Need 3bet vs Steal?

[m4ximu5]

Is there a stat for 3bet vs Steal (incl. SB, BU and CO) or have to built i them?

[kraada]

That stat would be Raise vs Steal Att, which is available by default.

[m4ximu5]

Aha thank you, i overlooked it.
It counts only for blinds, not for 3bet BU vs Steal from CO?

[WhiteRider]

That's correct - we can't currently tell where the initial raise came from when the player is on the button, and they are not recorded as facing a steal since they haven't posted a blind. We're investigating a mechanism to allow us to access this kind of information in the future, but I don't know when it will be available yet.

[eroock]

A related question: I want to calculate the stealer's 3-bet calling range. Something like [ATS * Call 3bet (when attempted to steal)]. How would that look like as a stat?

[kraada]

You'd need to build the Call PF 3Bet After Stealing columns yourself then you could do the multiplication in the value expression of the stat you want.

One word of caution before putting more effort in here - you'd need a large sample on any given player for this sort of thing to be useful. If you're playing 6 max, he can steal half of the hands when he's folded to - so say 30% of hands overall. Then if he steals 50% of the time (which is high), you're at 15% of hands where he steals. If his opponents 3bet fairly liberally at 15% you're now at 2.25% of hands. So about 1 in 50 hands you get a data point. If you want 10 data points, you'd need 500 hands at least. If you want a standard sample of 30, you'll need 1500 hands. Also, if he steals less often you get samples less often. If he gets 3bet less, you get samples less often; if you play full ring, you get samples less often.

There's certainly some value to doing analysis on large bodies of data here, but it's going to be tough to get much of a read over anything less than 500 hands at a bare minimum.

You just might want to take that into account before putting a bunch of work into building this stat.

[eroock]

Thanks for the heads-up. Don't really have that many hands on my opponents to make it work.
Going with your explanation, would you say that for a small sample size the PFR of a late position opener is more accurate for his hand range than his ATS?

[kraada]

You'll always have more chances to PFR than you will to ATS - you steal only when everyone folds to you, but you can raise regardless of what they do in front. So PFR should converge to meaningful numbers faster - but it may not necessarily be the number you want, as it includes the hands where he was facing a raise or a 3bet, etc - it depends on the situation you're in.

Basically, the more data you get the better. Mostly I personally try really hard not to pay attention to any statistic with a sample of less than 10, and I prefer samples of 30, 50 or higher - obviously you can't always do that, so when you can't you go with the best data you have and fall back on playing normal poker more heavily when your data is inconclusive.

[Mike_]

Fold to 3bet when raised is (cnt_p_3bet_def_action_fold_when_open_raised / cnt_p_3bet_def_opp_when_open_raised) * 100

This shows Fold to Preflop 3Bet Percentage When Hero Open Raised And It Wasn't A Steal


How does it know it wasnt a steal attempt >

[kraada]

It's in the definition of the columns in question - not(holdem_hand_player_statistics.flg_steal_att) - you can see the column definitions on the Columns tab.