At the Blackjack Ball one year, Tommy Hyland came up to me saying he had a question. The preface “I have a question for you” is always a little unnerving, especially coming from someone you don’t talk to very often. So then he asked me … [wait for it …]
“Why do you recommend KO?”
I was surprised, not by the question itself, but because it indicated Tommy must have read something I wrote. I think most professional gamblers who know me don’t really read any of my stuff. But everyone’s been reading Colin Jones’s The 21st-Century Card Counter, which espouses the same principle—simplicity gets the money—but draws a slightly different conclusion.
Mies and I are Less-Is-More apologists, and CJ, who knows more counters than I do, agrees: “All the successful card counters I know (I’m talking six- or seven-figure earners) have made their money not by the complexity of their systems, but by aggressively attacking the casinos and getting in thousands of hours of play.” I would make only a slight modification. The most successful counters I’ve seen would aggressively attack certain casinos, but totally avoid others.
Some counters will go ahead and play anywhere, but I don’t think that’s optimal. There are casinos out there that will allow a shocking number of hours. They just don’t know what a counter looks like, or they are unwilling to take countermeasures against the player. Successful counters generally play these unicorn casinos often and hard, sometimes camping out for months. The late Big Red camped out on the same table for years! Think about that—spreading from 0 to 2x$300, with no cover, for 20-40 hours per week, for years. And getting points and comps on a player’s card on top of it all. (That might seem like heaven to some, but I’m not sure if I ever saw the man smile—not once.)
CJ presses the case for choosing a simple count system: “You can learn a much more complicated count, but when you consider 1) the complexity, 2) the time needed to really master it, 3) the room for error (which comes at a cost), 4) the amount of mental energy it takes to use it, and 5) the value of rounds per hour, I question whether you’ll actually make more money. … And my strongest argument for HiLo is that every major card-counting team I know of has used it.” To me, that last argument is actually the weakest.
Ironically, one of the teams CJ is talking about would be Hyland’s team, but Tommy basically said the same thing to me at the Ball. When Tommy started out, the Knockout Count (KO) didn’t exist, and other teams were using HiLo. So everyone uses HiLo because everyone else uses it? It’s a fiat count system!
Okay, HiLo is actually a good system, but if, on paper, HiLo is comparable to KO, then I’m very confident that KO will destroy HiLo in the wild. Why do I say that? Because extensive observation of actual APs in the wild shows that their performance is consistently below the system specs underlying the computer-optimal EV.
People have a skepticism regarding running-count systems, but there are reasons that a running-count system can produce results comparable to a true-count system, even on paper. KO’s indices are more precise than HiLo’s, due to KO’s effectively finer stepsize when we restrict indices to be integers. A single point in the running-counted KO represents roughly a 0.2% change in EV, while a point in the true-counted HiLo represents about 0.5%. If we were to learn fractional HiLo indices, that KO advantage would go away, but no one would want to learn that doubling 9 vs. 7 should happen at HiLo +3.2 (or +2.8, or whatever). Potentially, a running-count system could have tags that provide a better playing correlation for certain critical decisions. KO counts the 7 as +1, which is better for insurance purposes. And KO has a good pivot, so that regardless of the decks remaining, a counter wants to be blasting when the count has reached the pivot. Though the edge E might be estimated less accurately than a true-counted system, a count above the pivot tells us that we’re safely in the E >> 0 zone.
But look, no one estimates decks remaining well, especially when the discard rack and shoe are opaque. And no one does the division fast enough, so hemming and hawing and stalling is introduced to the game. Then the execution isn’t as smooth, and time is wasted.
Another argument that I sometimes hear, but thankfully not in CJ’s book, is that HiLo is more compatible with shuffle tracking. Oh please. Just stop.
If we’re going to talk about other beyond-counting methods, then the case for a true-counted system like HiLo gets weaker, and KO looks good. When I’m playing a complicated blackjack game, if I count at all, the counting part of things needs to be really simple, and not slow down decisions that might already be tricky. KO is the limit to what I’m willing to do when multi-tasking.
