Page 2 of 5 FirstFirst 1234 ... LastLast
Results 14 to 26 of 56

Thread: Norm Wattenberger: CVData/CVCX Futures

  1. #14
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Thanks and More Suggestions

    > 1. The ability to generate a single BS decision or
    > single index number, w/o having to generate the entire
    > BS or set of indices for a count and set of rules and
    > playing conditions.

    You can do that now. Just select one index. Of course indexes are based on other indexes. So, you must have the indexes in the seed strategy that can affect the generated index or the indexes won't match.

    > 2. More progressions testing capability. We know that
    > a progression-based betting system cannot increase EV,
    > but some of our friends may need hard data to be
    > convinced of this.

    I tend to avoid that as I don't want to sell to progression players.

  2. #15
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Sessions Data

    This won't be a problem to add. However, I'll make these stats optional to keep from slowing down most sims. I try to avoid multiplies.

  3. #16
    ES
    Guest

    ES: Re: Sessions Data

    Thanks for adding more Sessions stats.

  4. #17
    Norm Wattenberger
    Guest

    Norm Wattenberger: Question

    Currently, the won sessions and lost sessions do not include sessions that hit the target or bankrupted so that the percentages total to 100%. But, this probably doesn't make sense for the stats you asked for. So, the current stats should probably be changed to match the new stats.

  5. #18
    ES
    Guest

    ES: Answer

    Here is the Session data for a sim I ran:

    Parameters Parameters Parameters Results Results Results Results Results Results
    Seat Bankroll Target Max Rounds Sessions Won Lost Pushed Hit Target Bankrupt
    1 50 100 400 1,630,297 52.76% 46.83% .41% 25.49% 21.23%

    The numbers are not directly below the headings.The size of each small bank is 50 units. The target is 100 units. The maximum number of rounds is 400. 52.76% of sessions were winners, 46.83% were losers and 0.41% were break evens. 25.49% of sessions saw the target reached or the bank doubled and 21.23% saw the bank totally lost.

    The additional data that I would like to obtain are the mean win and standard deviation of the amounts won for all winning sessions, including those in which the bank was doubled; the mean loss and standard deviation of the amounts lost for all losing sessions, including those in which the bank was totally lost and the mean result and standard deviation of the amounts won or lost for all sessions, including those in which the bank was either doubled or totally lost. Can you think of any other useful data?

    Here are some questions that arise because of half-unit payoffs for blackjack, half-unit losses on surrender and multi-unit bets or spreads to more than one hand because of a high count:

    Suppose that the session bank is 50 units, the target is 100 units and the betting unit is $10. If a player's total is $10, he bets $10 and surrenders, then is his $5 total, which is a positive number, but less than the minimum bet (unit), considered a bankrupt session? If so then will $0 or $5 be used in calculating the mean and standard deviation? I think $5 should be used. Suppose that the player has bet his last unit and has to split, double or DAS. How is a negative total handled, if the program allows him to reach into his pocket? If it allows "deficit play," then will a loss greater than the bank size be used in calculating the mean and standard deviation? I think a negative number should be used.

    Does the Sessions function allow a multi-unit bet if the player is down to his final unit, but the count is high? I think this should be a user option in the set up, as some players will adhere to a rigid maximum loss, while others will reach into their pocket to make a bet after losing a bankroll if the count warrants it.

    Suppose again that the session bank is 50 units, the target is 100 units and the betting unit is $10. Suppose that a players bank is increased beyond $1,000 because of a blackjack, a double-down, a split, a multi-unit bet on a high count or some combination. Will a win of "only" $500 or the total actually won in the session, say $520, be used in calculating the mean and standard deviation? I think that the "excess" win should be used.

    Use of the "excess" win, without allowing the player to be set up as reaching into his pocket when he is down to 1 unit or 1/2 unit with a strong positive count skews the result toward the positive. Finally, if the the player can be set up as reaching into his pocket when he is down to 1 unit or 1/2 unit with a strong positive count without some absolute deficit limit can skew the results to the negative. I hope that I have not convinced you to not add enhanced Sessions data to CV Data 5! :-)

    > Currently, the won sessions and lost sessions do not
    > include sessions that hit the target or bankrupted so
    > that the percentages total to 100%. But, this probably
    > doesn't make sense for the stats you asked for. So,
    > the current stats should probably be changed to match
    > the new stats.

