THE MYSTERY OF STREAKS

[Ray]

Everyone knows that streaks occur all the time in BJ games, but did
you ever ask yourself if there is some specific reason for this phenomena.
Some people say it's because the deck contains lots of high cards that
favor the player. Others will say it's just a matter of blind luck and that's
the reason that it don't happen too often. While others attribute streaks to
some magical way that the cards are shuffled or perhaps, not shuffled.
I favor the blind luck possibility because there are all kinds of streaks
that can be observed. For Example: We have win, lose and push streaks.
Now that sounds stupid, who ever heard of a push streak? Just before
last Xmas I played a session where I would have lost even if the dealer
had been playing face-up and I pushed on 20's and 21's seven hands in a
row, even on BJ's. Why do you folks think streaks occur?

[Walter Thomason]

Ray: Streaks happen because streaks happen... it's a natural outcome for a game with a 48/52% win/loss pattern. And one of the main reasons why win streaks are longer in duration that expected is because we win hands we didn't expect to win. Seems every long streak involves drawing that 8 to a 13, splitting 8's against a 10 and winning both hands, having the dealer bust in spite of his 10 upcard, etc. Winning those hands that the odds say you won't win leads to streaks that are longer than they "should" be...

[Ray]

Walter: I agree with your example. A very good example of "Blind Luck".
I might add: There is some reasonable logic that tends to support losing
streaks down in the negative counts. There are times when the 47.5 win
rate can drift down in the 30-40 range. The -1 or 4 loss walk idea serves
its purpose well.

[Grifter]

Ray - Better check those numbers......I'm busy right now, but I do not believe there is any way your expected WP will drop to the 30% range....just no way....G.

[Ray]

Grifter: Go to bjstats.com, select tables on the left, then select:
6 Decks
High/Low
No sort
No cover
S17,DAS
TABLE- cumulative win,loss&tied % by TC^

Look at the win loss numbers starting at zero and going both ways. It's
about midway in the table and I think it excludes pushes(47.5)

It takes high neg counts that don't come very often, but losing 10 in a
row don't come very often as well.

Ray

[Grifter]

Ray – O.K.…I looked at bjstats.com like you said, and I’ll admit those are some pretty “cool” tables…….but, Ray, they prove my point, not yours……Look again.

All the way down to negative counts of TC = -16, the numbers are between 47.39% and 46.04% (with one exception at -14). Then there are two anomalies at -17 & -18, but these are due to the small number of ‘occurrences’…..The actual average for all these minus counts from TC -1 to -16 is 46.62%, or less than 1% below the ‘norm’ of 47.5%.

This is what I expected to see, and it is the way it should be. There is no drastic drop in WP with a negative count. I noticed you were “taking the same tack” with Sage earlier on another thread, but I didn’t say anything…….You should be analyzing EV, not WP….. that is where you will see the ‘big’ change, and why you should ‘bail’ at negative counts.

Regards…..Grifter

[Ray]

Grifter: I understand the importance of the circumstances that exist at
the neg counts and I think I related this to Sage as a combination of
things which, of course affect the EV. They were: win rate, doubles,BJ and
hit/stand/split decisions as being less effective.

As regards streaks: The win rates for positive counts is, for the most part,
stable. It changes very little up thru high positive counts. Down thru the
negative counts the changes are greater. It does not matter how low the
percentages are in the run that we looked at because the changes are
very close to a linear translation and can be normalized or extraplolated
for the same volumes as the positive counts. The lower EV in the negative
counts is an effect, not the cause. The cause(s) are: Lower win rate, less
BJ and double posibilities, and less effective decisions for hit/stand/split.

In my opinion these are the circumstances that increase the probability
of losing streaks in high negative counts. No such circumstances exist
for winning streak probabilities that I can identify. But, in the high positive
counts the EV is much greater even though the win rate remains stable.

