Solver, bb/100 diffrence , what is a big deal and what is small deal?
Posted by ASAP1Z2
Posted by
ASAP1Z2
posted in
High Stakes
Solver, bb/100 diffrence , what is a big deal and what is small deal?
when i compare 2 simulations which diffrent betsize where sim1 is easy implementbale, sim2 a bit harder but sim2 has a higher ev, lets say 5bb/100 more than sim1,
its hard for me to messure which bb/100 amount is big and which is small if u looking at the bigger picture. so which bb/100 diffrence or ev diffrence i can neglect and when i should be careful .
do you have any rough guideline for this? for example on the flop everyhting about 10bb/100 diffrence in ev is a big thing
i know there are other factors like how many mistakes make the population vs these 2sims, how much they punish a simpler strat and so on, but i just want to solve the ev question here for me.
Because im complete clueless how I messure a bb/100 , ev difference and come to a conclusion if I choose sim1 or sim2
Loading 8 Comments...
1% is general rule of thumb for that(for many pro players), if less than 1% of total pot you are good to go with that strategy.
So u mean 1% of EV diffrence? So for example if the pot is 90 chips and
strat A captures me 70EV and strat B captures 70.8, strat B is okay to choose if its much easier to implement?
also do u know which pro player said that? i can only remember nick howard said something like this when he was working with these bet all flops 33% simplefication in his video courses.
Ben Sulsky mentioned it in last year video I believe. Few other pros from Rio as well (can't remember name)
I meant 1% EV difference compare to optimal strategy. In your example if pot is 90 chips we want EV difference be less than 0,9 chip compare to optimal.
I guess you are talking about different things. :) OP asks for differences in EV for different solutions (i.e. rounded betsizes, or restriction of betsizes from 5 to 3 - to make it more executable for humans, that lead to differences in EV for the solution).
What Jeff explained was EV deviation from optimum ("nash distance" as a measurement for the solution's precision). Naturally that leads to an EV difference as well.
I made a more detailed post somewhere else where I answered that exact question in greater detail, maybe you wanna search for it. But in short - the answer to both questions (what EV difference is acceptable) is equal: it depends on how often that scenario would happen in real life, hence, how big it's impact would be on your overall strategy (which is the base of your overall winrate). Like a "defend vs. cbet in SRP OOP vs. BTN-open, 100bb" happens like a gazillion times more often than a "defend vs. river x/r as PF-4bettor after x/c,x/c,x, 300bb deep". In the first scenario, 10bb/100 might be a gamechanger, whereas in the second scenario you might not recognize a difference if you had a 1k bb/100 difference.
BigFiszh
hey, thanks for the answer. really appriciate it, cause on this topic i feel, noone has a good answer. i will def search for your other answer.
but still your answer like 10bb/100 might be a gamechanger is still to vague for me, i still cant do a lot with this.
for example i study atm BTNvsBB 25BB deep, which happen in my format quite often, its like a bread and butter spot
the pot after btn mr , sb fold and bb call is 90 chips
lets take 9s 8s 2d,
IP EV(25,50,75,100) so 4 betsizings mixxed is 62.29
IP EV (25_100%) so 1/4pot 100% is 61.98
so the diffrence is 0.31
*100 = 31
1bb = 20 chips
so from playing a 4 betsizes higly mixxed strat on this board
to a 1/4 pot 100% cbet strat i loose 1.55bb/100
which seems not too much, so that i want to implement the 100% 1/4 cbet strat
but now when i took for example
9c 5c 2s
here is EV diffrence is 0.98
so 4.9bb/100 diffrence
so for this board, a 75% mixxed cbetstrat with roundings yields to a
better ev.
but still my question is kinda unsolved, where to draw the line in these spots
is 5bb/100 a big thing, is 8bb/100 a big thing, should i worry only if its >10bbb/100. so to ask a concrete question, in common spots, which bb/100 diffrence is a gamechanger?
