Risk Adverse Strategies
  • Taking off another thread, is there a list/discussion somewhere of the best risk adverse strategies (i.e. if you are risk adverse, the plays you can make to lower the variance with the most minimal drop in EV?). I'm assuming they all come into play mostly with doubling down and splitting.

    I can understand why they aren't too popular. Usually doubling down and splitting are the things that I have the most fun doing while playing BJ. Still, I would like to know about them.
  • FunkyChicken said:
    Taking off another thread, is there a list/discussion somewhere of the best risk adverse strategies (i.e. if you are risk adverse, the plays you can make to lower the variance with the most minimal drop in EV?). I'm assuming they all come into play mostly with doubling down and splitting.

    I can understand why they aren't too popular. Usually doubling down and splitting are the things that I have the most fun doing while playing BJ. Still, I would like to know about them.



    That is exactly what risk-averse indices are. They take a hit on EV, to produce a bigger hit on variance. A better scheme is to simply properly size your basic "unit" and your betting ramp to fit your existing bankroll and the RoR you are willing to accept, and play "by the book" using that spread. There are several "sets" of RA indices floating around. The general thing you will notice is that they raise the threshold for where you split or double, since those specific plays double your risk.

    You can run a sim to see what happens as you change split/double indices upward. You will see your EV drop, but you will also see the variance drop. The "optimal point" is the point where you maximize your EV (and you take whatever the variance happens to be). As you back off that optimal EV (by raising the index to split or double less frequently) the EV and variance both drop. I'm not sure how I would define an "optimal point" since any change hurts EV. The question you want to ask and answer with the sim is this: "for a specific play, and we can take the A-2 vs 5 from the other thread as an example, what effect does raising the index have on the EV and the variance. If it only slightly decreases the EV and significantly reduces the variance, you might choose to remember that index and use it instead of the usual one. Knowing that you are impacting your EV negatively of course."

    EV and variance will both drop as the indices go up. How much EV you are willing to give up to drop the variance a bit is a question open to debate. Again, my advice is to just bet according to some kelly-fraction and play optimally with respect to your EV, letting the kelly fraction be the thing that controls your RoR, rather than trying to limit the variance...

    that's what 99.999% of knowledgable players do...
  • Don Schlesinger covers them in Blackjack Attack. He gives the RA versions of the Hi-Lo Illustrious 18. He also shows you how to calculate them for your own system based on your tolerance for risk and bet spread.

    -Sonny-
  • Ray, Where can I get a more complete list of plays that reduce variance??
    Sometimes I lose the count, have a big bet out and get a split or double opportunity.
    Thus I would like to reduce the variance.
  • sage01- other than blackjack attack, I don't know of a better reference.

    I don't play RA, even in those situation where it may or may not be justified
    depending on how you look at things. Buying time, in a given situation, usually
    means that something else is wrong.

    Some would argue that if you wong and/or over bet your BR in some exceptional situation, a risk adverse strategy is the way to go. I personally think that the frequency of occurance does not justify the action. But, everyone does not share my opinions.

    The reference that I gave above will cover the most frequent double and split
    situations because you will be playing basic the greater percentage of the time.
  • Ray: I just looked at the Illustrious 18 in BlackJack Attack.
    If you don't know the count, how to you develop what are the best plays to not split or not DD? (both of which reduce the EV and the variance)
    I'm not sure how to read the table on page 61(2nd edition) without the index number
  • sage01- I don't have any references at my current location and I'm not sure that I could help if I did. It's not a subject that gets a lot of attention from me. You may get some answers by going to BJmath.com and doing a search on risk adverse strategies. You may be able to direct your questions to folks that have a greater interest in the subject.
  • sage01 said:
    Ray: I just looked at the Illustrious 18 in BlackJack Attack.
    If you don't know the count, how to you develop what are the best plays to not split or not DD? (both of which reduce the EV and the variance)
    I'm not sure how to read the table on page 61(2nd edition) without the index number



    RA indices are called "indices" because they are basic strategy departure plays that are based on the count. :) I suppose a pure B/S player could look at the wizard's web site, and look at the double/split plays that only slightly increase the EV, and then just remember to not make those plays, since any double or split increases risk/variance...

    But the house edge is going to creep upward as well, so you might decrease variance, but you decrease your EV as well...

