hands" and 1 % of that are actually splits then only 9670 delaer upcards and 100.5/9670= a number that would equate to the "W/L (%)" column in Don's tables.

I think lol.

You think this might be confusing, I do too and am never exactly sure what the numbers I may be looking at from the output of a sim might actually mean lol.

Let alone what "avg bet" may mean or what is a "hd" or "rd" when "spreading" is involved lmao.

Mostly, most likely a tempest in a teapot anyway lol. Worst is subtract 1-2% range (guess) from your definition of "hds" and assume "rds" and see what changes by how much.

But, you know, like your (winning) BJ % of 4.66%, as I recall, can't see now, seems "right on" based on 9768 "hds" but if only (9768 less 1 or 2%) number of "rds", then you have been winning more than your fair share of BJ's so far and, that alone, might account for a lot of your actual win exceeding your EV win lol.

Obviously your expected number of winning BJ's would have nothing to with split "hds" counting as "hds" and only have to do with the number of dealer upcards one has played against lol.

I have no idea how to figure out the expected number of splits for various sets of rules anyway lol. Pretty sure CVDATA does it - does CVCX do it?

QFIT · Oct 24, 2009

Simulations always deal with original bets because all risk calculations are based on the original bet. Split that into 8 hands, and the data is still based on the original bet. It's like Craps. A "hand" in Craps continues until you seven out, no matter how many times you roll the dice.

Kasi · Oct 24, 2009

QFIT said:

If spreading, say $10 to 4 "spots" vs one dealer upcard, would CVCX treat that as a $40 initial avg bet per/round or a $10 initial avg bet and 4 "rounds"?

To measure my actual vs a sim of this, should I count that as 4 "rds" or just "1"?

Is mu initial avg bet here $10 (based on one spot) or $40 (based on 1 dealer upcard)?

Is it true CVBJ would count splits as 2 or more "hands" when they happen and that one would have to adjust for this to compare one's actual results using CVBJ compared to the expected results of either CVCX or CVDATA?

sagefr0g · Oct 24, 2009

Kasi said:

i can't answer your question off hand but as far as the CVBJ log, yes it records splits as two or more hands.

Software simulator and unexpected results

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Attachments

New Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member