Today, I've been playing with the variance calculator.
I'm playing HT 10€ level. After 2k HT my winrate is marginal on this level(~51%). I'm getting 1% ROI. For that winrate is like you are on the side of the variance, and every month can be a 'losing month'. My conclusion was that probably I need to study much more and leave the tables for a while until I get something like 3.5% ROI which I think is possible because low rake (2.5%). (I barely do table selection because the low traffic.)
My problem and question is about the sample size per day. I'm playing 40 HT per day on average, because I play most of the time single table to be aware for every move of my opponent. But I guess that studying a little may help me to multitabling at 4 tables. Playing 40HT per day means something like 900 HT per month which is insignificant and highly variance dependent since I have low winrate. (I saw Jack the shipper playing 100HT on microstakes on a half hour video. 6 tables)
I think the best sample size per day would be something between 350-450 HT HU.
I know this may be subjective and It may be a 'fishy' question, but, just to get an idea, for a professional poker player what could be a standard sample size on low-mid stakes? Feel free to give some extra advice.
Thankyou for reading. Greetings
One thing to think about is that it's a common myth that increasing sample size lower variance. Actually the opposite is true, variance increases with increased sample size. It's just that relative variance gets lower. So what that means is you're more likely to experience really big swings in a higher sample, but the size of the swings will be proportionally smaller when compared to your EV over that sample. As an example, over 10 games a swing of 3 games isn't that big can feel really big since it's in a small sample, wheras over 100 games you'll maybe find swings of 25 games, which are bigger swings in terms of number of BIs, but are proportionally small simply because the sample was bigger so your EV over the 100 games was bigger too.
In the end I don't think there's any good reason to increase sample over increasing quality, especially at low stakes where ever ROI is important. Increasing winrate is the only actual way to lower variance, so focusing on improving in my book is much better than focusing on putting in a lot of games. Higher ROI also means you won't need as big of a bankroll to move up in stakes. Because of this I'm a strong believer that most newer players should only play one table as long as possible since it boosts ROI and also is better for learning (though be aware not all people are created equal, and I will concede there are some people who actually play better and improve more at two tables). In the end at lower stakes moving up in stakes is almost always going to be better for your hourly than adding tables, so I'd only add tables once you've very confident in your game and so don't need the extra improvement and the extra roi and are already at a decently high stake (or perhaps because moving up to the next level is very difficult due to reg wars and so instead of that you can add a table as the next stake at that point instead).
Hope this helps!
I'd never expected to be answered directly by the mathematician. All I needed to hear is contained on that text. The second paragraph contains all the advices I was looking for. It helps a lot, thank You Coffeyay.
The first paragraph let me a doubt: That means that we should have a bigger bankroll if we are expecting playing a bigger sample size? Says, 10.000 HT and taking your example: And reducing the relative variance to a five percent for example: 10.000*0.05=500 <- Could we face a 500 buy-in swing on such a big sample size?
Sacrifice minus profit equals zero
No, because larger sample size will give you winnings to pay the downswings with (as long as you don't cash them out). Bankroll is independent of sample size, only dependent on ROI (and variance, which is a function of format).
All macro poker decisions can be calculated as a function of win rate & rake.
Rake is a known constant.
Expected win rate & its probability distribution are difficult to estimate.