    As I've mentioned before, the most risk-averse play of all is to just not play. :) Otherwise, the game is full of risk...
  • stainless steel rat said:
    RA indices are called "indices" because they are basic strategy departure plays that are based on the count. :) I suppose a pure B/S player could look at the wizard's web site, and look at the double/split plays that only slightly increase the EV, and then just remember to not make those plays, since any double or split increases risk/variance...

    But the house edge is going to creep upward as well, so you might decrease variance, but you decrease your EV as well...

    As I've mentioned before, the most risk-averse play of all is to just not play. :) Otherwise, the game is full of risk...


    Yes, you are robbing Peter to pay Paul and I've never been able to justify that kind of strategy. What may I ask is the benefit other than a perceived smoothing of variance at some unknown loss?
  • Ray said:
    Yes, you are robbing Peter to pay Paul and I've never been able to justify that kind of strategy. What may I ask is the benefit other than a perceived smoothing of variance at some unknown loss?


    The plus side comes in when you wong in with a big bet. Say $2000. You are already "in" with an advantage, so you have a positive expectation. By using RA indices, you can maintain that positive EV (which is multiplied by that really large bet) but you can reduce the variance of adding another pile of chips (I suppose worst case would be another $14K assuming you split three times and then double on all 4 hands. An EV of +.15 times $2000 might seem safer than an EV of .11 times $16,000 when the variance goes way up and a single loss puts you in a big hole.

    I have never personally used 'em either. But I have heard/read the discussions. I always thought of it as a weak attempt to have your cake and eat it too. :) This damned game simply will always have a significant luck factor even though it can be made a +EV game easily enough. We'd like to see that "bell curve" become a simple point, but it isn't going to happen...

    I also liked Sonny's answer in that RA strategies don't make sense if you are using some kelly fraction for betting, assuming the fraction is < .5 or so. By the time you get to 1.0 kelly things are getting very wild and going to 2.0 is most likely going to be fatal, if you play long enough.
  • I don't think its having your cake and eating it too -- it seems to me to be a smart play. If you have a slightly positive EV, would you rather risk it all on one bet? or risk it over the next 1000 bets? That answer seems obvious. Thus, with a positive EV, its best to do as many trials as possible so that you actually hit the positive EV and kill the variance. (And likewise with negative EV games, the best thing to do is make one big bet).

    Of course, you can't make your bet too little or it will take forever to make any money and it won't be worth your time. But taking small steps to greatly reduce variance while only slightly hampering your EV seems smart to me. Wouldn't you rather bet $25 a hand over 100 trials with a +0.53 EV, say, than bet $2500 on your next hand with a +0.54 EV? Obviously that's an extreme example, but I think it makes RA strategies seem rational and attractive.
  • To win in the future a player will have to have at least three casinos within a 30 radius playing different shifts on short trips. Vegas will always have action because the future will require you to "Hit and Run." No more than 30 minutes at one house and in Vegas you have 70 (?) within a fifteen minute drive. No player's cards, no ids, etc. You just "hit and run." It will not be fun anymore, and only people who are not greedy and stay out of the radar will survive. Playing a house once every two weeks may even be too much. A counter at a secluded Indian casino, etc. will be made within an hour. I can see playing minimum for 20 minutes waiting for a high count to come out, bang'm and hit the road. If you don't pull a high count within a half hour, get out. Tech is here, and will continue to improve leaving only the popppies to have a good time and lose their money. Corporations today will not continue to let counters have their way. Does anybody else see it any different?
  • jimpenn said:It will not be fun anymore, and only people who are not greedy and stay out of the radar will survive.

    I agree with most of what you said, but I certainly don’t think it spells the end of BJ. Here in SoCal there are plenty of profitable games and Vegas is ripe with opportunity. Sloppy dealers, weak shuffles, great promotions, ignorant floormen and other factors make these games very worthwhile.

    jimpenn said:Tech is here, and will continue to improve leaving only the popppies to have a good time and lose their money.

    The technology is becoming more advanced but is definitely not improving. The MindPlay system was a joke and it died a quick death. Technology will never be able to beat the Advantage Players because the players will always find new ways to beat them. The only thing, and I do mean the ONLY thing, that can detect an Advantage Player is another skilled player. A sharp pit boss can sometimes spot a card counter in a matter of minutes whereas a computer can monitor your play all day long without detecting them. The computers are neat and all but they are worthless if an unqualified employee is using them.

