I played my first complete 100 hand match with Cepheus today - I finished +10. I had started to play a few other times but had been interrupted, usually by a real Bovada game.

I have a couple of questions about the program...

1.) Is it a useful exercise to play Cepheus if I usually play $1/$2 6-Max LHE? Cepheus makes plays that I don’t see at $1/$2 - for instance, twice, when I bet the river and he raised, I reraised and he folded. He was right to do so both times, as I had the nuts in each case. Cepheus also reraises most times you raise preflop, so I got pretty loose check-raising after the flop if I caught any piece of it. He also nearly always raises preflop from the button.

2.) It’s only once, obviously, but is it a decent indicator that I’m solid heads-up if I can hang with Cepheus for 100 hands? I’ve been working a lot on my blind/heads-up game, so I’m trying to figure out if I’m making some progress. I got some good cards, for sure. But he got some too.

My inclination is to keep playing Cepheus, but I’m wondering if I’ll pick up habits that won’t translate to $1/$2. Just looking at the difference between $.25/.$50 and $1/$2, there are calls I would make in $1/$2 that I would not in the lower limit game because people bluff more in $1/$2.

Thanks for any responses. I do not have an inflated ego and feel confident that long-term, Cepheus would bury me.

how many hands did you play? i did a guest account and played 100 hands. Cepheus played horribly i assume because it was learning.

my guess it is soon adapts and after enough hands will beat your strategy, forcing you to change tactics. i am curious about its learning curve.

I am curious also. I may misunderstand, but I don’t think it plays adaptively/exploitatively; I think it just plays GTO all the time. I say that because you can see those charts showing what Cepheus is likely to do given any hand and sequence of prior actions.

Maybe it does adapt - I will look into further.

Focus on what it does pre and flop. If you do just that you be 90% better than most. After that the game tree get so big that what you looking at the riv probably won't matter as much unless you making some really bad fold or bluff.

It’s an animal pre. Complete the SB, you’re getting raised. If it’s the SB, you’re getting raised. If you raise, you’re likely getting reraised.

It confirms what I’d been thinking about HU: never call anything preflop. You’re either in with a raise or you’re out. I admit I love the nearly automatic reraise to an open raise, and I’m going to test it out at .25/.50. Not ready to try it at $1/$2 yet.

That’s how I want to play HU: like an enraged feral cat whose cage has just been opened.

I train with cepheus 2+ hours a week, and im sure i have logged more than 150 hours total, and i think it pretty useful. Sometimes calling the river knowing i will lose to see what it has. I feel like my fundamental approach and balance on the flop and turn have both improved a lot. The only annoying thing is actually runnong queries to see its full strategy is pretty painful, so sometimes i learn less than if i could easily review that for more hands.

Also i normally play no more than 30 or 40 hands at a time, iv done 100 before but it can get prettty tiring, even though it feel good on days where you stomp it.

; “I think it just plays GTO all the time”

This is correct.

“”That’s how I want to play HU: like an enraged feral cat whose cage has just been opened.””

A few things:

82% open range from the button isn’t actually high nor low. It’s exactly where it should be.

3 betting preflop from the big blind is a function of the buttons opening range. Once you memorize the gto opening range and 3bet range, you’re in a much better position to exploit buttons opening range deviation.

It’s not about unchained aggression; it’s about your expected value and the ev sources that provide profit:

A) showdown winnings
B) non showdown winnings

We can further divide (a) into these categories: (unimproved showdown value) and (draw value).

Since it’s correct to call 100% after opening on the button, there’s no preflop 3bet decision that needs to account for (non showdown winnings) unless the opponents postflop deviation is so significant to the effect that (a)we must tighten up the bottom of our 3bet range, or (b) we must expand the bottom of our 3bet range, depending on the opponents deviation. Notice that the hands on the margin are the most likely to gain ev vs opponent deviation. Thus without a postflop deviation read on the opponent, non showdown winnings are not a part of the 3bet decision.

