$25----->25k Bankroll Challenge on Ignition

haha it's all good man I didn't think you were

Nice thread... good luck with it.

Ya, I think nitpicking about source data is a little meh. If you're explaining where it comes from, I think you're all good. As far as the data being against TOS, you're talking about specific online sites TOS, correct?

Thanks yeah I think most sites adhere to the no gathering of 3rd party data.

Okay late night studying some poker theory from Qing Yang's YT channel and came across this gem that is counter intuitive.

BTNvsBB SRP

We open BTN 2.5x, BB calls.

We have 9♥8♥

Flop comes T♦7♠4♦

We cbet B50 OTF (solver uses all sizing's here), BB calls.

Turn is J♦ - we turn a straight but flush completes. BB checks to us.

Solver NEVER bets turn!

Reasoning is if we bet OTT. Solver is mixing B33/B50/B75. Let's say we B50 OTT.

None of BB's one diamond hands are folding OTT given flop calling range so we are denying zero equity!

Very cool spot.

Okay time for bed.

Serious question doodoo:
You study so much - why not just stop this challenge , stop the studying and try to grind as much money as possible at 100/200NL and get to 500NL.

Isn‘t it the goal to make money with poker?

Imagine if it was possible to make even more money based on how much more knowledge you had over the rest of your opponents.

Just imagine.

will this mda data work in zone play? good luck with your challenge.

That can't be the only reason we are checking, because we would still gain plenty of EV by betting and charging villain if he had, say, AdTc (20% equity).

As an extremely simplistic model, say that we can bet turn 0.5x pot and check river.. or simply check the hand down.

If we check the hand down, our EV is: 11bb * 80% = 8.8bb

If we bet 5.5bb and check the river, our EV is: 16.5bb * 80% - 5.5bb * 20% = 12.1bb

If we bet 5.5bb and villain folds, our EV is simply 11bb.

So in this simplistic model, we WANT him to call with a diamond rather than "deny his equity".
So obviously, inability to deny equity vs one diamond hands can't be the sole reason for why we are checking.

Seems like solver shenanigans since they bet a good amount after 33% flop and a bit after 75%, and some of the offsuit non-diamond hands still bet

Yeah it works everywhere. It's player tendencies, they don't change much based on how much time you have to make a decision...although that could be a cool experiment.

Subbed. Gl, this is very useful to me that play at Ignition

Yeah I think your take is intuitive, you clearly want to put money into the pot if you are an 80% favorite so we still want to bet our hand in a vacuum vs all 1 diamond hands.

I think QY's point was to understand calling ranges and the GTO reasoning behind why we bet.

That's one of my gripes with the theory experts (I don't claim to be one although I understand theory). A lot of the examples are esoteric spots that almost never happen in game.

I still like his example because it illustrates the broader concept of how bet sizing works and helps me understand the spot better. You are correct though.

Oh another thing I wanted to bring up that I get wrong sometimes and wanted to get better at understanding.

TBJ and I were going over this hand (any Bayesian experts that want to crunch the numbers on this would be appreciated).

Here is the link:

https://forumserver.twoplustwo.com/69/on...

So we know this guy is 11/11 (VPIP/PFR) over 35 hands.

Most people will assume this is a nit even though we don't know for certain, correct?

But this is where statistics gets counter intuitive (at least to me). It's actually WAY more likely that he is not a nit and is just a regular running bad.

I'm hoping someone can crunch the math here but I'll post TBJ's data analysis and conclusion.

Out of a DB of 97.5 million hands from players with VPIP/PFR <30. Only 42,000 hands were from people with a VPIP of 13 or less.

42000/97,500,000 = .0000043 of all hands

Let's say the average VPIP for a NIT is 18 VPIP, now what are the odds he has an 18 VPIP over 35 hands when the data show's 11 VPIP?

Okay well we can use Tombo's calculator (shoutout to tombo from GTO Wizard).

Okay so there is around an ~82% chance that this unknown is a nit with an 18 VPIP or less after showing 11/11 over 35 hands. But then we realize only .0000043 of all hands were from players with 13 VPIP or less.

So now instead of it being likely he is a real nit--------->it becomes extremely likely he is just a reg that is running bad (yes other people can run bad not just you! lol).

