HoldemResources.net: Beta/Feedback - Page 26 - Free Software -- Two Plus Two Poker Forums

kevinjet · 07-24-2012, 09:10 AM

Quote:

Originally Posted by plexiq

SQL Table: I just meant i will add an "action" column similar to the query table (ie: display the X/R/F/C actionstring)

Oh, you meant "Query outline" and "hand outline"...
Action letters that are generated dynamically, following spot selection (either selecting line at strategy or clicking a tick at outline) - it would be good. But its not as important right now, as to make the whole view enough to analyze the hand at one glance.
The polishing could be done while essential functionality is being beta-tested

Other items:
- Auto-select actual played line &
- show known hands and actual actions
- show cached ranges at hand list
are most awaited at the moment.

Quote:

Originally Posted by plexiq

And for unrestricted results this may be problematic re: memory usage if a large # of hands is batch calculated.

Ok, let's store whole tree only if user chooses analysis heavier than linear ICM and it found warning/error.

plexiq · 07-24-2012, 09:18 AM

There is currently an issue with the batch-calculation structure not being normalized correctly, update will be up in a few minutes.

kevinjet · 07-24-2012, 09:27 AM

The hands like
BU 3.0 Ra
SB 13.6 C,X,X,X
BB 13.4 C,X,X,X
is now shown as error for SB, and will also show error for BB in the same config.
It seems if more than one player are non-allin at showdown -> this hand should not be analyzed. The detection code for each player I've already posted

Code:

HM1:: playerhandstourneykeycolumns.streetwentallin >0
PT3/4:: tourney_hand_player_statistics.enum_allin # 'N'

plexiq · 07-24-2012, 09:35 AM

I do the detection within the program when batch calculating, since i already have access to all stacks etc at that point. Currently the requirement is that the opening action is a raise.

I am not sure if it makes sense to exclude hands like these. Is SB really folding to a BB shove here? I think treating it as all-in is not too bad overall. If pushing is not profitable here, then the spot likely deserves a closer look anyway.

kevinjet · 07-24-2012, 09:46 AM

Quote:

Originally Posted by plexiq

I am not sure if it makes sense to exclude hands like these. Is SB really folding to a BB shove here? I think treating it as all-in is not too bad overall. If pushing is not profitable here, then the spot likely deserves a closer look anyway.

The best play here is collusion(as shown), but I had some funny hands when BB shoved at river, or even at turn. Then, best play for SB is fold.
Treating it as all-in would really be bad. Much better to leave it unanalyzed.
So, user will see, why and see it was play C,X,X,X - and understand it was collusion - good play.
Right now the actions are shown only as
BU R
SB C
BB C
And hand is marked as error. And I have to individually analyze it, look to handhistory and found thats ok. Its just a waste of time.

Thats why I asked for full action strings(up to showdown), and to mark hand as batch-unanalyzeable.

plexiq · 07-24-2012, 09:57 AM

Agree about showing postflop actions somewhere.

About excluding or not:
There are plenty of spots where treating hands as all-in will be perfectly fine.

I guess that's up to personal preference, i think calculating these hands and manually discarding them is better than missing them. But probably best to add a checkbox "ignore all non-allin hands" in the batch dialog, easy enough to do.

Ignoring all will be icky sometimes though, since ppl do stuff like:
BU 10BB
SB 8BB
BB 9BB

BU raise 7BB

Obviously he is not folding here, ever. And plenty of regs do these random commiting raises instead of actually shoving.

kevinjet · 07-24-2012, 10:48 AM

Quote:

Originally Posted by plexiq

But it is a nice idea to update the range + quickgraph window when navigating through the list, this can easily be done by simply caching the hero range & eqdiffs.

Yet, there's also a need to cache hand outline, and show it instead of query outline.
It seems outline windows should be finally unified. It seems only differences at query outline would be
for all hands:
- ticks column is unclickable (grayed).
for non-analyzed hands:
- EQs, range, "fit in range?" columns are empty.