Showdown winnings are where most of your ev comes from in limit holdem. This is because of the size of the bet relative to the blinds, which allows the caller to profit with seemingly very poor hands. However this doesn’t imply that you want action when you have a decent hand. Hint if your opponent is making a profitable call, you would prefer he folds. Only the nearly unbeatable hands are so profitable that you own part of the opponents stack.

This is why it’s best to think of betting as effectively reducing the opponents profit, because you cannot erase it.

So whenever you 3bet preflop, except with the strongest hands, you would actually want your opponent to fold, because calling is clearly profitable for your opponent.

Thanks, Bob, for that detailed and incredibly useful response.

As someone who has finally started to figure out HU, an 82% open range feels incredibly wide - I know it's not, but it feels that way, incredibly aggressive. I've widened my open range considerably from my early days, when I was quite an exploitable player due to my tentative play on the button/HU. But, I still struggle to muster up the courage to open hands like Q6o, J6o, 87o, T6s (for example) all the time, especially against a good player post.

At the limits you play, I hear what you're saying about the 3bet pre. But at lower limits, the 3bet pre has good fear value - many weaker players lose their composure and either a) fold immediately post if they don't hit anything, even when they have odds to call, or b) call down anything because they don't want to be pushed around, or take some weird line with garbage while attempting a "genius" bluff. I know for sure that no one is folding pre to that one more bet - opponents of any skill level will complete the third bet.

I understand, though, that that fear value would diminish to near 0 at a $20/$40 game or higher. And it doesn't make me better to push around inferior players at $.25/$.50.

Still studying Cepheus every day. I definitely feel like it is making me better HU. I'm a lot more comfortable in 3bet situations pre.

I'm not sure what the pool of live players that are willing to gamestart looks like, but if you gamstart 20/40 online, you're expected to be pretty good.


Exploitive profits in limit holdem are a smaller fraction of the pot if compared with no limit, but this doesn't negate your opportunities to profit more by 3 betting more or calling more.

........(nash equilibrium)
.................../\
................./....\
(deviation A).....(nash equilibrium)

Since (nash equilibrium) is the strategy that maximally exploits* (nash equilibrium), gto is synonymous with nash equilibrium. However, (deviation A) will create a new set of comaximally exploitive strategies that is different from (nash equilibrium) and the ev of both strategies for this new set of strategies will go up(because the pot has either been (a) bloated by a poor bet or call, (b) abandoned by a player that is unwilling to bet a hand that should bet, (c) abandoned by a player that is unwilling to call a hand that should call the flop or turn (d) going to be abandoned on the river.

if we play our part in the nash equilibrium, our ev for (deviation A) will be higher than the ev for (nash equilibrium) after the decision point. This will be true if we play the nash equilibrium, or the comaximally exploitive strategy(new set of gto strategies).

Those are the nuts and bolts of it. Now remember what I was saying about ev sources? Well there's ev we gain both whether or not we actively exploit due to the law of conservation of ev:

Ev lost = ev gained in a rake free game.



Rake is an exception, as are certain tourney situations(because chip ev doesn't equal prizepool ev). With multiple players in tourneys near the bubble, players that call shoves too much can actually transfer part of the prizepool to the other players that folded already because they increase their chance of cashing when there is an all in situation.

If the decision point where the deviation occurs(folding too much on the turn or river, for example) is on a future street, there may or may not be a good preflop exploit(If they raise 72o at 1% frequency there's really nothing to change). We want big deviations that we as humans can recognize and easily exploit.

Remember though, that your opponent's cards are basically a contract that grants shared ownership of the pot. Best hand wins. Since you will never erase your opponent's ev, your share in a (strategy vs strategy) sense will always be less than (pot). This is more true in limit holdem, but it's also true in no limit except with irrational play(insert ramble about sujective rationality).