I'll try to calculate this:

Probability of being a nit P(A): 82%
Probability of being a reg P(reg): 18%

Posterior Probability = Likelihood * Prior Probability / Evidence

Prior Probability = Random player with 11/11 stats over 35 hands being a nit (18 VPIP or less ) = 82%

Likelihood = .0000043%

.82 * .0000043

Probability of evidence = Probability of likelihood (.0000043%) * Probability of being a nit (82%) + Probability of likelihood (.9999957) Probability of reg (18%)

Only two possibilities for purposes of this discussion. He is either a nit or a reg.

Probability of evidence = .1800034486

Posterior Probability = .000003526 / .1800034486 = .00001956 = About .0002% = This player is actually a nit and not a regular just running bad.

I won't lie I probably butchered some of this math so feel free to correct me. I won't it take it personally.

The main thing to take away is that even very good players will incorrectly identify a player as a nit when it is with very high certainty that they aren't a nit and instead a regular that is running bad.

P.S

Check my math

Nice analysis! I'll double-check the math

Let's define a "NIT" VPIP < 20%

According to the population data, 2M / 97.5M players have a VPIP < 20%, so the portion of nits in this sample is 2%

Using my HUD stat calculator, if we take his stat at face value there's an 88.7% chance that this player's VPIP < 20%, making no prior assumptions about what the Stat ought to be.

Now calculating the posterior:

P(Nit|Evidence) = (2% * 88.7%) / ( 2% * 88.7% + 98% * 11.28%)
P(Nit|Evidence) = 14.3%

So, given TBJ's population data, we calculate a 14.3% probability that this player is a nit with a VPIP < 20%.

----

If we instead define a nit as VPIP <17%, we get these numbers:

Portion of nits = 0.42%
Evidence that VPIP < 17% = 77.8%

P(Nit|Evidence) = 1.4%

---

In short - even though his initial stats look nitty, there are so few nits in the iggy population that chances are his true VPIP is > 20%.

Thanks Tom! I appreciate the awesome breakdown. A+ post.

Yeah in my analysis I used <13VPIP which is why the numbers get a little goofy, I like your definition of a nit better (<20 VPIP).

So if we define a nit as <20 VPIP then we come to the conclusion that an 11 VPIP over 35 hands has an 85.7% chance of not being a nit.

That is such an important conclusion and why I really liked this exercise since it is conceptual so we can apply it to other situations. I think learning to label players correctly or at least what is "likely" is almost a separate skill set apart from being good at poker fundamentals since you need to understand at least basic statistics to apply it accurately.

Cheers!

Imagine making one quarter of a BB/100 less because of studying 5h less per day (after having thousands of studying hours already) but playing 3000k per day instead of 500…

Can't shake the impression OP's approach might actually be working on some level...

I lost the part where OP said he's studying 5+ hours a day. Def agree that no matter how good you are, if you never put the volume, you don't make $.

I'd say 8bb/100 qualifies as crushing, especially if you put in a lot of volume.

At PS NL200, over the past two years, only one reg has a winrate of >8bb/100 over more than 200k hands.

Same at NL100.

You are making the same mistake as the person who says the 11/11 guy is likely a nit over 35 hands is.

These graphs have selection bias because the better players move up. They aren't accurate representations of what real winrates are since the results are skewed towards mediocrity.

Whaaattt?

8bb at nl200 is exceptional wr, you can make close to 100k a year if you put lots of volume.

The biggest crusher of pokerstars small stakes, Seta-Beni, got about 6 bb/100 at 200z and 7 bb/100 at 100z. He was very exploitative and went heavy on MDA back then. You can check Onklebs journals on runitonce, there are 3 I believe.

Not sure about ishter11 but from a comment he made on meale thread (the prop bet of 10 bb/100 at z50), I'm sure he doesn't believe anyone is getting even close to that, and z50 is basically lowstakes.

