Pokertracker - aggression factor useless?

[Supern]

I have found the aggresion factor quite useless for loose players.
They may have an extremely low aggresion factor due to calling alot.
But they may bet quite loose as well and bluff alot.

1. How do you determine if a bettor is a loose or a tight bettor?

2. How do you determine if a caller is a loose or a tight bettor?


My answers:
1. I guess the best way to go is the Bet/Raise/Ch-raised stat for each street. And notes of course if you have any.

2. The best use for the aggresion factor is to determine if a caller is a threat I think. Low aggression and loose player means he is calling with all kind of crap.

[excession]

1. How do you determine if a bettor is a loose or a tight bettor?

Not sure exactly what you mean - if you are trying to spot bluffers from those who have the goods then aggression is quite useful - ultra aggresive players won't have the goods all the time and tend to bluff more, passive players who bet at/raise you probably do.
You can also link it into pre-flop tightness - the tighter they play the more premium hands they will have so you can expect more agression from them post flop (as they should have the best hadn more often). On the other hand, a player who has a Vp$iP of 50% and and aggression of 3+ is trying to do a lot of stealing..

2. How do you determine if a caller is a loose or a tight bettor?

Ah - spotting a calling station - pre-flop check their 'cold call%' if it's over 5% that's pretty fishy IMHO.. post-flop check their fold to flop bet% - less than 50% they like to call down too much... - players who call only with a good hand (weak-tight) will have high fold to flop bet% (over 80%), low WtsD% (below 20%) and high W$SD% (over 55%) - these folks can be bluffed on the flop but slow right down without a monster if they stay with you..

[Supern]

[ QUOTE ]
1. How do you determine if a bettor is a loose or a tight bettor?

Not sure exactly what you mean

[/ QUOTE ]

I have found there are players with high VPIP and low aggression (below 0.6) that bluffs alot. If you see them bet and look at their aggression stats you can't see that because they call alot as well.

I guess aggresion stats works better for tight players.

What I want to spot is how likely they are betting a draw or something like second pair.

[K C]

The only odd thing here is that I can't believe more people don't realize this. Aggression factor calculated the way PT does is more than useless - it's harmful

Without getting into too much detail here, and I know that there are going to be plenty of skeptics who see this stat on a pillar:

The primary purpose of a stat like this is and needs to be putting a player on a given range of hand strength. There are secondary uses such as determining the probability of as certain move like betting or raising in a given situation but those are best left to other stats. Oops, I forgot, PT doesn't have these stats /images/graemlins/smile.gif

Essentially the real problem with AF as a measure of bets+raises/calls is, as our friend mentioned, the fact that it doesn't account for how loose or tight a player is.

Let's take 2 different players and see what can happen. Player 1 is a rock who only bets and raises with the best hands. Player 2 is a loose aggressive who bets and raises wuite a bit as well as plays a ton of hands. Player 2 calls more as a percentage though so his AF is going to be lower than the rocks, perhaps even quite a bit lower in fact.

Looking at AF though how are we going to put these players on a range of hand strength? We aren't, and it's easy to see how this stat is useless. The rock is playing very strong hands and the LAG is playing weak ones. This stat doesn't help us at all here and actually if we don't understand it it can have us folding too much to the rock and not playing back enough against the LAG.

You cannot calculate aggression in any useful form without considering degrees of looseness, period. I've actually had to write new formulas for Poker Sherlock as there isn't anything out there at present which even comes close to calculating this properly, and I'm looking to modify things even further soon to get an even better read on things. We actually had to put the standard AF in the program as so many people think so highly of it that they may not consider our program without it, but the ultimate goal is to educate them about the fact that this stat can be very deceiving in a lot of instances and isn't really reliable at all.

KC

[MyMindIsGoing]

[ QUOTE ]
You cannot calculate aggression in any useful form without considering degrees of looseness, period.

[/ QUOTE ]

Why do people say things like this? "period" If you are not willing to discuss things what the hell are you doing on a forum?!?!

[K C]

Hey what makes you think I'm not? I'm stating my opinion though and it's up to you to provide counterarguments - not just some lame attempt at flaming me.

KC

[Mendacious]

I play mostly PLO8, but aggression is calculated the same way. I think for this stat to be utilized successfully, you need to take into account how often the player sees the flop as a baseline for what types of hands your opponent likely has.