Yes, I really felt like that all had to be said before I come around again with the ev source thing:

When your opponent deviates, this creates an exploitive opportunity. Sometimes this deviation is so great that we may adjust preflop on the margins(I would never raise 72o, haven't for a long time). So let's look at the margin:

hope I remember the cepheus opening range:

22+, A2s+, A2o+, K2s+, K2o+, Q2s+, Q2o+, J2s+, J3o+, T2s+, T4o(only hand to mix raise or fold at frequency I think, implying 0ev open raise), 92s+, 95o+, 82s+, 85o+, 72s+, 75o+, 62s+, 65o+, 52s+, 42s, 32s.

There is a limping range(which has a wide variety of hands in it), but that's another story.

think about these hands:

J2o/J3o
T4o/T4o
T3o/T4o
T4o/T5o
94o/95o
84o/85o
74o/75o
64o/65o

Notice that the left side ev is assumed to be either equal to zero(fold), or negative ev (-ev raise) vs cepheus.

This is why these are the preferred hands to add vs a player that folds too much. I have this idea that if one hand on the left of this list is a proper exploitive preflop open raise, then they all are proper raises.

Vs a player that calls, bets, and raises too much(both preflop and postflop), you don't get to realize as much equity. So you gotta fold more preflop with the hands that are unprofitable when called. I believe these hands are those on the right side of the list, and if one is a fold then they all are.

Link to play Cepheus?

Just type "Cepheus Poker" and it's the first page that will come up.

Bob, please write as much as you like, because I really enjoy reading about poker currently.

I was inconsistent about discussing stakes... I keep seeing $1/$2 online equated with $20/$40 live, so I should have said that the intimidation factor would start to dissipate at $1/$2 online. Just played against a player like that HU, who opened every hand except for one pre and was happy to go 3 if I opened pre.

I assume $20/$40 online is for pros only and I would get obliterated HU in short order given my current skill level.

Those hands you related were eye-openers, because most of them I currently fold HU unless I know I'm up against a tight passive player. I will try to work some of them in HU.

And I'm inferring hands like T6s, 87o, Q6o etc. are slam-dunk opens, which I've started to do in the past couple of days.

http://poker-play.srv.ualberta.ca

The game interface is a horror. You'd think they'd put $3.50 into making it better, but, nah.

Three more rounds:

50 hands: -400 (brutalized by flush over flush)
100 hands: +170 (I return favor with nut straight over straight with capped river)
100 hands: +260

I'm working a lot on when and with what frequency to lead out.

If you mean to bet from the big blind after calling a raise, I suggest doing it only on boards that will cause the in position player to check more often because of the high number of draws available. The most threatening boards are those that allow the most straights(567r allows more straights than 569r) and flushes(monotone boards happen, opponents make flushes on the turn and river, cest la vie). The boards that threaten the in position player the least are those that allow full houses and overpowering two pairs(notice how your 3 bet preflop range affects your ability to make full houses). I check a lot more in position on the former(more dynamic) board than the latter(more static); the big blind should bet more because of this in the former, and check more in the latter.

If you mean to bet after 3 betting preflop, I suggest only checking a small showdownable range(K9s can definitely win a showdown a decent fraction of the time, but that hand doesn't like to bet every flop). second pair no kicker or bottom pair on dynamic boards make good flop checks a fraction of the time(the weaker the pair the lower the betting frequency imo), pocket pairs don't improve very often on the turn nor river, which increases the need for more protection value in order to realize equity(paying 2 big bets is often better than paying 2.5 big bets), which is why I only check under pocket pairs a small fraction of the time after 3 betting preflop. Then I hardly ever fold after checking the flop with anything that I got there with. Only the truly hopeless hands like J9 on 456. Idk what cepheus does there but I check fold that hand without ye olde 100% cbet read.

I'm not sure of the bot's preflop 3 bet range. This is how I look at it, which is similar to how I handle the -ev/+ev margin at the bottom of my range, 3 betting however is a choice between two profitable options(calling getting 3:1 implies that when you have between 25% and 50% realizable equity you're profiting up to +25% of the present pot. It also implies that 3 betting must give a higher profit than calling in order for the 3 bet to be correct).

