|
|
Skill ICM
Ok guys here is version 1.0.
Disclaimer: Using this model requires making judgements regarding your relative skill level in comparison to that of your opponents. Since this is a highly subjective practise, any losses or dip in form as a result of using this model is enitrely you own fault and not the fault of the writer.
In Poker everything you do is a decision based on imperfect information. Rating a player is no different.
There is no objectively correct way to do this, however as players we learn to become skilled in reading players so the idea of rating them should certainly not be abhorrent. If anyone wants an argument about this im there.
So First thing you need to do is rate the players left on your table including yourself. I suggest using a scale from 1:10.
How you do this will depend on you and your reads and ability. Multitablers may want to use only HUD stats, some people may want to base it on hands shown down, or bets made or past games or whatever.
The important thing is that each player on your table, from the point you want to do an analysis has a rating or relative skill level.
Let Si = stack size for player i.
Let n = number of players on table.
Let x_i = rating for ith player. (assuming from 1:10. You can have 6.5 if you want!!)
We require W_i to be the weighting of each chip in a players current stack and we require that W_1+...+W_n = n.
Let W_j = (n*(log(1+x_j)))/(log(1+x_1)+log(1+x_2)+...+log(1+x_n))
Clearly W_1 +..+W_n = n.
The n and the denominator in the equation fro W_j are merely normalising constants designed to ensure that W_1+...+W_n = n.
I used the log function to act as a smoother. This is like the mathematical version of luck. I.e. a bad player is never that much worse than you because of chance etc.Log is nice as it is monotonic so preserves the order of rating and allows more of a difference between very bad players and average players, than good players and average players.
Now to use ICM, simply calculate the relative chip stacks W_j*S_j and carry on as normal.
So there it is. Have a play, see if it is ok. It is only version1.0 so feel free to suggest ammendments. Post any ratings systems you find useful, or ask any questions.
Remember the main function of this model is to allow for skill to play its part. If working this should mean tighter pushing ranges against Donks, and more aggressive against good tight solid opps. This seems intuitive but ill need SNGPT before i can really start experimenting myself.
Hope it has some practical use
GW
|