Beyond that, it gives you an idea of what strategies are most likely to succeed against a player.

Don't bluff twice against a low aggression players generally, if he called you once, he'll call you again. Play your good hands fast against low agression players.

Do slow play and re-raise aggressive players. Sometimes a second bluff is needed to take down a pot.

A re-raise from a non-aggressive player means you are beaten, from an aggressive player you may have to consider calling.

etc. etc.

[MyMindIsGoing]

It was no flame, I just think it sux when people have that "I am right, you are wrong, no point even tanking more about it.".

[Supern]

[ QUOTE ]
Beyond that, it gives you an idea of what strategies are most likely to succeed against a player.


[/ QUOTE ]

This is very dangerous. I have several very loose players that bluffs quite a bit and have an aggression factor of around 0.6.

So if you don't have any notes on him and he bets you look at the agg and say: "he must have something".

Some players are betting semibluffs and bluffs and calling och checking their real hands.

It seems like the Agg factor works quite bad for loose players (VPIP 30+).

What about using the following stats instead and look at the W$SD%?
* "Bet/raised/ch-raised" for each street to see if he is dangerous when betting. This stat includes raises as well which is bad but it gives you a clue.

* "Just called" for each street to see if he if he is dangerous when calling.

[OrianasDaad]

The best way is via observation. Watch what they bet and call with, and take copious amounts of notes.

K_C is correct about the inherent inaccuracy of aggression factor, and does a good job of explaining why. I won't elaborate, and probably couldn't improve upon what he said.

I will note that I often see moves from players that their stats don't support. I watched a 70/0/.3 check-raise a LAG with an OESD today, and if I am observing closely, find that these anamolous situations fairly frequently.

I've been lax in my note taking requirements, but since I've moved down to one table (and loving it!), I make at least one note on a player for every showdown. It's not long before I've got notes on everyone at a table, and usually not much longer after that I have really good notes on everybody at a table.

These notes are usually supported by statistical reads, but sometimes they aren't. It's these differences that allow you a better chance to make correct decisions at the table.

[Supern]

Glad to hear from you Dad!

It seems like the only reliable stats are VPIP and PFR.
And to some extent Went to SD% to see if he is easy to bluff or not. /images/graemlins/wink.gif

I agree that notes is the way to go.
If the player:
betting trips or slowplay
betting draws
bluffs alot
induced bluff
how he behaves after pf raise - bluffs on flop, turn
if he bets against a pf raiser - what hands
etc

[SackUp]

The agg. factor is actually a really good indicator, but just not necessarily on its own. You need to use the stat in conjunction with the other stats.

The other posters are correct that an AF of 1.3 may be misleading b/c a dude sees 50% of flops and is just calling so often. That doesn't mean the AF of 1.3 is off, it just means you need to account for the vpip as well. I wish I had some of the link to other discussions on this. Maybe someone has one as a favorite or can do a search for them.

But essentially you have to look at how many hands the guy is playing and then base the agg off of that. A guy who is playing 50% of hands with an agg. factor of 1.5 is betting/raising a heck of a lot more hands than a dude with 20/1.5. So you first need to look at the range of hands the guy could be on and then the aggression.

This is obviously more complex than just look at the AF and know how the guy plays...but hey it cannot to everything for you /images/graemlins/smile.gif Maybe there is a way to combine the numbers into something more meaningful. If you could combine the vpip and AF numbers.

[excession]

As anyone who has read my PT articles is aware there is serious problem with the way that Aggression is calculated at the moment. Someone who checks/folds 16 times out of 20, never raises, calls once but bets out first 3 times will have a PFA of 3 for example. If you just check to him on the flop hoping that as he is 'aggressive' he will bet back at you and you can trap him, you are going to be very disappointed.

What aggression really shows is how likely a call is. A player with agression 3+ will call you rarely. One with PFA of .3 will call you down a lot.

When I was doing the last NL autorates I thought that this flaw was fatal and spent a lot of time looking for other ways of measuring true 'aggression' or the likelihhod of escalating a pot(which is really bet+raise ratio to check+fold with calls being pretty neutral) on the PT stats available for autorate. None really works in practice. Actually it is very rare that someone's aggression stats are very misleading especially after a decent number of hands - as a rule of thumb an aggressive player can be checked to and you can still expect a bet and a passive player is showing strength if he bets out and you should pick another spot to run a bluff. This means that for most situations you can still use the agression stat for what you need it for.

