We'll talk about some ways to make the new bot better here.
03-16-2014 11:47 PM
#1
| |
beta testing new botWe'll talk about some ways to make the new bot better here. | |
03-17-2014 04:31 PM
#2
| |
|
I only had time to play a few quick hands, so this is based on a very small sampling. Ninja seems better than WAZZUP, but still makes some bad/weird plays. |
03-17-2014 06:42 PM
#3
| |
Thanks, Phi! | |
03-18-2014 03:46 PM
#4
| |
|
Played a few more hands just now. New bot is definitely an improvement over WAZZUP, but is still not very good. |
03-18-2014 04:17 PM
#5
| |
I don't like the way he set A down low utg here in hand 2 on table 7837. The ace should almost aways be set front or mid. We don't set the ace in back often unless it is part of a straight or flush. | |
03-18-2014 04:19 PM
#6
| |
Phi, the bot is live now. There is a BetaBot button for him next to the Waz buttons. | |
03-18-2014 04:39 PM
#7
| |
| |
03-18-2014 04:56 PM
#8
| |
03-18-2014 05:02 PM
#9
| |
|
Should the bot have played this differently? |
Last edited by Keith; 03-18-2014 at 05:05 PM. | |
03-18-2014 05:12 PM
#10
| |
Yeah, he could have set 88 front for the 3 royalty points and 4455 mid here with the straight - that way he gets 2+3 royalties with the straight set. | |
03-18-2014 06:56 PM
#11
| |
| |
03-18-2014 07:45 PM
#12
| |
|
Just played another session, and there are some hands that reveal some pretty big holes in the bot's strategy. |
03-19-2014 02:56 AM
#13
| |
That's a good way of putting it. In general, if a FL card is not being used for a flush or a straight then it should not go low (in other words it should not go in back). Of course we can put aces or even kings mid if we're drawing to queens in front such that we beat them with a single pair. Queens, Kings and Aces generally don't go in back unless they are part of a straight, flush or full house draw. | |
03-19-2014 05:39 AM
#14
| |
|
Should the bot have played this differently? |
03-19-2014 05:44 AM
#15
| |
|
why discard 2h on 7th when he could put it on the bottom and discard the 5 |
Last edited by Keith; 03-19-2014 at 05:48 AM. | |
03-19-2014 10:57 AM
#16
| |
|
A new version was put up last night, supposedly plays a little better. Thank you guys for your comments - everything is duly noted. |
03-19-2014 04:14 PM
#17
| |
A few hands I played | |
Last edited by givememyleg; 03-19-2014 at 04:24 PM. | |
03-19-2014 04:26 PM
#18
| |
Alex - if you have any questions as to why I would play some of these hands differently, or place a card in a certain spot, let me know and I can go into more detail than just saying "I would have put this up top" etc. to help give you a better understanding. | |
03-19-2014 04:50 PM
#19
| |
|
Carl, thanks for your thoughts. |
03-19-2014 10:23 PM
#20
| |
The bot went Q / 9 / 383 but Q / 8 / 393 is better. | |
03-19-2014 11:10 PM
#21
| |
#5 | |
03-20-2014 11:39 AM
#22
| |
|
Yep, found a bug. Will have to call on the Wizard with this one. I'll let you know when it's fixed. |
03-20-2014 04:42 PM
#23
| |
03-20-2014 11:07 PM
#24
| |
|
OK, a new version is out there. Let me know what you guys think. |
03-21-2014 12:44 PM
#25
| |
The 5 card set is clearly wrong. He should go Q / X / 22TT instead of X / Q / 22TT. In other words, the queen goes front instead of mid. | |
03-21-2014 01:03 PM
#26
| |
|
Hmm, we had it working just fine. I'll check again. Thanks, Eric. |
03-21-2014 02:36 PM
#27
| |
I think his second draw was wrong - he should have a pair of fours mid. | |
03-21-2014 02:41 PM
#28
| |
His first draw is wrong - the king goes front instead of mid. | |
03-21-2014 02:48 PM
#29
| |
ShHere is another example of a five card set issue, the queen goes front and not mid. | |
03-21-2014 02:51 PM
#30
| |
The ace of clubs on the first draw goes front instead of mid. | |
03-21-2014 03:41 PM
#31
| |
|
OK, more of the same - I'll check into it tonight. |
03-21-2014 04:56 PM
#32
| |
The 2 should not go in front with his five card set. | |
03-21-2014 08:30 PM
#33
| |
| |
03-22-2014 11:51 AM
#34
| |
The 6 on the 2nd draw should go mid instead of front. | |
03-22-2014 11:56 AM
#35
| |
This 5 card set of | |
03-22-2014 11:59 AM
#36
| |
|
Agreed. I also dislike the set. I'd argue that |
03-22-2014 12:00 PM
#37
| |
I thought the bot was adjusted | |
03-22-2014 01:01 PM
#38
| |
The 6 on his first draw should go mid instead of front. | |
03-22-2014 02:46 PM
#39
| |
Queen on first draw goes front instead of mid. | |
03-22-2014 03:26 PM
#40
| |
He should have set the Q in the front instead of the back. | |
03-22-2014 04:04 PM
#41
| |
The bot made sure he would foul with the way he played his third draw. Setting four to a flush mid with just a straight in back, he guaranteed himself a foul. | |
03-22-2014 05:16 PM
#42
| |
|
New version is out - seems rid of "stupid tricks". Let me know how it does. |
03-22-2014 08:14 PM
#43
| |
|