Top 3 winners when we aggregate all limits from 25nl to 500nl on stars, and which show up having a favorite limit of at least 200nl currently, are:

CRAZY_YEHOOO 8.9 bb/100 over 1.3mm hands
energyvadym 8.7 bb/100 over 671k hands
2FAT2BTRUE 8.3 bb/100 over 738k hands
luismi1912 8.2bb/100 over 812k hands
Random_Bluffer1 8.1 bb/100 over 1.1mm hands

luismi1912 had his main limit at nl400 (probably this data is old I guess), and at that limit he got 7.5bb/100 over 296k hands

The others are/were 200nl regs, and this is their performances at their main stake:

CRAZY_YEHOOO 8.8 bb/100 over 791k hands
energyvadym 9.6 bb/100 over 485k hands
2FAT2BTRUE 8.6 bb/100 over 700k hands
Random_Bluffer1 6.1 bb/100 over 470k hands

Can't attest for the actual poker ability of any of these guys, but this is what the top winrates of regulars playing regular tables who might have moved up look like. So, sorry but 8 bb/100 on pokerstars low/smallstakes games is crushing 😀

Source: statname

Okay this concept will print you money once you understand it. This is very subtle but we have MDA to back up our conclusions.

Like everything in poker, it's all about nuance. A slight change in a board runout can mean the difference between a river call and a river fold.

I'm going to focus on the paired river vs paired turn specifically for this spot.

Let's look at two boards that look similar but are vastly different:

8♦7♦2♥7♠3♠

8♦7♦2♥3♠7♠

Most people won't give much thought to these two runouts because they look similar, since they are both paired OTT/OTR and we are in a 3BP facing a triple barrel.

But let's dive into the data:

Note: This data is filtered for COvsBTN3BET B30-B-B lines (my HH is HJvsCO but the same concept applies they will just be slightly tighter since it was B50-B-B and one position earlier).

The samples are not great but the discrepancies are undeniable. If you are facing a triple barrel on 87273 or you are facing a triple barrel on 87237. You fold the first runout and you SNAP CALL the second runout.

This is because your opponent expects you to fold low and middle pair to a turn barrel so you won't have many trips on the river in the 2nd runout but you will have more trips on the 1st runout. Remember----->the stronger your perceived range is the less they bluff.

HH that inspired this deep dive.

This HH isn't perfect (they rarely are) but the concept is what is important to take away.

Hand History driven straight to this forum with DriveHUD 2 Poker HUD and Database Software

NL Holdem 0.05(BB)
HERO ($6.70) [VPIP: 29.2% | PFR: 24.3% | AGG: 36.7% | Flop Agg: 41.4% | Turn Agg: 34% | River Agg: 37.6% | 3Bet: 11.5% | 4Bet: 13.9% | Hands: 336187]
CO ($5) [VPIP: 23.9% | PFR: 17.4% | AGG: 39.1% | Flop Agg: 55.6% | Turn Agg: 28.6% | River Agg: 28.6% | 3Bet: 15% | Fold to 3Bet: 0% | 4Bet: 0% | Hands: 46]
BTN ($5.43) [VPIP: 20% | PFR: 20% | AGG: 150% | Hands: 20]
SB ($8.28) [VPIP: 25.6% | PFR: 20.9% | AGG: 30.8% | Hands: 46]
BB ($5.27) [VPIP: 42.9% | PFR: 42.9% | AGG: 0% | Hands: 7]
UTG ($5.86) [VPIP: 18.8% | PFR: 15.6% | AGG: 60% | Hands: 34]

Dealt to Hero: 5♠ A♠

UTG Folds, HERO Raises To $0.10, CO Raises To $0.35, BTN Folds, SB Folds, BB Folds, HERO Calls $0.25

Hero SPR on Flop: [6.04 effective]
Flop ($0.77): 5♦ 6♦ 3♥
HERO Checks, CO Bets $0.37 (Rem. Stack: $4.28), HERO Calls $0.37 (Rem. Stack: $5.98)

Turn ($1.51): 5♦ 6♦ 3♥ K♣
HERO Checks, CO Bets $0.85 (Rem. Stack: $3.43), HERO Calls $0.85 (Rem. Stack: $5.13)

River ($3.21): 5♦ 6♦ 3♥ K♣ 6♣
HERO Checks, CO Bets $3.43 (allin), HERO Calls $3.43 (Rem. Stack: $1.70)