Enter the SECT

[JPFisher55]

Yes, I know that SECT cannot take non-showdown hands into account. True, you will not know when your advantage hand held up in non-showdown hands. OTOH, unless you always call down on the river with TP, TPTK 2 pair or even a set no matter what the board texture is, then you will also fold hands in which you were outdrawn. SECT is not perfect, but it is a useful tool in evaluating your luck.

The problem with AIEV is that the sample of hands for one player is too small. For the last year, I have run below my AIEV at AP. To check the site's legitimacy, I ran HEM AIEV for other players with hand totals between the several of them about mine. Sure enough they ran above their AIEV. I was jealous. When I added all of our AIEV hands together, the AIEV just about equaled the actual AI winnings. I agree that if you have a mult-million hand database, then AIEV is the best measure of luck. But if your hand database is in the hundreds of thousands, then SECT can help supplement it.

I agree that your SECT results do not alone measure your luck. You need to use it, your AIEV results and other stats. Percent to win for all streets would be another nice stat to have to help evaluate both luck and skill. I hope that HEM adds these stats in the future.

[pzhon]

Originally Posted by JPFisher55

Yes, I know that SECT cannot take non-showdown hands into account.

This introduces a huge bias in the examples I gave.

SECT will declare players with some playing styles to be lucky regardless of whether the player gets lucky or not. Other playing styles commonly used by serious players will be declared unlucky even if the player ran well above expectation, and should have lost even more money.

OTOH, unless you always call down on the river with TP, TPTK 2 pair or even a set no matter what the board texture is, then you will also fold hands in which you were outdrawn.

Different playing styles will have radically different rates of going to showdown with various types of hands. So, the errors in the SECT method do not cancel out.

SECT is not perfect, but it is a useful tool in evaluating your luck.

Please enlighten me as to a use other than to fool yourself into thinking that you are getting outrageously unlucky when you are not. The people raving about how wonderful SECT is do not understand that it is biased. It is returning the wrong answer. How is misleading people useful?

Suppose I make a program called "Optimal NL Solver" which returns a play in any situation. People say, "That's just what I need! I want to know the optimum play." But, my program always says "Fold." Sometimes it's right. Often it's horribly wrong. On average, it recommends playing with an overly tight style which loses money rapidly. Is my "Optimal NL Solver" useful? It would be if it lived up to people's expectations, but since it does not, it is useless.

The problem with AIEV is that the sample of hands for one player is too small. For the last year, I have run below my AIEV at AP.

So what? This means nothing. If you flip 100k coins, you are very likely to come up with more heads than tails, or more tails than heads. This does not mean you need millions of coins for them to be fair.

If it confuses you, feel free to ignore AIEV. It's useful to some players, particularly tournament players, high stakes NLHE players, and PLO players who often get all-in preflop or on the flop. It says little for players in nitty games where people don't want to get all-in for 100 bb with JJ, but what little it says is unbiased, so it might mean that it only takes 85k hands to get the reliability of 100k unadjusted hands.

I agree that your SECT results do not alone measure your luck. You need to use it, your AIEV results and other stats. Percent to win for all streets would be another nice stat to have to help evaluate both luck and skill. I hope that HEM adds these stats in the future.

If you want an estimate which converges to the right answer, and you have the SECT information and what HEM now reports, the right thing to do is to drop the SECT information on the floor. Ignore it completely. It's not what people want it to be.

More people might think SECT does what it promises than realize that it doesn't. However, mathematics is not a democracy.

This should not be implemented in HEM.

[JPFisher55]

Fine how about just adding your chance to win by street for preflop, flop and turn for all showdown hands like it has for all in hands and leave out the actual expected value?

[pzhon]

Originally Posted by JPFisher55

Fine how about just adding your chance to win by street for preflop, flop and turn for all showdown hands like it has for all in hands and leave out the actual expected value?