    There will always be beatable games and there will always be untrained casino employees. That is why I believe the game will always be profitable in some form or another. In the past few decades the profitability of the game has skyrocketed and I don’t see that slowing down at all:

    http://www.bj21.com/frames/index.html

    -Sonny-
  • Sonny, I think you are correct concerning Vegas Play because of the number of casinos available to play. Not being familar with Southern Cal. I can't comment on it. I do think for example, Aztar, Indiana is a good example of playing at a secluded casino, along with some Indian Casinos in Up State NY.

    Do you think the days of the player's cards are numbered for counters, in addition to limiting one's play to short durations and then moving on?

    Thanks Sonny, Jim
  • FunkyChicken said:
    I don't think its having your cake and eating it too -- it seems to me to be a smart play. If you have a slightly positive EV, would you rather risk it all on one bet? or risk it over the next 1000 bets? That answer seems obvious. Thus, with a positive EV, its best to do as many trials as possible so that you actually hit the positive EV and kill the variance. (And likewise with negative EV games, the best thing to do is make one big bet).



    That's the point Sonny was making. If you are playing a real Kelly-fraction, you won't be "risking it all" on one play. Because you would never bet more than a fraction of your actual advantage (using fractional-kelly).

    The point I was making was that your expected win rate is your EV times your average bet. You can afford to take a hit on your EV (if you choose) when you are betting really large units (such as team play with a big player) since a big unit times a small EV can still be a big number. And if it reduces variance signficantly at the same time, some might like the play.

    But for those that actually use fractional-kelly betting, it is not a good plan for bankroll growth overall since Kelly already optimizes for growth and risk and considers variance...






    Of course, you can't make your bet too little or it will take forever to make any money and it won't be worth your time. But taking small steps to greatly reduce variance while only slightly hampering your EV seems smart to me. Wouldn't you rather bet $25 a hand over 100 trials with a +0.53 EV, say, than bet $2500 on your next hand with a +0.54 EV? Obviously that's an extreme example, but I think it makes RA strategies seem rational and attractive.



    I don't disagree there. But the comparison doesn't make sense for a kelly bettor. 25 vs 2500 would be either way under-betting (25 unit) or way over-betting (2500 unit) assuming your bankroll does not change while you are making the above choice...
  • stainless steel rat said:You can afford to take a hit on your EV (if you choose) when you are betting really large units (such as team play with a big player) since a big unit times a small EV can still be a big number. And if it reduces variance signficantly at the same time, some might like the play.

    My biggest concern with RA play is that it may actually increase my long-term risk. In many cases the RA play will decrease both variance and EV but not proportionally. You will be losing more in EV than you are saving in SD. This will results in smaller short-term swings but because the ratio of EV:SD is smaller you could be increasing your ROR by making the “wrong” play. I don’t know how much of an effect it will have, but I think that several plays worth of lost EV will show a noticeable difference in long-term risk.

    I don’t think there is anything wrong with using RA plays, I just chose not to use them myself. Different players have different goals. Overall I try to play the strongest game I possibly can. I’m not worried about short-term fluctuations. They’re going to happen and I’ve learned to live with that. I’m focused on the long-term results. I want to win as much money as I can as quickly as I can. My main concern is maximizing my EV and minimizing my ROR and N0. That’s why I use my playing to maximize my EV and my betting to minimize my variance.

    -Sonny-
  • You've seen me explain my betting scheme on other forums. I use CVCX to sim the exact game/rules I will play, and then let it tell me the optimal betting ramp, so that everything is optimized, leaving out the guesswork. It's always nice to know _exactly_ what the expected win rate is for a particular game and bankroll. Which means for me that the idea of RA indices really don't come into play since my "risk" is already covered by the CVCX kelly-betting calculations.

    It is pretty easy to change a playing index number to be more "risk-averse" and then see exactly what it does to N0 and RoR/variance, as well as hourly win rate. Unfortunately not everyone knows about (nor understands) software such as CVCX/CVData and how much information they can provide to eliminate guesswork that can eliminate the small advantage we play with if one is not careful...

    Some think the optimal bet ramp is min or max, with max at TC of +2 and min at anything less. It is quite informative to see what such a scheme does to variance and RoR (it does help hourly win rate of course) rather than just guessing that it doesn't hurt much at all.