Every player swears that his execution is flawless, and it’s the other guy who should consider a simpler system like KO, but the data says otherwise. In every instance where we do a pop quiz or secret audit to test a player’s skill, the results are disappointing. (This is the scene in the movie clip where we all stand in a crowd, and one-by-one step forward to announce “I am the underperforming AP.”)
In recent years, I’ve become a bit disappointed by the AP community, primarily because the skill level overall is so poor, and the posers are tolerated. And I think there’s some idiocracy going on. I think we’re getting worse, despite all the new tools available to learn. I think the millennials are not as serious about the game, and maybe CJ unfairly takes some of the heat for that. But the best way to improve collectively is to be realistic about our abilities in the field, and utilitize the simple but powerful systems we now have available. By promoting HiLo, Colin Jones’s BJA empire is a step in the right direction (from UstonAPC, RevereAPC, WongHalves, and others). For a veteran or a new player, I’d recommend KO, but CJ’s doing good work preaching HiLo. It’s aiight.
Never miss another post
“ If we were to learn fractional HiLo indices, that KO advantage would go away, but no one would want to learn that doubling 9 vs. 7 should happen at HiLo +3.2 (or +2.8, or whatever).”
1. Most BJ authors that I’ve read never went to that level of granularity; the exception would be Stanford Wong. So this would a chicken and egg argument.
2. In (1D) VBJ, the APs I know do count cards and since the number cards seen is pretty much limited, the exact penetration is known with each (additional) card drawn. And, since we play each hand identically, any gain in EV accrues immensely, especially given 1,500+ hph playing speed.
The deck estimation, division to TC to one decimal point, etc arguments do not apply to VBJ (1D, 2D, etc).
1. “So this would be a chicken and egg argument.” I don’t follow that. Many people don’t understand how it is possible that a system that does not do true-count conversion (KO) could be comparable to a system that does (HiLo). I’m giving one reason: KO has a finer step-size on its indices–that’s a fact, not an argument. Then as an aside I point out that if HiLo were to use decimal indices, HiLo’s “granularity” would match KO, and HiLo wouldn’t be giving up any ground in this aspect, but no one uses decimal indices for HiLo; ergo, HiLo gives up ground in this aspect. How much of a benefit KO gets from this I have no idea. I am not about to run the sim showing KO vs. Integer HiLo vs. Decimal HiLo. I know it’s not a lot, but the total difference on paper between KO and HiLo is not a lot.
2. I think it’s pretty clear in all of my writings that I’m not talking about machines. As far as I know, Colin isn’t teaching BJA players to play video blackjack, but I am not a BJA member, so I have no idea. There is certainly nothing in Colin’s book about video games. But if your defense of HiLo is that on video BJ the true-count conversion doesn’t suffer from player inaccuracy in deck estimation, then I’d point out that APs who hammer Kitty Glitter never make mistakes on their true-count conversion either.
“but no one uses decimal indices for HiLo”
The closer I’ve seen from this was MIT using 0.5 rounded Hi-Lo indices.
I didn’t know they did that. Is that mentioned in a book (I’m behind on reading). Did they have some sims or reason to believe that was worth doing? As EVBandit mentioned, the insurance index is the only one often shown in decimal.
I simmed the MIT indices against HiLo RA indices and the difference wasn’t worthwhile using current, generally available rules. I haven’t seen it mentioned in a book. The trick with KO is to use indices with more hands, not more accurate indices. REKO indices are all +2 no matter how many decks.
“Many people don’t understand how it is possible that a system that does not do true-count conversion (KO) could be comparable to a system that does (HiLo).” – the Dean of APs
So Grosjean is talking about the *people* who didn’t learn BRH’s Unbalanced TC Theorem. Any unbalanced count, e.g. KO, can be true counted. You can even true count KO with a side count of Aces. In the case of KO being TC’ed, a lot of the indice-plays are negative (vs traditional positive TC’s in balanced counts).