  6. #19
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Answer

    The current "Disallow big bet with low bankroll" option overrides the betting strategy when the bankroll is not high enough. It cuts the bet to save for a double or split. But, the code does allow you to go a bit into your pocket for the rare resplit.

    The new screen would look like:


  7. #20
    ES
    Guest

    ES: Questions and Comments

    Thanks for showing the screen. It raises many questions and a new suggestion for Sessions Data.

    My sim was for a $500 session bankroll, a $10 minimum bet and the aspiration to double the bank and the willingness to risk losing the entire bank, with a 400 round time constraint. I used what I call "best available Strip rules and conditions," i.e. 6 decks, S17, double with any two-card hand, DAS, LS and resplit aces. I used an 8:1 bet spread and 70.8% (91 cards or 1 3/4 decks penetration).

    52.76% of sessions were winners, 46.83% were losers and 0.41% were break evens. 25.49% of sessions saw the target reached or the bank doubled and 21.23% saw the bank totally lost.

    Clearly, the range of the size of the wins was between $5 (half unit) and some amount over $500 (in the event that the amount of the winnings was $495 before the final hand of the session and the final hand was a bet in excess of $10, the final hand was split, the final hand was doubled, the fian hand was a natural or some combination. I have no idea of what the mean of such winning totals was, but assue that it was $350. I have no idea of what the standard deviation of such winning totals was, but call it s. About 2/3 of sessions wins will be in the range of $350 +/- s; About 95% of sessions wins will be in the range of $350 +/- 2s. It seems to me that s will be much less that $350, or less than m in general. If not, then the lower confidence limit for the mean of session wins will be a negative number, i.e. a loss!

    Similar results should hold for the mean and standard deviation of the sizes of session losses. If s for sizes of session losses is not much less than m in absolute value, then the upper confidence limit for the mean of session wins will be a positive number, i.e. a win!

    The data on your new screen show the mean loss greater in absolute value than the mean win. This would indicate a very poor game, a very poor playing and/or betting strategy or both. This data also has s > |m|, which would produce questionable confidence limits, as described above. The new screen show a high s for target reached and bankrupt sessions. It seems to me that these standard deviations would be near zero because almost all the data points would be the target in the case of target reached sessions or zero in the case of bankrupt sessions, with a few of the target reached sessions having a win in excess of the target because of a high bet, split pair, double down or blackjack on the final hand and some of the bankrupt sessions having a total of 1/2 unit because of surrender on the final hand or a total of 1 1/2 units before the final, losing hand and a loss of a split pair, souble down or combination thereof of the final hand.

    It would seem to me that a relatively long time constraint, coupled with a relatively low target and relatively high bankroll size would have a relatively low standard deviations for sizes of winning amounts because, under these conditions, many session results will be reaching the target, i.e. many data points are equal. It would seem to me that a relatively long time constraint, coupled with a relatively high target and relatively low bankroll size would have a relatively low standard deviations for sizes of losing amounts because, under these conditions, many session results will be losing the entire bankroll, i.e. again many data points are equal.

    It would seem to me that other sets of time constraint, bankroll size and target data would produce relatively high standard deviations. Would they produce standard deviations large enough to have negative lower confidence limits for wins and positive higher confidence limits for losses?

    Your new screen shows m and s separately for winning sessions and losing sessions, but not an aggregate m and s for all sessions. I would like to see aggregate m and s for all sessions.

    Your new screen shows m and s data but not the percentages of winners, losers and break evens and the percentages of target wins and total losses that the current screen shows. I would like to see this data as well as the m and s data.

    Finally, I would like to see average durations and standard deviations of durations for winning, losing and total sessions. This is useful for the low stakes player in a casino location with no low minimum tables, e.g. Atlantic City most of the time, who wants to get an idea of how long he will likely be able to play given his small bank and the high minimum bet.

    > The current "Disallow big bet with low
    > bankroll" option overrides the betting strategy
    > when the bankroll is not high enough. It cuts the bet
    > to save for a double or split. But, the code does
    > allow you to go a bit into your pocket for the rare
    > resplit.

    > The new screen would look like:

    >

  8. #21
    ES
    Guest

    ES: Re: Thanks and More Suggestions

    > I tend to avoid that [progression analysis] as I don't want to sell to progression players.

