I'm solving some spots here with the flops subsets from piosolver, I'm trying to find patterns on how the solver plays on certain kind of boards, but categorizing boards with categories like (dry, wet, static, dynamic, etc) doesn't looks like the best aproach, because two flops considered dry can have very different frequencies and the range advantage too isn't always to the same player even in textures considered the same with this method, then I tried with only 3 categories, low cards, mid cards and high cards, example (LLL, LMH, LHH) but with this method some flops have very different frequencies as well (A33 the solver bets almost the half than K33 in a 3bet pot spot that I solved), wich is the best way to categorize them that isn't a very complicated method but still have some precision? wich method do you use? a method that isn't too complicated, have some precision and is usable in practice, I'm not trying to mimic the exactly strategy from the solver, but would be very good to find a pattern on his average frequencies on different board textures

For something useful that you can use at the table check out these 25x flop types. From what I've read near gto solutions can be obtained using 184x flop types but unless you're rain man that is going to be really hard to remember.

https://cdn.shopify.com/s/files/1/07...46264467327999

Here is the post I got it from
https://www.piosolver.com/blogs/news...the-whole-game

I goup together by things easy to remember. I have a folder aceRagRag. Where i have a bunch of Ace high to rag boards, then study frequencies on just that board texture. Then i have kingRagrag. I have 5high, i have 5hightwoTone, 6high, 6high two tone.
I have aceHighLowPair, kingHighLowPair.
note: You are on the right track with what you are already doing. Notice the differences you got with the A33 and K33.
Run a aggReport on A22, A33, A44, A55, A66. then do the same for a folder with K22, K33... Compare to eachother, compare to a "Entire game flop subset"
Now when you play and see a KingLowPair or AceLowPair board you should be good! (move on to studying turn and river lines on A22, K22)

midConnectedTwoTone.... a bunch of 7s8sTh, 6h7sTh, Stuff that looks similar, keep it simple, start with a few similar looking flops and build from there looking for patterns.
midConnectedRainbow.


Another trick is group the way the HM2 CbetSuccess tab groups. (This makes node locking population stats easy)

Just go at it in a way that makes sense to you. TwoFaceRag folder.
look for patterns and prove stuff.

group all nineOrLess flops and compare the frequencies You get to one of the "Entire Game flop subsets"

This way if a nineOrLess flop comes you can adjust accordingly.

I spend alot of time doing this. You start noticing lots of patterns.

I also do agg reports on turn and river lines, look for frequencies that stand out using color coding in spreadsheets.

Just do what makes sense to you. You will start noticing patterns and its less work than what it sounds like. Do this for all positions. Comparing positions is also another way to find patterns.

Understand what makes those differences different on the flops you have studied so far.

Thanks man, great post Enviado de meu XT1033 usando o Tapatalk

I don’t have solvers, but if I did then I would look at the difference between (bottom of pair betting ranges) and (top of draw betting ranges), looking for how the board affects the frequencies.

I tried grouping flops into categories when I did some work on about 45 flops with Snowie, but I couldn't really draw many conclusions. Ultimately, I just sort of divided flops into 3 categories:
* Good for my range, with villain folding a lot.
* Bad for my range, with villain continuing a lot.
* So-so flops where neither player has a clear advantage.

I am working on the same problem. There must be something in between a set of representative flops and definitions that can be practically used at the table.

There are several sets of representative flops available, 103, 24, 45, etc..

My approach to this is to start with a representative subset of all flops and through simulation find a categorize them with a weighted value that represents a degree of goodness or badness. Like an EV for a flop based on your range. No results yet, but I am actively looking for ideas on how to do this.

For example, if in a set of representative flops you find that number 13 is the most profitable, how can you determine when some random flop is like number 13.

I believe that simulation will provide the most practical result most quickly. Any ideas on how to do this?

