On Thu, May 8, 2008 at 6:57 AM, Bradley M. Kuhn <[EMAIL PROTECTED]> wrote:
>> Regarding interface, my first thoughts have been to rigidly define PF
>> hand range notation (eg, 64+ includes 86o, but not 98 or AA),
>
> This sort of hand-range notation is a very good idea; I've read that
> some of the proprietary tools have this sort of notation.  It would be
> great if you wrote up a spec for the notation and put it in the Wiki!

Well, pokerstove understands "64+", but I'm not clear on what it
covers exactly. I have a feeling it means "64, and any hand that rates
better against two random cards than 64".

> Yeah, to make a pun -- I think this needs to be flushed out more.
> Defining the quality and value of draws is difficult: just think about
> the complicated language often used by expert players to describe them.

Maybe. I don't know. When I play, I don't look a lot deeper than
"drawing to nut straight against possible flush draw". There's a clear
difference between a nut straight draw and a bottom end straight draw.
There's a clear difference between a 2 card draw (45 on 23x) and a 1
card draw (5 on 234). There's a clear difference between a 9 out, 8
out, or 4 out draw. While there may be further subtleties, hopefully
one can come up with a model that makes maximum use of the clear cut
areas, and neatly packages up the complexities into some "advanced"
parameters. Or something.

> This is a level of complexity would should probably ignore in the first
> implementation, so we don't go overboard before making something usable.

Yeah. When I picture this tool, I picture something that lets you
drill down, or zoom in, to further levels of detail. So you might
start out a given analysis with a basic model of Hero and a basic
model of Villain, defined as "generic NL10 player". You then start
adding parameters like you have AK and raised PF. You could then zoom
into certain kinds of flops, those where you have a pair, those where
you miss. For certain of those kinds, you could zoom further in,
tweaking the behaviour of the model, like "what if he raises 50% of
the time with bottom pair" etc. Then tweak even further like "what if
he semi bluff raises with flush draws, but only if there's no ace, and
he doesn't hit any pair".

You get the idea - the complexity is not there, but can be added if
and when you want to.

> The complexity of your example worries me.  It's not readable without
> really learning the notation well.  You'll note that as I respond to

Pretty much all notation can be graphicalised.

> your subsequent notation proposals above, I get less and less enthralled
> with each successive notation suggestion.  By the time you are writing

Yeah it was kind of ugly. I've vaguely thought of accepting a wide
range of English-like syntax, so "flush draw", "2nd pair or better"
and "TPTK" might be accepted as alternative notations for the more
precise notation I gave an example of. But again, much ugliness can be
prettified with a GUI.

> Expert poker players stumble over describing these things and the
> distinctions between the quality of various situations or choices of
> various actions.  I've seen even in the last five years in poker
> substantial confusion about what terms mean, and the hunt for simple
> terms to describe the complex situations.  We shouldn't therefore jump
> too quickly to find the "perfect notation", but instead make something
> that reads reasonably well to an bot/simulation programmer as an API,
> and then worry about the "assembler-style" notation later.

The engine for all this seems the most straightforward part,
particularly if PokerSource what I think it does. Given clear
instructions like "Raise 50% of the time UTG with KQ" and clear
questions like "what % of the time does he hit TPTK on a K high
board", the programming is straightforward. It's the human side which
is harder.

> The minimal hand range notations you suggest at the top should
> definitely be done, though.  That part is simple and clear, I think, and
> would fit well with the API I suggest.

Ok, I'll try and write it up somewhere.

Steve

_______________________________________________
Pokersource-users mailing list
[email protected]
https://mail.gna.org/listinfo/pokersource-users

Reply via email to