    Once one has the right tools, and then uses them, much of these kinds of discussions would not ever pop up, and beginning APs would be much better prepared to win.
  • Jim- We had the technology to completely automate a BJ game/table 15-20 years ago. The casino's have done some stupid things down thru the years, but even they learn given enough time. What can be done and what should be done are two different things.
  • Ray....Thanks for response. I now believe your right. People (like me) have been been concerned since the first Mindplay unit was installed. I guess I'm the type of person who worries about every thing before it even happens. I'lm 58 and still worry about situations that haven't happened and maybe that's why I don't play the market. LOL

    I'm not a seasoned gambler and believe in the old saying that I don't buy anything that I can't kick. It's not the fear of losing money. I get high by getting over on the big houses who have the advantage.

    Take care, Jim
  • The point of RA is not to decrease risk; but to increase win rate. If you generate optimal RA indexes, and then generate an optimal betting ramp using those indexes, the EV will indeed drop. But, win rate will increase. Why? Because you adjust the unit size to bring the risk back to the same level. The increase in unit size is greater than the decrease in EV thereby increasing overall win rate (SCORE.) So, you make more money at the same level of risk.

    In practice, everything cannot be optimal because you cannot bet $23.87. So you will always be over-betting or under-betting to some degree. But, on average, the use of RA indexes increases win rate by about 3%.

    norm
  • I seem to be missing something here.

    If I run a sim and ask CVData to produce optimal indices, are these therefore RA indices? I've always thought that reducing risk also must, by definition, also reduce the overall EV.

    How does RA indices avoid that pitfall and why would anyone use anything but RA indices if they maximize the EV and are better than the normal indices. What am I missing???
  • stainless steel rat said:
    I seem to be missing something here.

    If I run a sim and ask CVData to produce optimal indices, are these therefore RA indices? I've always thought that reducing risk also must, by definition, also reduce the overall EV.

    How does RA indices avoid that pitfall and why would anyone use anything but RA indices if they maximize the EV and are better than the normal indices. What am I missing???


    CVData can generate either Risk Averse or EV-Maximizing indexes. If you ask CVData to generate RA indexes, then use CVCX to generate a betting ramp based on those indexes; then EV will drop, unit size will increase by a greater level and SCORE/WinRate will increase - given the same risk. This is one of the reasons that EV is not a great measure of a strategy.

    The only reason not to use RA indexes is if it is impractical to increase the unit size (e.g. you have a huge bankroll.)
  • It appears to me that optimal RA indexes would stray from using mostly integers and that would complicate things. Instead of +2 it may be +2.15,etc. Otherwise, it would seem reasonable to say that RA numbers should apply to all count systems and strategies. If this is the rub on RA, I can see why when you consider things like rounding. Rounding to what?
  • QFIT said:
    CVData can generate either Risk Averse or EV-Maximizing indexes. If you ask CVData to generate RA indexes, then use CVCX to generate a betting ramp based on those indexes; then EV will drop, unit size will increase by a greater level and SCORE/WinRate will increase - given the same risk. This is one of the reasons that EV is not a great measure of a strategy.

    The only reason not to use RA indexes is if it is impractical to increase the unit size (e.g. you have a huge bankroll.)



    OK, so what you are saying is that using RA indices I can bet bigger, with the same RoR and lower EV, but have a better hourly win rate because of the bigger bets??

    If that is the case, why would not everyone just use RA indices all the time, and call the "other" indices (Won'g usual hi-lo numbers for example) something else since they would appear to be worse in most all situations???

    I guess what I am asking is why are not the "optimal hi-lo indices" RA by default, since that seems to maximize the win rate?? It appears to be so because of changing the bet unit/ramp which leads to confusion???

    This kind of goes back to an older discussion (not with you) about variance, and whether you think game X has higher or lower variance than game Y. And one says "X" because the SD number for that game is bigger, while the other says "Y" because the rules for that same are so good you can use a bigger betting unit to obtain the same RoR...

    Why don't all counting systems use these numbers to start with?
  • Ray said:
    It appears to me that optimal RA indexes would stray from using mostly integers and that would complicate things. Instead of +2 it may be +2.15,etc. Otherwise, it would seem reasonable to say that RA numbers should apply to all count systems and strategies. If this is the rub on RA, I can see why when you consider things like rounding. Rounding to what?


    There is always some method of integerization whether or not you use EV-Max or RA indexes. But, rounded RA indexes are slightly better than rounded EV-Max indexes on the average.
  • stainless steel rat said:
    OK, so what you are saying is that using RA indices I can bet bigger, with the same RoR and lower EV, but have a better hourly win rate because of the bigger bets??