So, the list of gathered columns for desired unified outline window:
tick, pos, name, stack, BB, action, cards, "fit in range?", range.

kevinjet · 07-24-2012, 11:02 AM

Quote:

Originally Posted by plexiq

Ignoring all will be icky sometimes though, since ppl do stuff like:
BU 10BB
SB 8BB
BB 9BB

BU raise 7BB

Obviously he is not folding here, ever. And plenty of regs do these random commiting raises instead of actually shoving.

a) If anybody folds, there won't be showdown. Hand will not be rejected with my rule.
b) If somebody calls, so usually caller would went all-in to showdown -> so hand will be analyzed.
c) If somebody calls, and all further actions will be just checks, -> two players will be not all-in, and thats interesting hand, and it should be examined closer.

I guess these hands like c), and mine should be marked as "attention, not sure about correct analysis" in some additional way. Like "?" icon next to result icon...
So your hand at case c) would have "?" mark only. And it would be false attention.
And my collusion hand would have both marks: "X ?". And it would be correct attention.

kevinjet · 07-24-2012, 02:04 PM

Just found folded hand which marked as error, but linear analysis got EQ=-0.418, unrestricted got EQ=-0.438. For the hand, payouts 50/50.

Code:

PokerStars Hand #999999: Tournament #99999, $3.67+$0.07 USD Hold'em No Limit - Level II (50/100) - 2012/01/01 21:54:54 MSK [2012/07/22 13:54:54 ET]
Table '99999 1' 6-max Seat #1 is the button
Seat 1: Villain1 (340 in chips) 
Seat 2: Hero (650 in chips) 
Seat 3: Villain2 (980 in chips) 
Seat 4: Villain3 (1030 in chips) 
Villain1: posts the ante 20
Hero: posts the ante 20
Villain2: posts the ante 20
Villain3: posts the ante 20
Hero: posts small blind 50
Villain2: posts big blind 100
*** HOLE CARDS ***
Dealt to Hero [7c Kh]
Villain3: folds 
Villain1: raises 220 to 320 and is all-in
Hero: folds 
Villain2: folds 
Uncalled bet (220) returned to Villain1
Villain1 collected 330 from pot
*** SUMMARY ***
Total pot 330 | Rake 0 
Seat 1: Villain1 (button) collected (330)
Seat 2: Hero (small blind) folded before Flop
Seat 3: Villain2 (big blind) folded before Flop
Seat 4: Villain3 folded before Flop (didn't bet)

So I'm in worry - it means fast analysis are way too rough. Difference between linear and unrestricted was never so huge. At other side, there's possibility fast analysis could mark errors as good hand. This error could be also up to 0.5% (maybe more?)

So, a tuneable threshold should be specified. If action's EQdiff>threshold, that's ok: it was extremely good action for sure. no need to more precise analysis(for ex, AA call, 23o fold against 2 allin). Otherwise, perform next-level analysis.

It makes sense on both transitions (fast->linear, and linear -> unrestricted), but with different thresholds.

As a default, it would be set to +0.5% for switch fast->linear.
And to 0% for linear -> unrestricted. Those who really learns to fly, could set it above zero.

As a best, I see it as radiobutton set "fast/linear/unrestricted", and two inputboxes near it. 1st threshold inputbox near "fast" radiobutton , 2nd threshold near "linear" radiobutton.

kevinjet · 07-24-2012, 03:07 PM

Quote:

Originally Posted by plexiq

Currently the requirement is that the opening action is a raise.

You miss opening call all-in, when opener has not more than bb+ante. I see some examples already. These hands should be analyzed as usual.
However, as it was discussed, it could be "?" mark, based on other's actions.

kevinjet · 07-24-2012, 03:13 PM

Bugreport:
Select some tourneys, handlist below will contain all the hands of all selected tourneys. Then select one tourney within selection. Handlist below will not change. But it should be replaced for hands of selected tourney only (as it does, when selecting a tourney outside of selection)

kevinjet · 07-24-2012, 03:38 PM

This hand has gray "X" mark for unknown reason. It seems to me as ordinary hand, that should be batch-analyzed.