    There is a risk that when a niche product begins to appeal to a broader, less elite market, the producer will sense that he can cut corners and make more money. This is more so when the producer is taken over by a larger company. Quality control, research and development, etc. will diminish and the product will suffer. I have full confidence that you will neither sell out literally by allowing QFIT to be taken over nor figuratively by abandoning cutting edge practice and simulation software if adding the ability to disprove voodoo were to give you a majority of ploppie as customers. I do understand your committment to CV products as they currently are and personally am more interested in expanded sessions data than in expanded progressions analysis.


  9. #22
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Questions and Comments

    I don't see how confidence levels can be calculated in that manner on half of the curve. I'm not convinced that breaking the data into two hunks and calculating SDs on each half has any use.

    BTW, I have always had a bit of a problem using standard deviation to create confidence levels since this assumes no skew. That is, a symmetrical curve. But, it seems to me that card counting has skew due to the fact that we bet larger when we have an advantage and that advantage tends to continue for multiple hands. This is why I don't use SD for "random walk" examples in my software, but use simulation instead. On top of that, we have fixed limits that chop off the ends of the curve non-symmetrically further messing with confidence level calcs. I don't think that this makes a huge difference to our general calculations and don't wish to disparage the use of confidence levels. But, it can cause aberrations.

    At least that's how I read it.

  10. #23
    ES
    Guest

    ES: Re: Questions and Comments

    > I don't see how confidence levels can be calculated in
    > that manner on half of the curve. I'm not convinced
    > that breaking the data into two hunks and calculating
    > SDs on each half has any use.

    The amounts won in winning sessions and the amounts lost in losing sessions are separate sets of data m, s and confidence intervals for each "half" of the total data gives information about what can be expected for winning sessions and for losing sessions. m, s and confidence intervals for the two "halves" aggregated (and for break even sessions) are also useful, as they give information about what can be expected for the totality of sessions. I requested this aggregate data in my prior post.

    > BTW, I have always had a bit of a problem using
    > standard deviation to create confidence levels since
    > this assumes no skew. That is, a symmetrical curve.
    > But, it seems to me that card counting has skew due to
    > the fact that we bet larger when we have an advantage
    > and that advantage tends to continue for multiple
    > hands. This is why I don't use SD for "random
    > walk" examples in my software, but use simulation
    > instead. On top of that, we have fixed limits that
    > chop off the ends of the curve non-symmetrically
    > further messing with confidence level calcs. I don't
    > think that this makes a huge difference to our general
    > calculations and don't wish to disparage the use of
    > confidence levels. But, it can cause aberrations.

    > At least that's how I read it.

    There is less skewness in sessions data than in "regular" sims because the nature of the session places a cap on the size of the win as the session is terminated if the target is reached.

    Just a reminder of my other further requests:

    Your new screen shows m and s data but not the percentages of winners, losers and break evens and the percentages of target wins and total losses that the current screen shows. I would like to see this data as well as the m and s data.

    Finally, I would like to see average durations and standard deviations of durations for winning, losing and total sessions. This is useful for the low stakes player in a casino location with no low minimum tables, e.g. Atlantic City most of the time, who wants to get an idea of how long he will likely be able to play given his small bank and the high minimum bet.


  11. #24
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Questions and Comments

    > Finally, I would like to see average durations and
    > standard deviations of durations for winning, losing
    > and total sessions. This is useful for the low stakes
    > player in a casino location with no low minimum
    > tables, e.g. Atlantic City most of the time, who wants
    > to get an idea of how long he will likely be able to
    > play given his small bank and the high minimum bet.

    By winning and losing, are you including target hits and bankrupts?

  12. #25
    ES
    Guest

    ES: Re: Questions and Comments

    > By winning and losing, are you including target hits
    > and bankrupts?

    Yes, unless you have a reason not to include them.

  13. #26
    Norm Wattenberger
    Guest

    Norm Wattenberger: Re: Questions and Comments

    I think I see what you want. I just have to figure out how to calc eight SDs with non-zero means in a reasonable amount of CPU time.

Page 2 of 5 FirstFirst 1234 ... LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  

About Blackjack: The Forum

BJTF is an advantage player site based on the principles of comity. That is, civil and considerate behavior for the mutual benefit of all involved. The goal of advantage play is the legal extraction of funds from gaming establishments by gaining a mathematic advantage and developing the skills required to use that advantage. To maximize our success, it is important to understand that we are all on the same side. Personal conflicts simply get in the way of our goals.