PERFECT Poker | Reviewing Solver vs Solver 6-Max Cash Hands.

[@peripheral8]

Best educational poker content on KZbin.

[@TH94191]

Pete I would like to see you play against 5 solvers in a 6 max to see you react to their play, that would be some good content!

[@LightYagami-tf7ig]

this please

[@andthenwhat9603]

I feel like I do that everyday on the site I play lol

[@TedJones-ye1ud]

That is a great video idea 💡

[@dchen0_alt]

Tried running the 88 vs KQo spot in Wizard and it looks like there are some weird multiple-order blocker effects: On the turn, when you give BTN 50% and a smaller size, BTN prefers betting 50% with A7cc/K7cc and smaller with A8cc/K8cc (presumably because it unblocks more continues). When BTN takes the 50% line, BB counters by calling 77c but folding 88c on the four-flush river (as 77c blocks flushes with the 7c). So on the river, BTN prefers bluffing 77 no club but giving up with 88 no club.

[@jmfitzy2]

That’s wild lol thanks for sharing

[@johannesgerry4319]

Check back at 20:15 might be lots of calling A/Kx flush draws and some Q/Jx flushes on the river that just cant fold. Aswell some flopped flushes in villains range. Aswell you unblock value which is bad.

[@raoeoiku7549]

Great video concept. For the analysis, I think on some of the trickier spots it could be worth bringing up actual solver output to better understand frequencies and range construction, as occasionally we'll probably end up with fairly obscure lines being shown.

[@FrankJMEscurra]

This!

[@Solwa54-y7c]

I'm gonna take a wild guess and say 88 doesn't bluff the river because it only blocks Ac8x (which may even be tempted to bluff prior to the river? Maybe not.). Any 9,T blocks AcXx, KcXx and JcXx he calls pre (and those combos could be more likely to call F and T). So those should make better bluffs. But then again, I'm struggling at 25NL, so what the hell do I know.

[@TedJones-ye1ud]

Best informational poker channel on yt imo

[@simonaylwin9467]

Another point: first two hands are straight out of Doyle Brunson 1978 classic. 1) betting into the opener with a set hoping for a big pair and 2) playing the low full house ‘real fast’ to get action from the trip queens. Fascinating

[@Lichtinsicht]

love this format. i would really like to see some smaller pots too, as i really like to learn when to fold!

[@coldblood92]

Would like to see the solver looked at in some of these hands. First example being 88 in position. This way we can see maybe a lower pair decides to fire again or everything gives up. Like the concept. Would love a deeper dive in some situations

[@janfredrikhalvorsen3611]

Yea agreed

[@jonathangroover1111]

I could watch this type of content for hours. Amazing stuff.

[@simonaylwin9467]

Really good. It would be fascinating to see what the other hands the solver would play the same. For example what other hands than 66 bet lead on the first board.

[@professorvegas]

Pretty cool; kudos to your student who came up with this. Kind of interesting that the script doesn't get the showdowns correctly; you would think that would be the easiest part to do. I think that the solver output does not include who wins when there is a showdown, and there would have to be a second component to the process that evaluates the final hands to add that feature. You mentioned that there were 40k hands completed here. I wondered how big the game tree is here; you mentioned it is pretty restrictive. Also, is the solution limited to head's up, situations or are the results from a multi-way solver??

[@phrog6073]

legit every fucking person has done this lmao

[@michaelarend4787]

No it is not multiway and the show down is fixed. You are right, the evaluation is not done by solvers. It was a simple error. Everything works fine now. I am working on cool stuff with this.

[@redraw0160]

@@michaelarend4787you’re the student that built this I assume? Can you explain what we are seeing here and how you built it? Out of interest

[@michaelarend4787]

@@redraw0160 Sure. What Pete uses are hand histories created by an app I built. The app uses Preflop Ranges and Solved Postflop Trees to let 6 Bots play against each other and creates Pokerstars like Hand Histories. These hand histories can be used to find statistical deviations in your play or in the play of the pool and for so much more. But right now, what you see are only the raw hand histories in a replayer.

[@glenterris3980]

More marginal/close spots. Less coolers, we know that money piles in, in those spots. Tougher spots like having 99 OOP on K82 two tone with no fd blockers as an example.

[@joeythelips9668]

Did no one see AQs folded MP vs UTG? I don't have a chart but is this really folded ever?