Ray

[Mike]

"It does not matter how low the
percentages are in the run that we looked at because the changes are
very close to a linear translation and can be normalized or extraplolated
for the same volumes as the positive counts. The lower EV in the negative
counts is an effect, not the cause. The cause(s) are: Lower win rate, less
BJ and double posibilities, and less effective decisions for hit/stand/split. "

Whoa... Does this mean I double down ace/duece against a 6? (jk)

[Ray]

Mike: I should have made it more clear that I was talking about the
percentages listed in the sim and not related to the play of a hand,etc.
Sorry for the confusion it may have caused...........

[Mike]

I Was just kidding....It was a futile attempt at humor based on your articulation of your formula. It's kind of refreshing to actually have to think about what I'm reading. Carry on, my friend

[Grifter]

Ray – Nice post, but you basically said the same thing you said before with more words. I honestly don’t know where you are getting ‘off base’ so I will start from the beginning again.

The ‘quick and dirty’ numbers I gave you yesterday were close enough, but you sidestepped them so here is the whole gamut. The results below are based on the numbers you told me to look at, and reflect the cumulative WP for all negative counts from -1 to -10 inclusive; which covers 99.94% of all occurrences.

1. Ray said, "There are times when the 47.5 win rate can drift down in the 30-40 range.”

The above statement is false…….See following:

Counts………………………Cum. WP
-1................................47.39
-1,-2............................47.33
-1,-2,-3........................47.28
-1,-2,-3,-4....................47.26
-1,-2,-3,,…-5................47.24
-1,-2,-3,,…-6................47.22
-1,-2,-3,,…-7................47.22
-1,-2,-3,,…-8................47.22
-1,-2,-3,,…-9................47.21
-1,-2,-3,,…-10..............47.21

As you can see, the ‘drift’ is negligible……only 0.18% going all the way down to -10, and never gets even remotely close to the 30 - 40.

2. Ray said, “The win rates for positive counts is, for the most part, stable. It changes very little up thru high positive counts.”

The above statement is false……The following is a reverse of the above, showing the cumulative WP for plus counts of +1 to +10, inclusive.

Counts………………………Cum. WP
+1...............................47.93
+1,+2..........................48.15
+1,+2,+3.....................48.46
+1,+2,+3,+4................48.77
+1,+2,+3,…+5..............48.96
+1,+2,+3,…+6..............49.08
+1,+2,+3,…+7..............49.14
+1,+2,+3,…+8..............49.17
+1,+2,+3,…+9..............49.19
+1,+2,+3,…+10............49.20

Your statement above is not only incorrect, it is actually backwards. As you can see, with positive counts over the same 1-10 spread we now have a drift upward of 1.27%.

Hope this helps, and let me know if you have any questions.

Regards……Grifter

[Ray]

Grifter: Because I can't find a good description of the numbers in the
table, I'll say it's possible that my understanding of the numbers is lacking.
For Example: It appears to me that the numbers are Cum WP
up or down thru the counts and need no further elaboration. As best as
I can determine, this is what they mean when they say each position
contains the results of the previous.

There are anomalies in the negative counts that probably show up only
because of the limited action, but to that I would say; losing 10 in a row
is an anomalie that would, in my opinion, require something far less
favorable than a 47.5 win rate.

I'll revisit the site and continue looking for a more detail exp. of the no's.

Ray

[Grifter]

Ray - That's it!...There's your glitch.....You just said, "It appears to me that the numbers are Cum WP up or down thru the counts and need no further elaboration. As best as I can determine, this is what they mean when they say each position contains the results of the previous."

No, unless I'm missing something entirely, the numbers on the chart are not cumulative. They are stand alone numbers for each event (i.e. -1, -2, etc.)......However, my charts above are cumulative, thus the difference; and cumulative is what you want to use.

Re Anomalies: There are anomalies that occur in both the individual positve and negative counts, and yes they do show up because of low occurence frequency. That is why I stopped my charts at +/- 10......That covers 99.94% of the total occurences.