What it doesn't indicate so well is weakness (folds to flop bet%) but now with PAHUD you can have that stat up for all players at all times anyway (as well as cold calls pre-flop% and WtSD% which are also quite handy)

Remember you can pull the actual stats up for any player and if you are looking at a big pot it would be worth doing this - in the meantime until PT (or maybe PokerAce for his HUD) redefines aggression the current PFA ratings are flawed but still useful.

[Supern]

How many hands do you think is needed to draw any conclusions besides VPIP?
I think 200 is the minimum for aggresion and 500 for BB/100.

[Supern]

Which stats do you use?
And which ones should really be avoided?

It seems like the "bet/raised/ch-raised" and "just called" with W$SD% is a good indicator.
Or is it some kind of trouble with them?

I have to agree that the value of the AF stat is very questionable. In my opinion though, the biggest problem is that it is horribly named. I think most people would agree that in most situations raising or folding is a better option than calling; I've seen that stated as a general rule by many people and in numerous poker books. This is the general philosophy that the AF stat seems to be trying to rate. But is that truly a measure of aggression?

If you have one player who NEVER calls, their AF is infinite, seemingly indicating that they are inifinitely aggressive. But what if that person does nothing but check/fold except when they have the nuts (which they raise)? Is that person really aggressive? Conversely, you have a player who never checks or folds; they either bet, raise, or call everything and see a showdown every time. Is that person really passive?

The fundamental problem with using the AF stat for measuring aggression, is that checking and folding are not factored into the equation at all. Those are almost by definition the MOST passive actions you can make! How can you possibly rate how aggressive/passive someone is without taking their most passive actions into account? By folding you're basically saying, "I give up, you win". By checking, (unless you're slowplaying) you're saying, "My hand sucks, I'm just hoping for a cheap showdown." That's about as passive as it gets. By comparison, calling a bet seems downright aggressive. At least you're in there trying to make something happen and giving your opponent something to think about on the next street.

AF has some value in the sense that good solid players will tend to have a higer value because they're following the "fold or raise" philosophy, but that's not really a measure of aggression in my book, just a measure of solid play.

[excession]

It depends on the Vp$iP - if someone is ultra-tight (say 12% Vp$iP) then you will need 5 times as many hands to get a good read on aggression as against an Xtra-Loose player with 60% Vp$iP.

For the normal Vp$iP of 20-30% I'd say the PFA settles down somewhere in the 40-50 hand range..

[allenciox]

Ok, the real answer is that what we mean by aggressiveness is different in different situations. Once I have more than 100 hands against a player, I look at his detailed stats. Then I rate him three ways, which I put on his notes. I update this again for 200 hands, 300 hands, etc.:

checks/bets ratio: for high ratios I rate "ps" (pot stealer), for low ratios, I indicate "nops". For neutral ratios I do not indicate. A pot stealer is someone that tries to pick up pots when noone has yet shown strength.

raises/calls: for high ratios, I rate post-agg (postflop aggressive). For low ratios, I rate post-pas (postflop passive). Neutral ratios get post-neu.

calls/folds: high ratios get rated cs (callstation), low ratios get rated nocs(no-callstation). Neutrals get no rating.

Finally, I have one other measure: Based on first action after preflop raise:

bet/check: high ratios indicate con (continuation better). Low ratios are nocon (no continuation better)

It would be great if PAHUD or PT calculated these automatically, but until they do, I have to do the self-ratings.

[Supern]

I really like your idea.
What numbers do you use for your ratios?
And do you use the Total% or Flop% (too bad total includes pf)?

ps, nops, cs, nocs, con, nocon

In NL, if the AF is above a 2.0, he's probably full of it about 40 % of the time.

[Mendacious]

I agree with this.

Also, I think that it might be helpful to calculate aggression giving greater weight to Raising than to simply betting. In my view, raising is at least 2-3 times as aggressive, in every sense as betting. You can't begin to equate the standard post-flop lead out of 1/2 the pot to a raise.

This tweak alone I think would greatly increase the usefulness of the statistic.

I would use the formula (Bets + 3xRaise)/calls = Post flop aggression. The number would inflate a bit, but the really aggressive players would stand out the most.