Variance Calculations and bb/100 role

For winrate, use EV bb/100. It's a more accurate estimate of your true winrate since it reduces all-in variance.

For standard deviation, most people just use regular standard deviation in bb/100. That's quick and to the point, but will underestimate confidence (wider intervals).

The optimal solution imo would be something like this:

• Use SD of EV bb/100 for winrate confidence intervals (since EV bb/100 removes all-in variance, so you're more confident)
  
• Use SD of bb/100 for bankroll / downswing / RoR calculations (since its important to get the real magnitude of variance correct for these applications)
  

But the issue is "SD of ev bb/100" just isn't a standard stat included in hand trackers. You can approximate it by measuring the SD of non-all-in hands.

By the way, if you want a more accurate estimate of your real win rate, try a Bayesian approach.

Video Explainer (go to 9:43)

Spreadsheet Calculator

Edit: I created a custom stat to calculate the SD of your all-in adjusted win rate. You can find it in the PT4 community library by searching for "Standard Deviation of All-In Adj. Winrate"

https://www.pokertracker.com/custom/stat...

I don't see it when I search. Can you post the formula?

Here's a download link.

Formula (at the column level):
stddev_pop(cash_hand_player_statistics.amt_expected_won / cash_limit.amt_bb) * 10

thanks

Not sure if this is applicable here but info may be useful.

The proper way to compute the standard deviation for sessions of varying cash game duration was given by Bruce Z, former Probability Forum moderator.:

SD= Sqrt(1/n)* Sum[(Xi-M*Ti)^2/Ti]

where

SD is standard deviation
Xi is the amount won in the ith session (dollars or bb)
Ti is the duration of the ith session (hours or hands)
M is the win rate per unit time ($/hr, bb/hand, etc.)
n is the number of sessions

Hey man! Totaly get why your brain is spinning, the math gets pretty wild here. If you want to know how good you actualy are, focus on ev bb/100. It's the "skil" line because it cuts out the all-in luck, making it a way better mesure of your true winrate. When you apply that ±3.1 confidence interval to your EV of -1.1, you'll see that a positive winrate is definately still possible! The results line (-4.1) just shows what hit your bankroll, but it's too noisy for serius variance calcs. About the trackers like PT4 or H2N, they use your actual bb/100 results to calculate that standart deviation. Even though it's based on actual winnings, it's the industry stander for messuring your volatility. Using this SD to build an interval around your EV is the most acurate way to judge your progres. Basically, aim for a high EV line and ignore the results line, it'll save your mental health lol. Hope that helps

EV bb/100 is the stat you want to anchor on for estimating true winrate because it strips out the luck from all-in spots. The CI should be centered on your EV bb/100, not zero, so your range is actually around -1.1 give or take 3.1, meaning you could realistically be anywhere from a small winner to about a 4bb/100 loser. Also worth noting that the SD your tracker shows is measured on actual results, not EV results, and EV bb/100 has its own tighter distribution, so the real uncertainty around your EV number is probably a bit narrower than that 3.1 figure. At 70 SD over 200k hands you are still in pretty noisy territory, I hit 300k before my graphs started making any kind of sense.