Code:

PokerStars Hand #7777: Tournament #99999, $3.73+$0.07 USD Hold'em No Limit - Level III (100/200) - 2012/01/01 18:11:07 MSK [2012/01/01 10:11:07 ET]
Table '99999 1' 6-max Seat #3 is the button
Seat 2: Hero (2080 in chips) 
Seat 3: Villain1 (860 in chips) 
Seat 5: Villain2 (60 in chips) 
Hero: posts the ante 40
Villain1: posts the ante 40
Villain2: posts the ante 40
Villain2: posts small blind 20 and is all-in
Hero: posts big blind 200
*** HOLE CARDS ***
Dealt to Hero [Qc 9h]
Villain1: raises 620 to 820 and is all-in
Hero: folds 
Uncalled bet (620) returned to Villain1
*** FLOP *** [4d 5d 4c]
*** TURN *** [4d 5d 4c] [8h]
*** RIVER *** [4d 5d 4c 8h] [Ad]
*** SHOW DOWN ***
Villain1: shows [3d Ks] (a pair of Fours)
Villain1 collected 360 from side pot
Villain2: shows [2d As] (two pair, Aces and Fours)
Villain2 collected 180 from main pot
*** SUMMARY ***
Total pot 540 Main pot 180. Side pot 360. | Rake 0 
Board [4d 5d 4c 8h Ad]
Seat 2: Hero (big blind) folded before Flop
Seat 3: Villain1 (button) showed [3d Ks] and won (360) with a pair of Fours
Seat 5: Villain2 (small blind) showed [2d As] and won (180) with two pair, Aces and Fours

kevinjet · 07-24-2012, 04:40 PM

And vice versa, that's limper hand, but it was analyzed for unknown reason..

Code:

PokerStars Hand #6666: Tournament #99999, $3.67+$0.07 USD Hold'em No Limit - Level II (50/100) - 2012/01/01 2:39:52 MSK [2012/01/01 18:39:52 ET]
Table '99999 1' 6-max Seat #2 is the button
Seat 1: Villain1 (699 in chips) 
Seat 2: Hero (946 in chips) 
Seat 3: Villain2 (105 in chips) 
Seat 6: Villain3 (1250 in chips) 
Villain1: posts the ante 20
Hero: posts the ante 20
Villain2: posts the ante 20
Villain3: posts the ante 20
Villain2: posts small blind 50
Villain3: posts big blind 100
*** HOLE CARDS ***
Dealt to Hero [2h Qh]
Villain1: folds 
Hero: calls 100
Villain2: folds 
Villain3: checks 
*** FLOP *** [Th Td Ac]
Villain3: checks 
Hero: checks 
*** TURN *** [Th Td Ac] [4s]
Villain3: bets 100
Hero: folds 
Uncalled bet (100) returned to Villain3
Villain3 collected 330 from pot
*** SUMMARY ***
Total pot 330 | Rake 0 
Board [Th Td Ac 4s]
Seat 1: Villain1 folded before Flop (didn't bet)
Seat 2: Hero (button) folded on the Turn
Seat 3: Villain2 (small blind) folded before Flop
Seat 6: Villain3 (big blind) collected (330)

plexiq · 07-24-2012, 04:56 PM

Thanks for the reports, will check them.

Regarding the -0.4x% Hand, maybe the Action is not read correctly or something. The difference definitely should not be that big. (As mentioned, the batch feature is not very well tested atm. Definitely possible that there are still bugs.)

Just to make sure, did you get the update with the payout-normalization fix a few hours ago?

kevinjet · 07-24-2012, 05:09 PM

Actions are good - F at handlist, at outline: F R F F.
When I set "Error EQdiff" to -0.43% -> warning mark, to -0.42% -> error mark.
Definitely some internal marking bug.

I've updated instantly at that time. Just checked again - I'm using latest version.

PS funny numbering issues:
1) enter "Error EQDiff" as -0.01, run batch, and then enter batch dialogue again. This number will be weirded, but still correctly treated...
2) entering number not in range [-1..1], this change is silently ignored, and calculation goes with old value.