K-Means Clustering
Posted by Frankie Carson
Video Player is loading.
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Resume Video
Start from Beginning
Watch Video
10
Loading 10 Comments...
Frankie,
Very interesting video. Lots to take away from this one.
Thanks
Interesting way to look at data. I have never heard of clustering.
The differences in strategy based on the gaps on the flop was interesting. Just a subtle shift in the gap makes a fair bit of difference.
Seems like the closer or nonexistent the gap the less rng and nut advantage we have so we should size down and bet less often.
I often equate a small bet with a larger frequency but that is not always the case.
I also associate large bets with lower frequency as having a nut advantage but not a range advantage.
Do these seem like reasonable heuristics?
Thanks Frankie!
Hey Sound.
Yes that's fairly sound (no pun intended ha).
Really the key is to understand why we have these shifts for seemingly subtle movements in the flop. Most the time it's a silly thing like if V plays 1 particle offsuite combo that would change sizing from x to y. But the juice of knowledge is knowing why the change is occurring. Because with this you can unlock a treasure trove of heuristics in game when thinking about ranges and interactions of board textures.
hey Frankie wicked video here my guy thank you
Thanks True :)
Hey Frankie, that's a very interesting way to cluster together different board textures! It seems like the algorithm did a pretty good job in grouping together the boards. Were there any big outliers, i.e. boards that didn't seem to fit in their group at all? Or just a bunch of borderline boards that can go into multiple categories? Did you do the K means clustering in Excel? Is this an in-built feature of Excel?
Hey Matlittle.
I thought it generally did a good job. If I did less clusters it would likely start forcing more boards together that shouldn't really be.
I think what is really interesting, since the algo is agnostic to poker cards and only focuses on math, is the grouping of boards like AQ7 and T22 in same bucket. Seems like a mistake, but when you look over the boards further makes a lot of sense. OP has tons of air on both and shouldn't carry much 2x (different in lower rake and/or lower OR sizes) while IP nut region is similar from a mathematical lens as AA, KK, etc garner almost as much EV as AQ/A7 does.
Yes I did but unfortunately it does cost money. No free built in functions that I'm aware of.
These 2 boards were presumably both cbetting mostly for the overbet sizing?
They are actually both splitting range small/overbet since there is a sizable range advantage but also a nut advantage.
Really great video.
would love to see the analysis on Blind vs Blind /3bet pot using same kind of approach.
Be the first to add a comment
You must upgrade your account to leave a comment.