I think these hands are all very profitable either as 3 bet or call and I'll play them at frequency either way as default vs unknowns:

A8s, A5s, A4s, A3s, K9s, Q9s, J9s, T8s, 97s, 87s, A9o, KTo, QJo.

Anything better, except A7s and A6s, is a solid 3 bet to me, and anything worse needs a read to 3 bet(this implies that if 76s is a 3 bet due to postflop fold equity vs a tight player playing a standard preflop range, then 87s is a 100% 3 bet vs this particular opponent).

This might be of interest:

https://forumserver.twoplustwo.com/5...31/?highlight=

https://forumserver.twoplustwo.com/1...35/?highlight=

maybe im dead wrong, so this is just my opinion...

im not sure if u have any knowledge in chess, but this is like the equivalent of saying u'll play a chess machine 1 game a day and learn a ton from it and thus be able to compete with ppl who's been playing for years. i think it's more likely u'll find urself misapplying stuff that's too complex for any humanly brain. ur question seems similar to https://chess.stackexchange.com/ques...-computer-help

1) I think the habits u acquire from playing cepheus will significantly hinder ur ability to play any poker game. HUHU is just an entirely different beast. it opens like ~80 otb and 3b like ~30 from the blinds. then postflop, there's like a gazillion combos in both player's ranges. i dont see how you can learn stuff from this to even translate to a 3 handed game (let alone a 6 or 9 handed game). the pf ranges and postflop play is just significantly different...

2) i think the stdev for huhu is something like 20ish BB/100. u can be losing 2 BB/100 against it, yet still be ahead 1/6th of the time after 10,000 hands.

At least in chess, u'll see urself losing everyday vs. a machine. In poker, u can win and fall under the impression that u've made progress (or lose a lot more and fall under the impression that u've gotten worse).

also cepheus is prolly way tougher than a chess bot, bc a chess bot isn't weakly solved while poker is...

I agree about 3+ handed. On Bovada, though, playing HUHU is a fact of life if you want to play. Believe me, I wish the tables were 4-6 handed all the time as I feel more comfortable that way. But a lot of the time HUHU is the only option, or you have to play HUHU for 10-30 hands before someone else sits down. My life isn't such that I can make it to casino very often, but I do have a lot of 15-30 minute windows during the week when I have dead time. So online is what I've got if I want to play.

Cepheus has helped me understand the agro HUHU $1/$2 players better. I did not know what to do in those situations - when I didn't have good hands I would just get blown off the ball by the aggression - but now I have a better idea of what to raise from BU and what to defend from BB, among other things. Like I had no idea something like Q5o was a feasible raising hand from BU; that hand is a complete joke is most scenarios.

But if I'm against a passive player HUHU... yeah, I need to forget about Cepheus. Bluff-catching opportunities nearly evaporate. These players are check-raising the turn because you are behind, probably badly. With Cepheus, you almost have the justification that it's betting to make you indifferent about folding versus calling, anyway, so why not just call. That line of reasoning is not helpful at a real poker table.

I still enjoy playing Cepheus - I just waxed it +535 the other day (I got really, really good cards) - but I've gotten much more out of Stoxtrade's book on the topic of shorthanded. Still learning - ways to go. There is much more to shorthanded than HUHU.

I don’t think Cepheus necessarily strives to make you indifferent by betting. Instead Cepheus strives for self imposed indifference; by betting and checking at the right frequency Cepheus both maximizes value vs the nemesis strategy and self imposes indifference vs everyone else (if Cepheus bets or checks at frequency, this implies that both opinions have the same ev, and that ev is always a positive value on the flop and turn).

In other words, Cepheus plays a strategy that guarantees a minimum ev of zero, and any deviation by its opponent will cause an ev exchange greater than or equal to zero in Cepheus favor. Cepheus has made itself indifferent to its opponents mistakes.

