Moved from PFC: HUD sample sizes

[AlfreD2]

http://www.pokerforum.ca/f53/hud-sample-sizes-29930/

Been wondering this lately...
What is a sufficient samples size for stats such as 3bet, fold3bet, steal, fold to steal, WTSD, W$SD, CBET, fold to CBET and aggression factor?

[JodaB.]

I'll get to answer this soon.

[RWPKRPLR1]

Nice thanks for moving this!

I have recently signed up for TPE and have been watching video after video. They have a good 6 part series on HM2 that goes through the mechanics, HUD and reports. Although its for HM2, I still watched it, as the HUD info is still relevant to PT4.

So my HUD now looks like this.

Line one: Live amount BB, Hands, VPIP, PFR, Aggression factor
Line Two: 3BET, FT3BET,STEAL, FTS
Line Three: CBET, FTCB, WTSD, W$SD

POPUP: VPIP, PFR, 3BET by Position (will be adding more to this when I get familiar with these)

In the series he briefly goes over sample sizes and his table examples only show large samples so you cant really get a good feel based on that. He suggests that the core stats should only be used when you have less than 50 hands and all the others should be taken lightly until you have 1k hands on a player. My problem with this is the significant gap between 50 hands & 1k hands!

In micro MTT's you rarely get the chance to get that type of sample size, You're consistently getting moved and the mix of players in the micros is astronomical. so right now, like in the video, I am only using the core stats under 50 hands, at 100 hands I start using my popup to see positional awareness and at 250 hands I start looking at my other stats(very lightly tho).

Kinda curious if this is ok?

Thanks, Ryan

edit: I'm also submitting a request to add a feature to PT4, min sample size for a specific stat. What do you think of that? I think it would be great so you do accidentally use a stat that isn't relevant.

[JodaB.]

Heres where some of the confusion is. When playing cash games, the blinds stay the same, the stack sizes generally stay the same, and the payouts stay the same (none). When you sit down for a few hours a few days a week with the same players, stats become very accurate and useful. If you play 10k+ hands with guys (I just made that number up) you can get reads like check raising turns too much, or doesn't bluff the river often enough an so on.

For mtts you are correct, we don't get enough of a sample size, not even in 180 man games. So we need to use the HUD differently. If a player over 10 hands is 100/0 and on the eleventh hand he raises we can feel like he limp a lot of garbage and now has a good hand. We don't KNOW that, but if we had to bet money on it thats what we would assume.

If we play 50 hands with a player and they are 11/10 we can assume they are tight abc. Theyve played only 10% of 50 hands which is 5 hands. I might assume this player knows how to steal in late position even. I might think they know how to bet an Ace high flop with air.

Aggression factor isn't so useful, I dont' know anyone who uses it.

Think about steal %. That happens from bu or cu, I think. An orbit is 9 hands and how often do you get a chance to steal, not too often. So 100 hands or 10 orbits you might only have 5 or 10 chances. So the % won't be an accurate indicator. But if someone has stolen 3/3 over 40 hands and they are stealing, you might reshove slightly wider, just not too wide.

What hands are standard to 3bet? JJ+ AQ maybe? thats only 4.2% of hands. So 4 in 100 times, but we know we might not get any of those hands in 100 times. And this is all provided someone has raised in front of us and we get a chance to 3bet in the first place. Or perhaps we got JJ and AA 2 hands in a row, so our 3bet % is 2/3 chances so 66%.

So what do we do? We think. If hes 15/12 over 100 hands we can expect him to 3bet a reasonably tight range. Because fish don't play 15/12 they player 60/30 or 30/20 or 4/4 (tight mouse).

What if we see someone 3bet and show down 66 on the river? The we might suspect they are 3betting 66+ ATs+ AJo KQs then we can check his 3bet stats and see if they support that.

Seeing hands at showdown is really the huge stat. If I see someone limp preflop mp with 72o then I only need 1 hand to know there vpip there is 100%. If I see a guy show JTo from the bu then I know he steals with any broadway.

Don't get stuck thinking we need to use stats as correlations to range %'s. They are a guideline to the types of opponents you will face.

Something that really helps is start labeling players based on there play: limper, loose, tight abc, easy simple accurate tags and then color people and compare it to there hud stats and see if everything fits and makes sense.