I will try categorize them specifing only the cards T-A and grouping cards 2-5 as L and 6-9 as mid, example LMTtt as the same of Ts8s3c and Ts6s2c, but I'm running the solver now so I didn't checked the results yet to see if this a valid method, but even this I think will not be so precise, example, Ts9s2c probably will have different frequencies than others, idk, I will check this later, depending on results maybe this could be a method of grouping flops

Enviado de meu XT1033 usando o Tapatalk

Might be a useful approach. I will try to simulate this. Something like this would actually be useful at the table.

I think that the soso flops are the most likely to produce large pots.

might it be a worthwhile attempt to categorize the flops by parameters of the equity distribution on the flop? flops with similar equity distributions should be played more or less the same and if you find 3-4 parameters that describe the flop distribution accurately (e.g. overall equity, "nuttyness", slope ?!) you might get only a handful of different situations.

Great post! It just might work. I will start by simulating the play of a large number of hands ( 1,000,000 ) using 25 flops and use the EV of the hand made. If there is enough difference in the EV values then I will have a starting point. It may be a few days, but I will post the results here. Thank you!

thank you for actually doing the hard work and testing my idea. looking forward to see how it goes

I ran a simulated play using the PIO 25 representative flop boards. I did prove that they are a good representation of all flop boards, as with over 1,000,000 hands played with both random flops and the 25, none of my results varied by more than a few percent.

I calculated the value of each flop bases on win rate but there was very little difference between the 25 flops.

Bottom line, I could not find differences between the 25 flops and thus could not find a way to use them to categorize flop boards.

I am now experimenting with flop boards characterized A-T high, 9-6 Medium, 5-2 low.
Also things like static, dynamic, pair, 2 suited, 2 straight and so on. I am calculating the value value of each flop board and that looks promising. More results to follow.

Thanks Pedro and ohly for suggestions.

You're welcome, I have to thank you for doing the tests, I think I was not very clear on the method I suggested, I already tested classifing them as T-A high, 6-9, mid and 2-5 low and it didn't worked, the method I was suggesting is to specify the cards T-A as their value, and 6-9 as mid and 2-5 as low, example JackMidLow, AceMidlow, QuenMidlow, grouping cards T-A and simply classifing them as high didn't give me good results in my tests, I think these high cards have so great impact on how the players hit the board that the strategy is very different, on A99 than J99, I believe that this why it didn't worked, so I suggested to group only cards to 6-9 as mid and 2-5 as low and not group the high cards, but what you said is good to add in the category, like JackLowLowpair or QuenMidMidtwotostraighttwotone

Enviado de meu XT1033 usando o Tapatalk

I will be trying a new approach. First, by using simulation, to get a value for each of the board types ( LMH, pair, flush draw, gutshot, etc.). Then I will use this information to assign a value to any flop board. Probably like wet / dry or dynamic / static but with a value assigned. The value being how much it helps my opponent. I already know what hand or draw that I made with the flop.

I have started an evaluation where the flop board is given a numerical ranking. Gutshot draws based on gaps, 0 gap, 1 gap, 2 gaps, 3 gaps. I give a numerical value to each. Same basic approach for straight draws. Then add a value for Flush draws, and pairs. Initial testing looks promising.

Next I will use the size of the showdown pot to evaluate..

If you make even the slightest change to the game tree, bet size, starting range, the strategy changes; usually quite drastically. Even with wide ranges.
Try running a 422 flop BB vs 40VPIP srp, keep the 40VPIP constant, but slowly add in the BB range weight on hands with a 2. Slowly add weight on A2s then add in K2s:.1 , K2s:.4 , K2s:.8 then add in a Q2s then start adding 32s, 42s....
And watch how the slightest change in range can make a flop go from a large bet size to checking close to %100

The only thing grouping flops is really good for is running sims. This has already been done, you can find them in the pio preflop folder.

Rather than grouping a bunch of flops together to develop a game plan, you are better of creating different game trees and learning how changes in the ranges effect the strategy. You will start seeing patterns of how a flop interacts with a change in the range, and how the strategy changes accordingly. Slight changes will make a difference, How much that difference effects EV is another experiment in itself.