had this error message , bot was in fantasyland |
03-22-2014 08:35 PM
#44
| |
|
Thanks, Keith, forwarded on. |
03-22-2014 10:08 PM
#45
| |
The bot's set isn't good: | |
03-22-2014 10:13 PM
#46
| |
I didn't see all the discards but I'm guessing the bot could have played the mid better. | |
03-23-2014 08:14 AM
#47
| |
It is probably simplest to program this bot to never set trips in front unless he's in fl. | |
03-23-2014 08:21 AM
#48
| |
This A / K 3 / 2 T set is terrible. | |
03-23-2014 08:44 AM
#49
| |
The bot made big mistakes on the third draw. | |
03-23-2014 10:46 AM
#50
| |
|
Thanks, Eric. I'll go through the placing logic once more. What is your overall impression of the current logic, is it getting closer to where you want it to be? |
03-23-2014 10:55 AM
#51
| |
Yes, it is getting better. | |
03-23-2014 11:26 AM
#52
| |
Bad set - the ace and king should be switched. | |
03-23-2014 11:34 AM
#53
| |
Bad set - Q / 5 / 229 is better. | |
03-23-2014 11:38 AM
#54
| |
The bot eliminated himself from fl on just the first draw by setting it wrong. | |
03-23-2014 11:44 AM
#55
| |
Bad set - the 3 and 9 should be switched. | |
03-23-2014 03:17 PM
#56
| |
|
The pattern that keeps emerging is the tendency to set a small pair with a small kicker in the bottom and two connected larger number cards in the mid. This is often catastrophic because unless you make at least trips in the bottom, you can't make 2 middle pairs in the mid, and therefore can't make FL. The program must be telling the bot to place any pair in the bottom and to value setting connected cards in the mid. The combination of those instructions must be what leads to sets like this. Instead, the algorithm should be something like: |
03-23-2014 03:21 PM
#57
| |
|
Eric, the case in the 6:44 post is getting into a really murky territory. In the absence of a FL card, with x/89o/322 it set a pair on the back with the first trey that has been played thus maximizing its chances of getting fh or better, while drawing to an open-ended straight in the middle. Switching 9 and 3 there would cut down on EV significantly in this spot. While I am not arguing the reasoning, programming it to set it as you suggested would be, in fact, programming against its current objectives. |
03-23-2014 05:59 PM
#58
| |
03-23-2014 07:13 PM
#59
| |
|
I must have been writing my response as Phi was writing his. Good thoughts, Phi. |
03-23-2014 09:55 PM
#60
| |
|
Got this error when both hit FL. Clicked ok and it dealt the hands. |
03-24-2014 12:03 AM
#61
| |
I think it's catastrophic but we can keep playing against the bot to confirm my thoughts. | |
Last edited by Eric; 03-24-2014 at 12:08 AM. | |
03-24-2014 09:35 AM
#62
| |
|
I've been monitoring that kicker on the 5 card set situation and it looks like it does put it in awkward situations in later rounds quite a bit. I will try to fix it. |
03-24-2014 01:05 PM
#63
| |
|
If the chances of making trips or better are 41%, that means the chances of NOT making trips or better are 59%. In those hands, you have a very hard time making a 2-pair hand in the back big enough to beat 9s-up or 8s-up in the mid. Putting yourself in a position on the set that will lead to fouling or making no royalties half the time or so just has to be a bad plan. |
Last edited by OneByPhi; 03-24-2014 at 05:16 PM. | |
03-24-2014 01:23 PM
#64
| |
|
To add a little more: |
03-24-2014 01:32 PM
#65
| |
|
Definitely, Phi - I am convinced. Looking through the hands the bot plays it is obvious that the higher kicker has to be in B in that spot. Thanks a ton. |
03-24-2014 01:58 PM
#66
| |
|
Well, like somebody said once "there is not a dead horse that I would not just keep on beating"... The odds of trips on a 223 set is 41%, runner-runner FH 22333, 13%, and 22223 quads 8% - for a grand total of 62% of having higher than two pair. |
03-24-2014 02:18 PM
#67
| |
|
the other 38% though are you making it a lot more likely to then foul since you are less likely to be able to beat your mid and your mid limits your chance to get to fantasyland without fouling. |
03-24-2014 02:28 PM
#68
| |
|
Should the bot have played this differently? |
Last edited by Keith; 03-24-2014 at 02:33 PM. | |
03-24-2014 03:18 PM
#69
| |
| |
03-24-2014 06:57 PM
#70
| |
|
Sorry for misunderstanding your math claim. |
03-24-2014 07:44 PM
#71
| |
|
Should the bot have played this differently? |
03-24-2014 10:05 PM
#72
| |
| |
03-25-2014 03:45 PM
#73
| |
|
I just played a 25-hand set with the new bot and it seems to be playing a little better. It's still making some pretty bad sets, however. Here are some problematic hands: |
Last edited by OneByPhi; 03-25-2014 at 04:34 PM. | |
03-25-2014 04:03 PM
#74
| |
I agree with Phi. There is a pattern of bad sets involving 2 suited cards mid. | |
03-25-2014 06:13 PM
#75
| |
The bot makes a big mistake on the 3rd draw here. | |