[@TheStein36]

I’m not sure how this solver is working exactly. I’m assuming it doesn’t do multi-way pots and hands are randomized to only be random Villain vs Hero.

[@dasein1137]

Great content! I love to see solvers in action. More please...

[@richcheckmaker]

13:41 AQhh folds from UTG+1 to an UTG open? Is that GTO?

[@michaelclifford6659]

This is the coolest thing in the GTO nerd shit realm I've ever seen

[@manuelgrub4738]

I was curious and ran the spot of the KQo vs 88 hand, and in my sim 88 basically never bets the flop. Am I missing something? Has anyone else tried running the spot and had different results?

[@liveslowlivesimple]

not every solver has the same outputs. there isnt just one way to play a given spot - its all part of a much larger tree

[@wilsonvinas5906]

Was 88 on the button?

[@88jouman]

Nice video! Haha I laughed so hard when u said that the bot was probably tilted after losing a hand😂😂 classic

[@ahaaha8462]

Why does qt call the 3bet on flop? Is there any bluff 3b by solver just curious

[@fruitpunchchan]

I'm curious, how does one get a solver to play against itself?

[@benjaminfreedman302]

love this format bro

[@Nytemaster]

This is good stuff. To irmprove this maybe you could show each player's strategy on their action. Even if it was just hero's it would definitely add to the context of where the play derived. Having all player's strategies would be a nice bonus (on their action of course).

[@ScottiMusicNow]

You were shocked by the flop 3 bet in the bottoms full hand because we shut down the bluffs but what bluffs does he have on QQ5? Maybe broadway badkdoor flushdrsws but that’s it.

[@RokasN999]

If solver does not have an all-in (AI) function working properly, as it has been seen in 2nd hand QT vs 55, where it did not shove AI IP nor it raised AI OOP, it may also skew the strategy calculations, because it could see no possibilities to get all money in pot in those lines where AI is not working properly. So logically solver may have to do more pot building moves, like betting more often and bigger then if let’s say compared to if it would have an option to get AI properly. So that could explain the flop re-raise. It could also be that flop checkraise is skewed to be in too high frequency too then if let’s say AI would work properly. I had very similar strategy outcomes when I ran my sims without proper bet sizings. Just my 2 cents. I hope I am wrong

[@couteauxduseignanx4625]

Amazing concept ! Please more !!!

[@yukoncpa1]

Where can this app be purchased?

[@marlofeliciano4800]

How to access this tool, man? New to your channel.

[@superjenspeter]

Hello Peter. Injoy your content, especially this one. If you have imported the HH into PT4, then maybee, there is a time issue. I had the same problem on the Ipoker network.

[@bears-dont-cry-ds4zq]

88 seems to be splitting jams and checks on the river.

[@michaelclifford6659]

Cool video carrot! I ran a solve on the 1st hand and found no donks. I tweaked ranges a tad to see if i could find a scenario with donks and could not. Do you think there is an error in the process? let me know if you found donks in the 1st hand. Thanks

[@michaelarend4787]

I checked that. EV of call, bet 17, and bet 36 are the same, so maybe your rake is a bit different or your preflop ranges are and that makes it a 100% check in your case.

[@michaelclifford6659]

@michaelarend4787 I tried tweaking preflop in pio by adding more/less Ax I'm btn or more/less pocket pairs in lj and found no donks regardless

[@w451-qx3kx]

​@@michaelarend4787 I nodelocked to b33 and OOP EV went from 15.1 to 14.1. I solved on the flop 6c4s2d and it donks 14% of range. Then tried on 6c4s3d and it donks 23%. I'm guessing it has to do with a few things: OOP has no flopped straights or OESD. OOP can't bet range to protect vulnerable overpairs or top pairs because it is being called too often by overcards with a club anyways. IP has a high amount of Axs so lower flush draws have bad reverse implied odds meanwhile AQs, AJs, KQs, KJs block IP's calling range. The few flopped sets (and therefore the best bluffs like A5s) want to protect the rest of the checking range from being too condensed. The donking range might be due to the use of subsets where 3 low cards would otherwise be a typically desirable donk.

