500th post thesis - the cost of marginal preflop decisions

[rzk]

[ QUOTE ]

can we agree that the tighter the range, the less marginal hands? That is, there are more hands I'm uncertain about for the button than for utg.

So if somebody has open-raised and it's to me otb, I think there will be considerably fewer marginal hands than if it had been folded to me otb.

Anyway, I agree that overall the effect of smaller sample sizes per situation will outweight the effect of smaller ranges per situation.

[/ QUOTE ]

i definitely agree that if somebody open-raises utg you have fewer marginal spots than if everybody folds to you, because there are no marginal Qxo and Jxo type hands anymore, whereas if everybody folded there would even be marginal Txo,9xo, and 8xo otb. basically, in step 2 i made a rough estimate for the number of marginal spots by assuming that on average there are 4 unsuited marginal hands per position: Axo - Jxo. but it's true, sometimes there are less and sometimes there are more, so a more careful counting is needed to get a better estimate.

[jr4284]

nice work, rzk.

While I appreciate the ideas of, as sample sizes approach infinity, refining starting hand ranges to increase overall EV by cutting the fat, this seems to apply much more liberally to the LAGs than the TAGs out there, as discussed already.

Unfortunately(?) I am in the TAG group, so can we make an argument for increasing ranges? Say once the sample size reaches 1M, sort by open raising results, and all hands that have an EV ~ .02 BB we expand towards the lower value hand? I.e. if our HJ KTo EV is .03, and we never play K9o from HJ, add it in.

Maybe some of us are missing the .34-1BB/100 on the other end.

[rzk]

[ QUOTE ]
Maybe some of us are missing the .34-1BB/100 on the other end.

[/ QUOTE ]

you are absolutely right. the final conclusion of my calculation applies equally well to a person who doesn't play enough hands as to a person who plays too many. the two types of mistakes in a marginal spot are equally costly.

the difference between a tag and a lagtag here is that it is much harder for the tag to extract the correct strategy from his data. the lagtag can simply "cut the fat", as you say. the tag would have to know (from elsewhere) the difference in ev between nearby hands. for example, if he knows that ev(KTo)>ev(K9o) by 0.08bb/hand, then if his winrate with KTo is 0.085bb/hand then he knows K9o is a marginal spot for him and he should tentatively add it to his range. without this knowledge he wouldn't even know what his marginal spots are.