Re 10 in a Row: but to that I would say; losing 10 in a row is an anomalie that would, in my opinion, require something far less favorable than a 47.5 win rate......That would depend entirely on the size of your sample. If you had 10 L streak in a set of 100 hands, in all likelyhood the WP for those 100 hands would be less than 47.5%. Conversely, if you had a 10 L streak in say 1,000 hands; it is possible that the overall WP would be over 47.5%.

Regards.....Grifter

[Ray]

SCORE ONE FOR GRIFTER:

The dam sim is reverse cumulative.....How did I overlook that(can't read)
I went to other sims to try and figure out what was going on and found the
numbers to be inline with yours.

One thing that I noticed that was a supprise to me was the increase in
pushes that affected the loss percentage more that the win percentage
on the positive counts.

Thanks Grif

[Grifter]

Ray - You're welcome, but no thanks necessary. Just return the favor and catch mine when I make 'em. :wink: ......Grif'

[Mike]

Ray - You're welcome, but no thanks necessary. Just return the favor and catch mine when I make 'em. :wink: ......Grif'


All that and not using the "F" word once, amazing what intelligent minds can do, That's why I continue to visit this site. :!:

[Sage]

Grifter/Ray If I understand your discussions?? The charts indicate the low counts do not affect the results as much negative as the high counts affect the results positive. (Is that because when the count is in the process of going low, high cards are coming out and that is good for the player?)
The charts confirm, when the high cards come out, (low count) the W/L rates don't change much from the expected, but when the low cards come out (high count) the W/L rate gets better for the player. Does that mean low cards are much better for the dealer than high cards are for the player? It is not an even swap, high cards are less important that low cards?
So I understand why to bet more with high counts. But it seems that neutral counts and low counts are about the same W/L, is that right?
If I understand it correctly, why quit the shoe with a low count if you don't also quite the shoe with a neutral count? :?:

[Ray]

SAGE: I think the situation at neutral or negative is more related to
what your options are after you receive your hand. You want get as
many blackjacks or 20's, but I think what is more important is the
reduced options on doubles(your best way to even out the odds).
There are 10-12 doubles that are not available,9vs2,11vsA,9vs7,
and the like. Also, you have a reduced chance of making the B/S dbls
and dbl after splits,etc. If your not Wonging, then you more than likely
will play thru some neutral or slightly negative shoes because your
going to play at one table for an extended time.

I try to keep good records and notes on every session that I play and
I can tell you that if you don't make your fare share of doubles, your
not going to win. I could have stayed home all of January and saved a
few bucks for that very reason........

[jedi]

Walt said:

Streaks happen because streaks happen... it's a natural outcome for a game with a 48/52% win/loss pattern

I agree, but with all due respect, flip a coin 20 times, you'll find streaks too, and that's as 50/50 as you can get! Are streaks less prevelant in a 50/50 than a 48/52???

[Walter Thomason]

jedi: Nope, but coin tosses are based upon independent trials, where a win is just as likely as a loss, whereas wins and losses in blackjack are dependent upon deck composition. My point was that any game with only limited outcomes (win/loss, heads/tails) will produce streaks, and that deck composition will effect the frequency and duration of streaks.

[jedi]

composition of remaining cards vs independent outcomes. yeah, that makes perfect sense. I just wrote without thinking it through. 8)
but it still is strange how streaks happen even in a 50/50 independent outcome scenario. the same probability to get 10 heads in a row as it is to get alternating 10 outcomes. the streak info in previous posts is very helpful, and encouraging, with the losing streaks I've seen so far, it should be payday for me from now on, and my kids too....and their kids....... 8)

[Walter Thomason]

jedi: I know the feeling... I actually remember winning a hand of blackjack in October of last year... or was it the year before? Last night was typical... I finally reached my max bet, split, doubled down on both hands, drew a 19 and 20, and the dealer drew five cards to make a seven card 21!!