[@philiphardy8650]

One of you best PC, really impressive, more please 👌

[@ЕгорИванов-и2о]

Carrot Corner, hi! A friend of mine adviced me this channel and I should tell you that the content is awesome!!! Answering your questions: 1. I’d definitely contact the support about the bug, cause I’m a full-time pro and don’t want it to happen with me 2. The topic covered in the video is super interesting!!! 3. Would be great if you made a video about the solver vs silver 6max stats (their WTSD, WWSF, WSD, etc.). Especially WTSD is interesting)))

[@smalltimep]

Thanks to the student who made this, I've been wanting to see something like this for ages

[@howjoec2638]

Im going all in every hand and hoping for the best

[@GTOfishh]

The QT vs 55 hand. U talk about a concept in which u want to "just call to leave the bluffs in". Is this actually a concept in GTO? Isnt there just equity distribution and the value hands in each range that define the strategy? And catching bluffs would then be sort of a side product of all of your middling value hands. I totally agree that as human its strange to 3bet flop though. What do u think about that?

[@epikouros93]

Fore sure yes. This is a special node. I would say the OP player on this kind of boards ( and when IP stabs more than the 50% rule on the floop) XRaises not only a polar range(bluffs - strong value). XRs also merged( 50-80 EQ hands, underpairs, 5x , overpairs.). Maybe vs that merged part of that range the IP want to decrease the EV as soon as possible...Also Peter refered smth related to the runouts that counterfit the bottom full house. Maybe that's a factor too...

[@epikouros93]

In addition hands like 44, 33, 22 can make this sick 3bet to deny EQ from all the bluffs that have EQ ( AJ,AT,A4s etc..)

[@GTOfishh]

@@epikouros93 Yes i like this a lot. Cause you wont push any equity vs a Q anyways. So why in the hell 3bet? It can only be to push equity vs pocket pair that have 2 out, to leverage vs merged x/r range. So "catching bluff" still isnt a driving factor in GTO theory. Thx for your insight.

[@epikouros93]

@@GTOfishh The 3bet floop don't deny EQ from 66-AA. Theses hands from CO are pure defends vs 3bet. Their EV is getting mutch lower. But the reason they dont fold( vs 50%3bet size) i think is because they make nuts on the future and cooler the BU strong value...One another factor i just thought is the SPR..When CO XRs marged range floop,will not invest in general in the future( and the spr after XR is quite big...So how IP destroys these hands? I guess to develop some 3bets. But again when you are IP and have positional advantage blah blah.... there is not a very strong argument for that 3bet flop strategy to be dominand. These were my thoughts. I am trying to understand 1st the game tree from the macro prespective ( distributions ,ip, op, spr)

[@hahaGGed]

Til in solver land a full house beats quads…no wonder it’s such a mysterious and confusing place!!

[@anhkiennguyenhoang2963]

GTO call flop in the 1st hand??

[@huggins_will]

Very interesting format.

[@davidricardo5629]

Cool format and I'd like to see more of it, maybe fiddling with the parameters somewhat (smaller opens from early positions, limiting the positions that coldcall). But from a business perspective, the potential marketable product here isn't a course based on this program, but selling an easy way for people to watch customizable solver fights and export the HHs

[@anthonyf596]

I agree. Id love to see solvers with different bet sizings, and deeper than 100 bigs. This content is top notch

[@FastPitch357]

With all the shade Pete throws at HCL players, we need to raise him a bankroll and send him to play with like a 3% freeroll for himself Edit to add: are the Preflop ranges that are used for raked games or just straight chipEV. I play in time charge games and recently found chipEV ranges and they allow for a lot more cold calls

[@coreyhuggins5785]

what sizes are allowed on flops?

[@michaelarend4787]

You can adjust them. Here I used the solves Pete uses: 33/75/150.

[@BenjaminRaven]

this is a sick video, more please!

[@koso2930]

Nice video dude - the solvers need better names like they got in chess :) Would love to see a video without limitations to bet sizes!

[@ChrisM-wv4gs]

If you arent 3betting the 3rd nuts what are you 3betting nuts what are you 3 betting? At that point might as well just say u don't have a 3betting range.

[@epikouros93]

There is a 3bet range ( not hier than 3-5%) the question is why such a strategy could be a thing....

[@ChrisM-wv4gs]

@@epikouros93 then what are we 3betting for value?

[@FizzyToni]

Which rake structure was this in? The preflop ranges

[@michaelarend4787]

It uses Carrot Corners NL100 Ranges, but every Range is possible to be used.

[@ekw555]

you said there were 40k hands? did a hand ever go multi-way? is that even possible?

[@michaelarend4787]

It uses Pio, so no Pio cannot solve multiway. But the preflop ranges are carrots ranges and with this ranges only 1% of the ranges will get multiway to the flop. So the stats are pretty solid.