* WALL OF TEXT FOLLOWING - SUMMARY (and SOLUTION) AT THE END *
Short answer: your overall winrate is formed by any decision in ANY single spot you get into. If you accepted a 5bb/100 trade-off in any single spot, your overall winrate drops 5bb/100. So, you define, how much you want to sacrifice.
Long answer: Start with a targeted winrate as a baseline. Define the trade-off you'd be willing to accept. Like, say you aim at a winrate of 5bb/100, and you'd be willing to accept a minimum winrate of 3bb/100.
Now, calculate how often the spot you're looking for happens. For example BTN open, SB fold, BB call. How often does that occur?
Let's make some assumptions:
a) In 17% you're on the button.
b) In (85% * 80% * 75% =) 51% the pot is unopened.
c) You open 45% when the pot is uncontested.
d) SB folds 80%.
e) BB will call 25% after BTN minraise open and SB fold.
f) BB will check the flop in 95%
Now, how often do you see a flop and get checked to after BTN minraise, SB fold and BB call?
P(x) = 0.17 x 0.51 x 0.45 x 0.8 x 0.25 x 0.95 = 0,0074.
That means, this spot will occur in 0.74%. Next question is, how often does this special flop comes up? The one flop mostly is worthless, as you won't be able to memorize the gameplan for each individual flop anyways, so you got to cluster them and build groups. Say, this flop category comes up 5%. This means, the spot we are talking about makes 0.0074 x 0.05 = 0.00037 (0.037%).
Now, if you accept a deviation in winrate of 5bb/100, this makes an impact on your overall winrate of 0.00037 x 5 = 0.00185 bb/100. Quite marginal, right?
But here's the deal: obviously, if you accepted a 5bb/100 offset for ANY imaginable situation you could come into, your overall winrate drops 5bb/100, turning your aimed winrate into zero. That means, in spots that makes the "bread-and-butter" of your game, you should accept a deviation that is not much bigger than your overall accepted reduction (2bb/100). In spots that occur less often, you can - proportionally - accept higher tradeoffs. So, reduce the accepted reduction for the major spots to -1bb and build some buffer for spots that come up less often (so you can reduce for example -20bb in rare spots and get to -2bb on overall average).
SUMMARY / SOLUTION:
Even though I love academic mathematical numbercrunching masturbation orgies, that is too much - even to me. :) Just forget about it. Study, go as far as you can - and don't care about the rest. Even with a simplified strategy and even if you lose 5bb/100 by simplification, if you manage to really internalize the basic idea behind that strategy and keep your ability to deviate where applicable, you will crush.
Hope that helps. :-)
BigFiszh
Thanks BigFiszh, awesome post, love it <3
I did also some Flop Categoriation , like 85% flops in nl are 9xx und higher
Axx 22%,Kxx 18%,Qxx 15%,Jxx 12%, Txx 10%,9xx 7%
so i guess for axx flops u want the lowest drop in bb/100, which is kinda good, cause axx flops are often automatic very suited for simplyfication cause the range advantage.
lower flops like 678s or something are rare, but very bad for simplefication.
so i think i will just try to get 9xx+ flops to a max of 0.5% and lower flops 8xx and below maybe a max 1.5-2% on the flop. still no one can say me if this is good or bad, but at least ive got a startingpoint, that makes logical at least some sense.
the problem im a perfectionist and this is stuff, where perfectionist often slows u down and theres no solution possible and decision paralyze is a very dangerous and frustrating terrain.
Well one thing to keep an eye on is the accuracy that you are solving the complex version to. You can certainly allow a reduction or up to that amount. I have seen 1% discussed a lot as well. But I would do sweeps of situations with increasing depth.
Better to have considered 10 situations even in rough simplified form rather than one in 10x the depth for the obvious reason that improvement rate on a given situation is probably sub linear in time.
Easier said than done when something gets intersting!
Be the first to add a comment