2. My “chicken & egg” comment was that Hi-Li Indices to two decimals for 1D, 2D, 6D, etc, respectively, were not in the public domain. I had to search, Theory of BJ, Wong’s books, BJMath.com (website no longer exists), etc for these types of information. Card Counters knew indices were either floored or truncated.
My point was it was one thing if Indices to two decimals were made publicly available and counters avoided them vs counters (like me) who didn’t have access to them in the first place.
My comment was about the codification of indices with decimals in the public domain; and, how the lack of codification results in this chicken & egg situation.
Kitty Glitter…. Haha!
JG, I am a fan of your writing and your sometimes odd jokes but that one had me spitting my food out.
Yes, any system could be true-counted, and maybe people know that, maybe they don’t. But if you don’t apply that true-count conversion, people expect the performance to be worse than other systems that DO employ the true-count system, at least if we are comparing systems of the same level (KO and HiLo are both Level-1 systems).
I don’t disagree with any of those points, but I still don’t understand how the term “chicken-and-egg” applies. Example of chicken-and-egg: “How to get experience without a job, but how to get a job without experience?” So there are no decimal indices in the public domain, so you therefore use integer indices and the performance of HiLo is “suboptimal.” But now what–because HiLo is suboptimal, there are no decimal indices in the public domain? “Chicken-and-egg” implies some circle that we can’t break. I don’t understand the circle you’re suggesting. Or are you suggesting a “Catch 22”–where there is no good option (use KO which has precise indices but lacks an easy TC conversion, or use HiLo which has an easy TC conversion but lacks precise indices).
No one is suggesting trying to use more accurate indices with KO, so I don’t follow “The trick with KO is to use indices with more hands, not more accurate indices.” But I added to the blog what I think you’re getting at: that KO’s different tags could give it better playing correlation on some plays, meaning that we have a useful index for a play that HiLo might not do well with.
As discussed in CAA, if KO has a good pivot, it will enable the counter to capitalize on the bulk of the good betting opportunities, even without a TC conversion. That might seem like a free lunch. But the thing KO gives up is that you might not be able to accurately estimate the edge. That is, KO is good at telling you WHEN the shoe gives you E >> 0, but it doesn’t necessarily give you an accurate estimate of E.
It’s funny to me that posts about comparison of count systems generate the most online engagement! Thanks for the comments, guys.
KO indices, as published, depend on a clever memory trick as opposed to accurate values. That’s OK. An index that is inaccurate is better than no index at all. So, in REKO, there is only one index value: +2. With that extra simplicity, you can easily add more indices without having to remember more numbers. (IRC is adjusted by number of decks.) The end result is an easier system with the same accuracy, rivaling HiLo.
As for KO calculation of edge, it is poor at the start of the shoe, where there is rarely an edge anyhow, and poor at the end of the shoe, where we no longer get to play. Edge estimate is fine in the sweet spot in the shoe and edge itself is superior to HiLo at that point due to the higher betting resolution.
And yes, discussions of systems always get the most responses and I avoid them if lengthy. Frankly, I’ve always said to each his own. But, I also dislike the complication of RAPC, Uston APC, and Halves. I’ve used RPC ever since visiting the craggy gentleman at Rexford Place; but that’s just force of habit. If asked, I nearly always suggest KO or HiLo.
I think KO would destroy HiLo in the wild also. Is it possible that a level 2 or level 3 card counting strategy could be optimized and configured to mimic the KO strategy without the need to convert to a true count, where the starting count is adjusted for depending on number of decks in play?
RE: millennials, I hate to stereotype, but sadly you are right and it’s not limited to gambling. It’s not their fault. As a generation they were completely destroyed by being raised in an “everyone gets a trophy” environment. What you describe is true in the business world as well. For the most part millennials feel entitled to a good job and a good salary and don’t think they need to work for it and can’t handle being told that they are doing something wrong, so they never get any better. As a fairly senior (ie, old) software engineer I see this all the time.