[@ekw555]

@@michaelarend4787 I understand what you are saying, but I thought solvers were only playing HU and this is set up like a 6 player table. so, I was not exactly sure how it was being done and what was possible. though I guess "can't" is "can't" so no matter what goes on preflop, it will never go multi-way postflop, unless it's a 3 way all in, lol.

[@michaelarend4787]

@@ekw555 Yeah that ist true. These hands are simply not in the included in the hand history.

[@d_b_]

The concept is interesting for entertainment, but as a viewer it seems like theres too many unknowns viewing this way. If there was an RNG to indicate what line the solver was taking, and a way to peer in to the PIO-style tables with actions across your ranges, it would be more meaningful. In this current format, a lot of hands left the viewers wondering "Why" and "What hands _would_ do so-and-so here"...

[@toineb777]

Love it! need more

[@michaelarend4787]

There will be cool stuff. The hand history is only the data base.

[@LokalgottPoker]

Good Stuff more please!

[@KabooseVAL]

loved this

[@josephsouthron5759]

Would like to see some of the more boring hands that dont show down on the river.

[@Dominik-pe6kj]

Actually nice content

[@YorneE]

This is so cool but people can't find out about your content and future content because it will mean less EV for me :^)

[@JohnSteill]

Hi Pete, I'm a big, big fan, but this video frustrated me. Sometimes I'm skeptical that humans can get near to GTO. And It enrages me when people in the coaching space make up s$%# to "explain" a GTO solution which they couldn't predict. I can do poker word salad too. Props to you for being transparent, but when the solver surprises you, no way I believe you can "figure out what it was doing" on the fly like that. I'm currently getting my ass kicked daily by hybrid poker, and while I'm just starting, maybe, to get better at the easiest edge cases, I feel like understanding is very far and getting farther. Why you shouldn't care about my opinion: I'm a shit-reg, barely beating easy low stakes games over several thousand hours. When you drop the "You don't deserve to be playing poker and should quit." You should flash my picture, and add, "like this guy."

[@mrbwatson8081]

If the computer gave me a pot I did not actually win…. I would spit my tea out all over my screen and shout WTF….. 😮then I would go all in every hand 😅

[@michaelarend4787]

unfortunately the bug is fixed :)

[@chepper8405]

Great!

[@pokersoftwarereviews5051]

Personally, I'd pay for this kind of content, but only if the villains were programmed to make the kinds of mistakes that real-life opponents make, such as e.g. (a) under-bluffing most turn and river nodes, (b) not check-raising their semibluffs frequently enough, (c) calling way too tight facing a c-bet on the flop, etc. Without some level of adjustment for these kinds of near-universal mistakes, I think the resulting content will probably end up being "interesting but too academic", in my humble opinion. Also, please make sure the rake levels are suitably large. I play at a whopping 10% rake (capped at $15 per pot) mainly because Australia has only one casino per state, and each state's casino is tight with the local government, allowing them to basically rob us blind each and every day. Anyway, obviously your preflop ranges have to be massively skewed towards value in the presence of such a ridiculous tax on every decision to put chips in the middle, and those of us who are winning players are forced to play outrageously exploitative poker. For example, I've folded AKs preflop facing a 3-bet, because I knew my opponent wasn't 3-betting pocket Q's in that spot. Another time, I folded pocket aces on the flop facing a 2/3 pot bet, because I knew my opponent was a complete rock who refused to put chips in the middle without two-pair or better. Basically, I've done terrible, terrible things in the name of exploitation, and in fact these adjustments are necessary where I play. In general, I'd like to see, not only how unraked GTO bots plays each spot, but also how raked exploitative bots play against bad exploitable bots in the same spots, especially when facing the same runouts on flop, turn and river.

[@Nytemaster]

I really like the premise of what you are asking for here. But for now I think watching a pure equilibrium game and understanding those dynamics would be important to follow first. Once that has come to critical mass then I do think your suggestion would be a very valid place to go. Some of the reasons being we have no context to what those bad players may be in a video format such as this. We would also have to understand just where the metrics are on what is "normal" play. Besides, it would be much easier to code equilibrium and validate it from a pure GTO standpoint first and move on from there working out the bugs like we've seen in this video. Definitely some merit to what you are asking though.

[@Samdeman90]

wtf are all these coolers

[@scottyheav]

wow make a solver that doesn't know when a pot is chopped. LOL