I am not attempting to group flops. What I am trying is to assign a more specific value to a flop than wet / dry or static / dynamic. A numeric value, easily calculated at the table, super wet through super dry. Something like that. This is by using simulation, not game trees.

I don't understand what you are trying to do. But you can replace everywhere I wrote grouping above to 'assign a value' and game tree to 'simulation'.
Keep working at it then post the results when you are done. You might find something cool. Good Luck.
I guess my point is, a flop is a flop and its how ranges interact with it that give it meaning, which can change drastically. Kinda. So this value you seek will unlikely be a constant.

Edit: My answer above was for OP who is using piosolver not running sims.

Yeah I started to believe that try to group the flops and use the same strategy wasn't being the best aproach to me, somedays ago I started a new study method, trying to find patterns and try to find reasons on why the strategy looks like this on the flop, I think this method is better to learn and adapt on different boards, trying to find why on board 934 for example the cbet frequency is a lot lower than 734, discoverying patterns for this like villain having a lot more 9x on his calling range giving his range a higher concetration in % of top pair+ than our open range in btn(hero) vs bb and on 734 cbet frequency goes up due to his range being very capped and having very low 7x giving us a big range advantage for example, At the moment I'm trying this method ally with an excel spreadsheet

Ps: sorry for my english

Yeah exactly, for sure. Which you knew in you're first post. The grouping thing was not working out. Hence, you started the thread. Looking for patterns works well for me, and what changes them. If someone does find something interesting post it, for sure.

I once did a flop categorization for simplification.

Basically using A, K, Q, J, h, m, x to represent the rank. h stands for T and 9, m stands for 8 and 7, x stands for 6 through 2. I'd use r at the end for rainbow and tt for two-toned. For monotoned flops I treated them all the same. If I was to do this now, I'd probably divide monotoned flops for flops with the ace and without the ace. paired flops can be represented with () marks to show they are the same rank.

Some example flops using the notation:
AAAr
Khmr
Khxtt
Q(mm)tt
Qmmr
QJxr
hxxr
mxxr

For a flop like 743 or 943 (as notated by the last two in the examples) I don't think it will be that much of a difference in the real world if you play them the same. However, the higher ranking cards will make a difference. A43 will be different from K43. A monotoned flop of 654 isn't going to be that much different from K72. Okay, there obviously is a difference, but it isn't detrimentally significant.

You can try and create a numerical system to classify flops, but they are pretty dynamic so it is going to be hard to do. QJT is amazing for a preflop raiser. Change QJT to QJ9 and it has already lost a lot of power for the preflop raiser. Take a flop like AKT which is good for a preflop raiser. Change that to AK9 and it actually becomes even better. In the first scenario the change from the ten to the 9 made it worse, but in the second it made it better.

Maybe a good system is to take all the nutty hand combos and subtract them from the callers nutty hand combos. Going off of this you can create a flop with a numerical ranking.

Perhaps take into account vulnerability of boards as well. This way you have 2 numbers. The first number tells who the flop is better for. The second number determines how vulnerable it is. I'm kind of brainstorming on the fly, but this seems like a good simplification.

Thank you! I don't see how to implement this yet but will certainly give it a try. Looks promising.

I ran a simulation that associated flop types with the size of the pot and with the number of times that it went to showdown.

Large and small are relative to the average.
_____ is in an average range.
Straight draw is any possible straight draw.



Flops Pot SD count
Rainbow 39% Large Small
Ace High 21% Small _____
High Cards 38% Small Large
Pair 17% _____ _____
2 Suited 59% Large Small
3 Suited 5% Large Small
Straight Dr 82% Large Small
2 Straight 36% _____ Large
Gap 0 39% Large Small
Gap 1 34% Large Small
Gap 2 28% Large Small

Interesting, but I don't know what to make of the results. Any thoughts?