However the river is the street where you do strive to make your opponent indifferent, or rather the gto strategy maximizes ev vs the best possible response, and gto achieves this on the river by betting a range that is as polarized as the available range would allow(the bottom and top of your river range might look different on all the possible boards; sometimes there is a group of middle strength hands but sometimes there isn’t; every situation is a bit different). Your indifference to call/fold river decisions is the product of Cepheus polarized range. Notice that there are many different possible ranges that would make you indifferent to a call/fold river situation, but only one range on that list maximizes value with strong hands vs the nemesis.

https://forumserver.twoplustwo.com/1...12/?highlight=

Maybe a more concise way of describing indifference in a gto setting:

if ev of gto is guaranteed to be equal to or greater than zero, this would imply that (ev gto vs gto = zero). To deviate would lose ev. Both strategies are indifferent to the game, and this indifference is self imposed by choice of strategy.

This is strategic indifference.

if gto bets and checks combo xx at frequencies on the flop or turn, this would imply that (ev bet = ev check in gto vs gto game). If neither player deviates, this value will remain constant and neither player will be able to increase ev by changing strategy.

This is self imposed indifference.

if river ev is maximized by both players, and assuming there's no such thing as a river that checks through 100% of the time(a reasonable assumption imo), this would imply that (a) the bettor has value hands, (b) the bettor has bluffs, and (c) the caller has bluffcatchers. If neither player deviates the evs for (gto vs gto) will be values that are less than or equal to (pot), which add up to (pot), and neither player will be able to increase this ev by changing strategy. If either player deviates from gto, there will be an ev loss greater than or equal to zero for the deviating player, which is inherently gained by the opposing gto strategy. Since neither player has incentive to deviate, and even though the bettor is earning much more ev than the caller, both players should be satisfied with this ev.

This is strategic river indifference.

Those types of indifference are not the same as the indifference of a river decision in equilibrium with a bluff or bluffcatcher. What happens on the river is a result of the bettor trying to increase his profit by value betting. The caller then considers the hands in his range that can beat a bluff(weaker hands don't matter here) and the caller makes the bettor indifferent to bluffing by making the odds he folds equal to the odds that the bettor is getting on his bluff(if im getting 9:1 on my river bluff, you should call me 90% of the time with the range of available hands that can beat a bluff). The result is that the caller will have a folding range in every possible situation when facing a bet on the river in a gto vs gto game; this creates the opportunity for the bettor to bluff exactly enough to make the caller indifferent with bluffcatchers.

This is opponent imposed indifference.

+1. 100% true; focus mostly on preflop and what it does on certain board textures.

You can also query cepheus to see what it does exactly.

An observation and a question as I continue to play HUHU on Bovada.

Observation: the rake is brutal. After 7-8 minutes, you're often both sitting there down $2. I haven't done the math, but the game seems near-unbeatable. The players are good enough - especially during the daytime when there aren't many players - that you don't encounter enough fish to get ahead unless you're really good. I'm still ahead on $$ after five months, but since I've started playing more $1/$2, I'm just treading water.

Question: What do I do with these hands from the button or facing a raise HUHU: 87o, 97o, T7o, 97s, T7s, J7s?

I feel like I just can't play these hands right, so I haven't been playing them much except as straight steals against players who fold the blind frequently enough (of which there are relatively few @ $1/$2). Against an aggressive raiser, I'm just folding them except for maybe J7s and T7s (then I usually flop garbage and fold).

I rarely hit straights or flushes, and I usually end up holding 2nd or 3rd pair, if that, against an aggressive player and getting bet at with a marginal hand that is vulnerable to redraws - if it's good at all - that has little chance to improve, and that thus holds up poorly against semi-bluffs.

I feel like if I'm going to get better, though, I shouldn't be folding these hands in these situations. Cepheus sure doesn't.

All thoughts welcome.

I raise all of them from the button and call all of them from the big blind, mixing in some three bets with the suited hands.

I think it’s important with those more middling hands is to understand you’re not going to win every single pot. So if the flop is akj and you have 87o, you can safely fold knowing you have nothing. Also if you 3bet 97s and get that flop, you can safely bluff knowing that if they raise you they probably have something good. Just play your range and range advantage and don’t get too out of line.