    If that is the case, why would not everyone just use RA indices all the time, and call the "other" indices (Won'g usual hi-lo numbers for example) something else since they would appear to be worse in most all situations???

    I guess what I am asking is why are not the "optimal hi-lo indices" RA by default, since that seems to maximize the win rate?? It appears to be so because of changing the bet unit/ramp which leads to confusion???

    This kind of goes back to an older discussion (not with you) about variance, and whether you think game X has higher or lower variance than game Y. And one says "X" because the SD number for that game is bigger, while the other says "Y" because the rules for that same are so good you can use a bigger betting unit to obtain the same RoR...

    Why don't all counting systems use these numbers to start with?


    Yes, RA indexes are superior. And some older strategies have some risk aversion applied to their indexes. For example AOII. But, not much thought was given to RA indexes at the time HiLo was developed. In fact, the published HiLo indexes aren't even accurate for EV-Max and IMO the indexes in the 1975 edition of ProBJ are superior to the indexes in the 1998 edition of the same book.
  • You only get a higher win rate if you keep your ROR the same, right? If you are open to having a higher ROR, then the better EV would give you a higher win rate -- you would just be risking more. Its only when you bet less with a higher EV to keep the ROR the same that you do not have as high of a win rate, right? or am I confused?
  • FunkyChicken said:
    You only get a higher win rate if you keep your ROR the same, right? If you are open to having a higher ROR, then the better EV would give you a higher win rate -- you would just be risking more. Its only when you bet less with a higher EV to keep the ROR the same that you do not have as high of a win rate, right? or am I confused?


    Yes that is all correct. And the way to maximize the win rate for a specified RoR and spread is to use RA indexes and optimal unit size and optimal betting ramp.
  • hmm seems risk averse indices have come a long way since Schlesinger wrote about them in Blackjack Attack. seems back then he likened using them to chicken soup for a dead man. the attitude being what can it hurt? :)

    best regards,
    mr fr0g
  • sagefr0g said:
    hmm seems risk averse indices have come a long way since Schlesinger wrote about them in Blackjack Attack. seems back then he likened using them to chicken soup for a dead man. the attitude being what can it hurt? :)

    best regards,
    mr fr0g


    No contradiction there. Don and I have always thought that a 3% gain is no big deal. I'm just saying that the point of RA indexes is to raise win rate, not reduce risk. And they do. Just not enough to worry about in shoes.

    Optimal betting OTOH is very important.
  • QFIT said:If you ask CVData to generate RA indexes, then use CVCX to generate a betting ramp based on those indexes; then EV will drop, unit size will increase by a greater level and SCORE/WinRate will increase - given the same risk.

    Okay, now I’m completely confused. I thought that RA indices were calculated based on the players bet spread (among other things). Doesn’t it strive to maximize CE (instead of EV) by taking into account the EV, variance, and average bet at a certain TC? Often times a RA play will be less optimal with a small bet but more valuable with a big bet.

    In that case, how much will it matter if you are using a different bet spread than the one you used to calculate your RA indices? I always thought you started with the bet spread and optimized your RA indices based on it, not the other way around.

    -Sonny-
  • Yes, you tell CVData where you expect to make your max bet when generating RA indexes. But, that parameter doesn't affect results by much. There is a chicken and egg problem. If you are obsessive, you can generate what I call iRAs. Iterative RA indexes. You run a sim, then generate RA indexes based on that sim, then a new sim with the new indexes, and then better indexes with the new sim:) Easy with CVData as you just set the iRA option. But that is way overboard.

    CVCX optimizes the betting ramp. This is not the same as the spread. You decide on a spread based on what you think you can get away with. Then using that spread and your bankroll, and desired risk, and exact strategy including indexes; the betting ramp is calculated. You need to generate the indexes before creating the betting ramp.

    On your point: "Often times a RA play will be less optimal with a small bet but more valuable with a big bet." This is the point of an RA index. The play will be different for large and small bets.
  • QFIT said:If you are obsessive, you can generate what I call iRAs...But that is way overboard.

    Of course I'm obsessive! Why else would I buy software with a “Beat to Death” setting? :)

    -Sonny-
  • Sonny said:


    Of course I'm obsessive! Why else would I buy software with a “Beat to Death” setting? :)

    -Sonny-


    I thought "Beat to Death" sounded better than 4.5SD:) I rarely use that setting.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!