[RWPKRPLR1]

Thinking in terms of exact % for ranges is exactly what I was trying to do, Which was a big leak of mine up until a week ago. Yes I used my HUD to define a players image, but I really didn't use it until I got a specific no. of hands on a opponent, That's why I was so curious about samples. Then I would use a stat, by itself to try and determine what their range was in a specific spot and try to exploit it. With no further knowledge of the program, I thought this was the right way of going about it, But NO it's not, like you and the Video's explained.

In the past week, with the help of the video's, the series explained in detail how we should go about using a HUD in MTT setting. (I didn't know there was a difference in use for cash and tournaments) Like you stated, we should use our HUD as a guideline to tell use what style of player we're playing and give us an idea of potential spots we can exploit.

So now with my new HUD setup I don't spend that much time worrying about having sufficient sample sizes before using it to define a exact %. I spend more time on defining who my opponent is and what their potential weakness's are. I use the total No. of hands to gauge how accurate my assumption may be.

I also learned through all of this that using an individual stat by itself is so wrong, we should be using stats together to define what their range may be. I will use your example of a 11/10 being able to steal.

So we're are in BB and we`re facing a raise from the button. We`re 30 bb deep. We take a look at his stats, over a 100 hand period we see that he is playing a tight 11/10 and is stealing 85% of the time with 6 opportunities, we also look at his VPIP by position to see if he is positionally aware. According to VPIPBP, he is positionally aware. Now we can determine that he is not playing 10% of hands on the BU, But based on everything we looked into he might be stealing with a looser range. If I`m in this spot with a premium hand, I will go even further and look at stats like F3BET, AF & CBET. Since he is opening up his range here and I think he may fold I may just flat if his stats look like he may fold a high % of the time. I may want to keep him in if his AF is high and is CBET % is high.

As you see I`m not worrying about the sample sizes as much, I am looking at stats together to gauge what a player is doing.

[JodaB.]

Ya very good post, youre learning for sure.

As for flatting premiums I don't recommend it. Heres the thing. If we feel our opponents will fold a lot to our 3bets then we should be taking some of the half decent junk we normally fold to their open and 3 bet it as a bluff AS WELL as 3betting our premium hands.

Think of it like this. Sure villains folds often, but we he does flat or 4bet get it in we cooler him always, that value and its the best way to win chips especially early game. If we don't get value from Aces then we aren't likely going for value with JJ QQ KK AK.

The flip side is, if we add bluff to our range we often pick up dead money postflop and can occasionally bluff or hit flops when flatted.

Consider ALWAYS going for value, its just if you think villain folds a lot then add bluffs. Don't take out your top hands just because villain doesn't call-value is where we make profit.

I used to have trouble with this, I didn't understand why we bluff and value bet in the same spot, this is a decent example.

[RWPKRPLR1]

OK OK thx! I'll be at the tables this weekend working on my stat reading skills, value betting and advance my play against the various types of opponents. After I'm done, I'm going to create a document profile for each type of player(I will post what I have in another thread on here). I'm going to attempt to pin point their common tendencies and how I may exploit them. Please feel free to critique these profiles.

Thanks for your help,
Ryan

[JodaB.]

OK OK thx! I'll be at the tables this weekend working on my stat reading skills, value betting and advance my play against the various types of opponents. After I'm done, I'm going to create a document profile for each type of player(I will post what I have in another thread on here). I'm going to attempt to pin point their common tendencies and how I may exploit them. Please feel free to critique these profiles.

Thanks for your help,
Ryan

glgl!

[RWPKRPLR1]

TY TY This NIT is going to try to evolve! lol Something off topic I need to ask. What screen recording software are you using? I'm looking for a good free program, so I can record my play. I have tried a couple of programs in the past like Microsoft expres​sion(only 10 min video in the free version) and one other(can't remember the name of it, but it crashed my computer). So I have been very hesitant to download another, any suggestions?

[JodaB.]

TY TY This NIT is going to try to evolve! lol Something off topic I need to ask. What screen recording software are you using? I'm looking for a good free program, so I can record my play. I have tried a couple of programs in the past like Microsoft expres​sion(only 10 min video in the free version) and one other(can't remember the name of it, but it crashed my computer). So I have been very hesitant to download another, any suggestions?

camtasia, not free, might be a trial.

[RWPKRPLR1]

Ya not free 30 day trial, $299 for full version

[Sponsored content]