That's a very hard question to answer. The reason it's hard is because a struggling player is mostly clueless to wether a stat constitutes a leak or not. You need parameters, and those aren't known for an average struggling player (I'm assuming you belong to such category of players since you are asking this).

1st Method: Compare your stats to winning higher stakes players' stats. Even though higher stakes winning players arent perfect, they are certainly doing more correct things than you (again, assuming you are struggling at lower stakes), so using them as a parameter is good enough for your purposes.

2nd Method: Hire a coach to point out those leaks for you.

3rd Method: Develop a deep understanding of theory and player pool tendencies to figure out whether a stat is a leak or not. For example, if you know that GTO Cbets at a very high frequency in position and, on top of that, your player pool overfolds to cbet oop, then having a low cbet frequency IP is certainly a leak.

The 3rd method is the best for your development as a player, but its the hardest and most time consuming. The 2nd method is the least time consuming, and very effective for your development (if the coach is good xD), however its most like the most expensive. The 1st method is likely the least expensive and it doesnt take too much time, but its probably the most ineffective, since the players you might compare yourself to may not be that good and you might misinterpret information in the process.

thanks Saulo, appreciate your thoughts

I am surprised this didn't get more responses, would have thought a lot of people are looking to identify their leaks through database analysis?

I would add that for a "learning player" it can be very difficult to "fix leaks" by the method you mention:
1/ identify
2/address
3/rinse repeat....

The biggest issue I see is that most of these stats are related either in a major way or at least a minor way. So for example if you think you are not defending vs a Cbet enough, you still need to know how calling or check raising more will change your actions and ranges across various runouts and opponent actions.

I would actually think there are only two viable methods to improve using a database.

1/ Find a coach you work well with, who understands both theory and database stats well

2/ Learn the theory yourself, and ONLY THEN try and fix your self identified leaks.

Truthfully, by the time you have accomplished #2, you will probably be a good enough player that you will be thinking in terms of "lines of play considering Range vs Range" rather than a specific stat in a database.

Just my two cents though....

TK

I'm sure there are people that read the thread and didn't respond to it, simpy because there is not much to add to the discussion

Broad statement! ;-D

I've just stumbled over this post and feel challenged. I've done (and still do) dozens of database analysis for my coachees (/ad :D). This just ahead to show where I am coming from.

OP did not ask for how to "fix" those leaks, he primarily asked for how to identify those. And actually this is quite easy - even with limited knowledge about theory.

For example, take the cbet defense% oop in SRP. It's X%. Is that too much? Too little? Just about right? Should we raise more - or call more?

The answer can be found pretty simple - by intelligently using filter functions:

1) Define a baseline for this scenario. That means, setup a filter for all comparable situations (SRP, marginal hand, 95 - 105 bb deep, etc.) where you found yourself oop facing a cbet and folded! Now check your loss in bb/100. This is the one decision we can always fall back to.

2) Next, substitute the "did fold" filter by a "did call" filter. Check your win/loss-rate again. Is it better than folding? In that case, you fold too much!! Simply said, because in any single case where you had called instead of folded, you would've gained money.

3) Next, substitute the "did call" filter by a "did raise" filter. Check your win/loss-rate again. Is it better than folding? In that case, you again fold too much! Is it better than calling? Then you're calling too often and not raise enough.

Simple, right? Beware though, these are NO fixes yet! Knowing that we fold too much, call or raise too much / too little is just a leak so far. We still don't know what the fix is. This is comparable to a Nash solver that can precisely tell us how far he is away from the Nash strategy, w/o having a clue how to get there or how the Nash strategy looks like.

Fixing it needs a lot more theoretical knowledge, comparison of proven winners, working with a coach or simply discussing here in the forum. But identifying leaks just needs a bit of craftmanship.

BigFiszh

PS: A "word" of warning along - regarding sample size!!

Be aware that scenarios ideally have to be fully comparable. If you find yourself with A-high + mid overcard on a 7-high board in a SRP oop with 100bb effective stacks, BB vs. MP, you should filter for that exact scenario, same holecards, same board structure, same positions, same eff. stack sizes.

It's very likely though that you quickly get to a sample size that just is insufficient! Don't ever underestimate, how much variance can influence results, if we speak about very tiny fractions of stacks.

Example: you have a sample of 1,000 hands in a given scenario. Accidentally, in one single hand of this sample you won or lost one entire stack (not visible in EVbb). 100 bb won or lost makes a difference of 200bb. Over the given sample this makes a swing of 20bb /100! This can easily lead to completely misleading interpretations.

Obviously this is even magnified if we talk about scenarios that are uncommon in and of itself, i.e. x/r as 4b oop.

The solution is easy. S-A-M-P-L-E-S-I-Z-E!!!

100,000 hands is not a database to work with seriously at all. 200,000 hands allow very basic scenarios with broad filters. 500,000 hands let you analyze most flop scenarios and a couple more complex things. 1,000,000 makes it interesting to dig deeper, up to the river. 3,000,000+ starts allowing analysis for unusual lines, i.e. x/r on the river and stuff.

Now, I hope I actually did contribute something substantial to the discussion.

Thanks BigFiszh, that’s exactly the kind of advice I was looking for and yes I was looking more for identifying than fixing at this stage.

Certainly going to play around with those methods!

Wonder if there are techniques lurking out there...

For finding preflop leaks you can do the same principle, and dont need this big of a database ;) Here you can also group a few hands for smaller databases.