I love your old post about handing off a seat. If the best player in the world needs to work that hard and pay that much attention to detail in order to get the money, how hard does a mere mortal need to work? No free trophies here.
It’s already been done with UBZ 2 and Uston SS. lol
I used Unbalanced Zen for a while because of exactly what you say. I think the true-count conversion is the messy part of counting for a lot of players (myself included), and doesn’t necessarily buy you much if the unbalanced count is well designed. I went from Unbalanced Zen to KO at some point.
It’s definitely easier using an unbalanced count in BJ compared to using a counting strategy like HiLo which requires a TC conversion, especially for a newer player to the game of BJ. I think it benefits a player in that they’ll be more accurate, effective, and relaxed at the table and they’ll be able to concentrate on betting the appropriate bet size when needed and they’ll be quicker at getting the bets placed which will seem more natural with the flow of the game at full speed with no need to look at the discard tray. The player will blend in better with players who are not counting cards.
If I was going to train players at BJ I would train them with an unbalanced system vs. a strategy that uses a TC conversion but that’s not going to happen anytime soon.
On decimal indices, one point that might be missing. There is a reason that insurance is the only index where a decimal index might make some sense. Insurance is the only linear decision. Pick any point on the chart of count versus edge, and all points at higher counts have higher edges. This is not true with any other index. When integerized, you can see trend lines, albeit they are curved, not linear, and can switch back. Attempt to magnify a section of the chart and look at fractions of a count, and you will see the line meander. That is, there are no points where a slightly higher count provides an edge that is always greater or always less.
As an aside, I find it funny when an EV is expressed in many digits of significance. If a number expressing the result of measurement of something has more digits than the digits allowed by the measurement resolution, only the digits allowed by the measurement resolution are correct. Taking this further, if you express a measurement in more digits than can be known, the number is misleading (another word for wrong).
In my simulations, assuming a jump spread at an advantage of slightly under 1%, and a simple depth-betting tweak for KO, KO was slightly outperforming Hi-Lo on 6D shoes, with around 50 indices each.
The paper advantage of KO on smaller numbers of decks is even larger than that. And in the wild, KO’s advantage in speed/accuracy/cover could easily add another 25% on top of the paper advantage.
One disadvantage of KO is that you need different index numbers for shoes and DD. The penalty for using DD indices on a shoe or vice versa is 15-20%. It might take more of an up-front time investment than hi-lo because of needing to learn both sets of indices, but it’s pretty easy to do with QFIT’s drill software.
Also, Hi-Lo tends to derive more marginal benefit than KO from indices beyond the 50th, because those indices tend to be farther from the pivot.
It’s unfortunate that the main book about KO has substandard indices in it. Everything below “Full” is simpler than the Illustrious 18, and “Full” is 4.26% weaker SCORE than the ~50 indices I use on 6D, and 15.4% weaker SCORE than the 50 indices I use on DD. I had to put a lot of original work into optimizing the indices. This is why there aren’t more people using it already.
CVData thinks it can calculate an index for 77v8 on 6D, and the index is not very high, but adding that index to the system does less than nothing for its performance. Maybe it’s nonlinearity, but maybe it’s just the variance and wasted high cards.
Yes, the fact that an index isn’t high does not necessarily indicate that it is useful. You can look at the chart and see if it is tight or not. You must also look at the index frequency. Splits have lower frequencies. IIRC, Zen didn’t even have splits in its original form.
I bought CAC2 a couple of days ago from Norm’s site and spent a lot of time running sims, and then decided to just stick with KO. In the real world I don’t care how big my edge is, I just need to be betting max when I have >1% edge. KO is good for that. With that kind of betting strategy CAC2 is only 8.2% higher SCORE on paper than REKO-F-6D. Speed differences and errors probably cost more than 8% no matter how much you practice the level2 true counted system. Plus KO frees up mental energy for other stuff.
James, what do you think about Red 7? Or perhaps double Red 7. It’s classified as a Level 2 count but functionally it’s nearly a Level 1 count.