Poker-AI.org
http://poker-ai.org/phpbb/

djhemlig LUT - Wrong Turn EHS
http://poker-ai.org/phpbb/viewtopic.php?f=24&t=2948
Page 1 of 1

Author:  ocara [ Wed Oct 14, 2015 3:42 pm ]
Post subject:  djhemlig LUT - Wrong Turn EHS

It is known that SteveBrecher - HandEval had a bug on the TURN calculation that was fixed later. The correct version can be found on his website.

The problem is that some other codes have used the old/bugged version of SteveBrecher and never fixed that. I was using djhemlig LUT, but only now I realised it is calculating the wrong EHS on the TURN. Really frustrating. :shock: :?

I'm sure that's why my bot is not making money!!! yet. ;)


I'm trying to build a workaround on djhemlig code to use the updated version of SteveBrecher and re-generate the LUTs.

Author:  ocara [ Thu Oct 15, 2015 9:24 am ]
Post subject:  Re: djhemlig LUT - Wrong Turn EHS

By now the best thing I found out is to use Steve Brecher 2008Apr27.0.

Code:
// 2008Apr27.0
//      fix hand6Eval's calls to flushAndOrStraight6


djhemlig was using 2006Dec05.0. You just need to replace HandEval.java.


The current version 2010Jun25.1 does not fit into djhemlig code. So we are missing all these updates and I have no idea how they will impact the evaluation:

Code:
// 2010Jun25.1
//      Fix handRazzEval return for full house for AAAABBB rank pattern (make the
//         lower rank the trips rank and the other the pair rank rather than vice-versa).
// 2010Jun25.0
//      Fix hand7Eval return value for quads+trips (case 2 of its switch).
//      Fix kicker ordering for handRazzEval one pair for AABBCCD rank pattern.
// 2010Jun24.0
//      Fix one pair value returned by handAto5LoEval.
// 2010Jun23.1
//      In Omaha8LowEval add short circuits to avoid evaluation if board has fewer than 3 8OB cards
//         and for pairs of hole cards which are not both 8OB.
// 2010Jun23.0
//      Fix Omaha8LowEval to use previous algorithm, but separately on each pair of hole cards.
// 2010Jun22.2
//      Fix coding error in Omaha8LowEval that would return incorrect result when
//         third and fourth hole cards make the best 8OB low.
// 2010Jun22.1
//      Omaha8LowEval rewritten with different interface.  Previous algorithmic trick
//         to avoid brute force of six (4C2) computations didn't work.
//      lo3_8OBMask array omitted -- no longer used.
// 2010Jun22.0
//      Fix initialization of lo3_8OBMask and loMaskOrNo8Low.
// 2010Jun21.4
//      Change name of method from hand5Ato5LoEval to handAto5LoEval.
// 2010Jun21.3
//      Fix return value of hand8LowEval.
//      Fix bit-wise rotation of parameter fields in low eval methods: "^"s to "&"s.
// 2010Jun21.2
//      Fix initialization of straightValue[0x100F], i.e., wheel.
//      Fix return values for A5432 and A5432 flush in hand2to7LoEval.
// 2010Jun21.1
//      Fix loop initialization in initializer block to fix eval result rank values.
// 2010Jun21.0
//      Fix deconstruction of most eval method parameters into suit fields.
// 2010Jun20.1
//      Fix return value of hand2to7LoEval for an A5432 flush.  This method was
//      introduced in 2010.Jun20.0.
// 2010Jun20.0
//      Conform method parameters and results to C version: CHANGED API and fewer arrays.
// 2010Jun19.0
//      fix rotations of ace bits from bit 13 to bit 0 in
//      handRazzEval, hand5Ato5LoEval, ranksMaskLo, hand8LowEval

Page 1 of 1 All times are UTC
Powered by phpBB® Forum Software © phpBB Group
http://www.phpbb.com/