It might be useful to be able to filter by such values, in case you want to review how your opponents play when they flop sets, for example.

The value averaged over the times you went to showdown doesn't have a clear meaning.

[JPFisher55]

Originally Posted by pzhon

It might be useful to be able to filter by such values, in case you want to review how your opponents play when they flop sets, for example.

The value averaged over the times you went to showdown doesn't have a clear meaning.

If you knew that you were 55% to win preflop on average and you only win 50% then you would know that your luck left something to be desired for those showdown hands. True, those may overweight the hands in which you were unlucky, but it is better than not knowing. Also, I would like to compare my win% on the flop to my win% on the turn to see how much it goes up.

I agree that this knowledge is not a perfect indicator of strategy or luck, but I disagree that it is better not to know at all.

[pzhon]

Originally Posted by JPFisher55

If you knew that you were 55% to win preflop on average

First, you wouldn't know that you were 55% preflop on average. You know that the conditional average would be 55% over the hands which went to showdown.

and you only win 50% then you would know that your luck left something to be desired for those showdown hands.

Second, you wouldn't know that. Poker is not about starting with the best hand, or having over 50% hot-and-cold equity. That's a common exploitable conceptual error. See my video on StoxPoker, "Poker Goals and Nongoals" #1256.

I would like to compare my win% on the flop to my win% on the turn to see how much it goes up.

You would not get your win% on the flop or your win% on the turn, since you don't know what happens on the hands which don't go to showdown. This is not a small error. It does not cancel. It is HUGE. Work through the examples I gave before.

It really sounds like you want to build another biased statistic which will tell you the wrong answer. It will not tell you whether you are getting lucky or not. It will reward some playing styles over others. It will tell you that you were just unlucky while you are actually lucky, and are playing losing poker. Is that really better than nothing?

You do not have the information about the hands which didn't go to showdown. You want to measure Sklansky dollars? Fine, that's one consistent accounting system. The Sklansky dollars you win or lose on the hands which don't go to showdown are absolutely critical, and they will be ignored, causing you to misevaluate playing styles systematically.

I'm done. Good luck.

[rextyrann]

as far as i have understood the conclusion is following:

AI EV is "almost" always correct but very very limited, like only useful for sng/mtt players,

Street EV does cover more ground i.e. all showdown hands but in the long run will be misleading since its ignoring the nsd hands which should be taken into account.

Sklansky Bucks does both so it should be better than either one...

i do not know how skansky bucks work but what i have read from the contra-sect post they seem all to agree that sklansky bucks would do a better job.
in that regard id suggest we vote for that to be implemented in hm since street EV is probably denied as the final conclusion by hm-support!

so guys i do want something better then AI EV. if you do too vote for sklansky bucks!
Graphing Sklansky bucks

[JPFisher55]

SECT calculates your Skansky bucks for all showdown hands. This is incomplete because it cannot include your non-showdown hands. It may be biased for some players. So it is not perfect. IMO, though it is useful. IMO, percent to win for each street for all showdown hands would be useful, but have the same incomplete and potential bias.

However, AIEV includes so few hands that it cannot tell whether you are lucky or not. No stat can perfectly tell whether you are running lucky or not. It will always be a guess. However, IMO SECT and percent to win stats for all streets for all showdown hands would help tell whether you are running lucky or not.

[Alexdb]

Having scanned over the thread again, I've become in favour of adding SECT to hold'em manager.

The people that understand the game already can ignore it completely, and it's awesome misinformation for villains that don't know that it's misinformation

[omaha]

HHad to reinstall windows, and sect doesnt want to work for me

I have downloaded sect 3.2.7, win rar, and have java se dev kit 6 update 18, java 6 update 17 and 18 listed in 'add or remove programs'

When i double click on the sect, nothing happens.

When i try to dos it, i get a 'unrecognisezed opion - jarsectv3.2, could not create the java virtual machine

Any ideas what i am doing wrong?