I really ought not to play on the second Tuesday of the month (ok, yesterday was the third Tuesday of the month, but it was my second Tuesday, after returning from LV).
I absolutely dropped my bollocks, to the extent that I got as far as opening up Pokertracker to start doing some research. Then I thought "fuck, I'm tired, and I have no idea where to start. I'm going to bed".
Clearly either I have got worse or my opponents have got better, and there are serious leaks somewhere. I suspect that at UB (where my results have got worse month by month since December, culminating in a minus $400 for April) I've shifted from +0.7BB an hour to something like minus 0.5BB. So it's a matter of finding one-point-five BB an hour somewhere to get that back.
At the moment, it looks to me to be top-pair-top-kicker territory. Opponents simply aren't semi-bluff raising me enough on the turn. The result is that my three-bet gets either picked off as a four-bet (do I cry-call to the end, do I fold?) or flat-called and then I call the river, and lose.
The problem with this is that it's counter-intuitive, because the call only has to be right about a quarter of the time, so you might feel the call is a losing play when, in fact, it isn't.
That's part of my general feel that opponents are getting out very early if they can't beat TPTK, and are flat-calling the flop and raising the turn if they can.
I can think of various possible counter-plays to this, but I can't really think which one is best.
I can also think of several metagame solutions (play something else, play lower levels, spend more time finding good games, good seats, don't play at all, just play at weekends, focus on NL, focus on MTTs, etc).
I'll really have to do a diligent trawl of a whole host of hands/scenarios. I'm fairly sure I'm doing something seriously wrong on either the flop or the turn (the numbers for the river and pre-flop don't look so bad) which is turning a winning scenario into a losing one. All I have to do is find out what the flaws are, and that is a long slow job.
+++++++
Mr Bowles responded in a PM about my previous piece, stating that it reminded him of two articles by Thomas and Chris Ferguson. The algebra was a little bit beyond me, but I understood the conclusion. In fact the article is covered in very similar ground by Chen & Ankemann in The Mathematics of Poker
Where I depart from this line of thought is that Ferguson and Ferguson "solve" the problem with optimal play on both sides. As was observed in another poker book, if you are playing against a player who is playing optimally, you should find another opponent.
Andy Ward has adapted his own style in tournaments on just these empirical grounds. Opponents do not tend to play optimally, so your all-in pushes have to be adapted in strength to adjust for an increased likelihood of folding. Note that for Andy's numbers, he doesn't need to adjust his hands for an increased likelihood of calling, but he could adjust the M factor upwards. Hmm, I hadn't thought of that before. That's a point worth following up.
So, I am not attempting to solve heads-up poker betting games, but to look for a system that provides the best play, no matter how sub-optimal your opponents' play is
As the Fergusons observed, in their game the "range" of bets goes; bet your worst hands, then check-fold, then check-call, then bet your best hands.
However, if you are up against an opponent who folds very rarely, the range of "bet your worst hands" shrinks. Against an opponent who folds often, this range increases.
Chen & Ankemann don't like this fuzzy line of thought. They say that opponents are dynamic. That they learn.
Indeed they do, and part of the skill in poker is staying one step ahead of them. If the opponent learns enough to play optimally, then you go somewhere else (a 'metagame' concept not touched upon by Chen or the Fergusons, because it is outside their remit).
What I would like to get is some system that distills a player's PT stats into an empirical likelihood. That gives a decision on whether to bet or not.
I know that this is straying somewhat into bot territory, which isn't what I'm aiming for. I want to create a system that can be applied aby a player, in real time, with some accuracy, that doesn't descend into the vagaries of "usually", "mainly", "nearly always" or, worst, inaccurate uses of the terms "never" and "invariably".
PJ
I absolutely dropped my bollocks, to the extent that I got as far as opening up Pokertracker to start doing some research. Then I thought "fuck, I'm tired, and I have no idea where to start. I'm going to bed".
Clearly either I have got worse or my opponents have got better, and there are serious leaks somewhere. I suspect that at UB (where my results have got worse month by month since December, culminating in a minus $400 for April) I've shifted from +0.7BB an hour to something like minus 0.5BB. So it's a matter of finding one-point-five BB an hour somewhere to get that back.
At the moment, it looks to me to be top-pair-top-kicker territory. Opponents simply aren't semi-bluff raising me enough on the turn. The result is that my three-bet gets either picked off as a four-bet (do I cry-call to the end, do I fold?) or flat-called and then I call the river, and lose.
The problem with this is that it's counter-intuitive, because the call only has to be right about a quarter of the time, so you might feel the call is a losing play when, in fact, it isn't.
That's part of my general feel that opponents are getting out very early if they can't beat TPTK, and are flat-calling the flop and raising the turn if they can.
I can think of various possible counter-plays to this, but I can't really think which one is best.
I can also think of several metagame solutions (play something else, play lower levels, spend more time finding good games, good seats, don't play at all, just play at weekends, focus on NL, focus on MTTs, etc).
I'll really have to do a diligent trawl of a whole host of hands/scenarios. I'm fairly sure I'm doing something seriously wrong on either the flop or the turn (the numbers for the river and pre-flop don't look so bad) which is turning a winning scenario into a losing one. All I have to do is find out what the flaws are, and that is a long slow job.
+++++++
Mr Bowles responded in a PM about my previous piece, stating that it reminded him of two articles by Thomas and Chris Ferguson. The algebra was a little bit beyond me, but I understood the conclusion. In fact the article is covered in very similar ground by Chen & Ankemann in The Mathematics of Poker
Where I depart from this line of thought is that Ferguson and Ferguson "solve" the problem with optimal play on both sides. As was observed in another poker book, if you are playing against a player who is playing optimally, you should find another opponent.
Andy Ward has adapted his own style in tournaments on just these empirical grounds. Opponents do not tend to play optimally, so your all-in pushes have to be adapted in strength to adjust for an increased likelihood of folding. Note that for Andy's numbers, he doesn't need to adjust his hands for an increased likelihood of calling, but he could adjust the M factor upwards. Hmm, I hadn't thought of that before. That's a point worth following up.
So, I am not attempting to solve heads-up poker betting games, but to look for a system that provides the best play, no matter how sub-optimal your opponents' play is
As the Fergusons observed, in their game the "range" of bets goes; bet your worst hands, then check-fold, then check-call, then bet your best hands.
However, if you are up against an opponent who folds very rarely, the range of "bet your worst hands" shrinks. Against an opponent who folds often, this range increases.
Chen & Ankemann don't like this fuzzy line of thought. They say that opponents are dynamic. That they learn.
Indeed they do, and part of the skill in poker is staying one step ahead of them. If the opponent learns enough to play optimally, then you go somewhere else (a 'metagame' concept not touched upon by Chen or the Fergusons, because it is outside their remit).
What I would like to get is some system that distills a player's PT stats into an empirical likelihood. That gives a decision on whether to bet or not.
I know that this is straying somewhat into bot territory, which isn't what I'm aiming for. I want to create a system that can be applied aby a player, in real time, with some accuracy, that doesn't descend into the vagaries of "usually", "mainly", "nearly always" or, worst, inaccurate uses of the terms "never" and "invariably".
PJ