I want to do a massive GTO+ DB, but my 3700x is slow and hydro is expensive in Ontario. I could get a better thread Ripper build for 3-5k, but then I may end up also spending 2k on hydro. Has anyone run the numbers and compared this to renting server space?
Thsnks

Sent from my EML-L09 using Tapatalk

I have a 2700x and it doesn't take that long, mby 3-5 months to get all srp,3b,4b spots solved for 1755 flops. tip, do the original solves with somewhat limited turn and river options if you're doing basic solves, you can add any sizes/lines after it's done and you won't have to resolve flop. you'd still have to use 2 sizes for turn and 3 for river to get accurate flop calcs, but don't go overboard with 5 sizes everywhere cause you just add them later in 2 mins at no time expense (for basic solves). all you care about is for the flop to be solved well enough for practical use.

I put together a 2nd 2700x pc with more ram to do preflop solves and that ended up ~750 USD. can do that and split the workload between the systems. I did spend a couple of weeks cherry picking some parts off ebay type places and saving on aesthetics but it can be done. Even a R7 1700 will do the job just 90% as well and mby you can still find these new for ~150 USD.

This helps a ton! I didn't know that you could add in turn and river sizes after. That is so cool. Mind if I ask you about your sizes? I'm not sure yet on my sizes but I'm thinking 33,75 for flop and 75,overbet for turn and river.

Also, how much more is your hydro bill when you do a ton of solving?

Sent from my EML-L09 using Tapatalk

you mean the power bill? I'd say it went up about 50-75% for the time I had 2 pcs on pretty much 24/7 but electricity is pretty cheap where I live anyway so no biggie, it's like 30$ a month with the pcs working.

sizes wise, I always do 1 flop cb and 1 raise size and would solve separately for any other cb size. normally tend do go a bit smaller everywhere than ppl will bet in practice just to have the solver mimic the higher freq betting people will have compared to optimal.

turn probing I do 40 and 110%, turn ip 66 and 125%, but you need to edit in a separate 33% size for when the flop goes x/r and oop x turn

river oop 25,75,150, ip 75,150. oh and let the solver jam at any point it wants to, have the add all-ins at the max 999% pot

I think this is all you need for practical use and you add bets or change sizes later on at your discretion. need to go to edit tree and the tool is pretty intuitive to use.

one more tip, if you're going to solve a spot that takes a lot of time, like srp which are 7+ days a pop for me, use the auto-save function, cause if the power goes out even for a second on day 5, you lose everything.

what you do is plug in ranges, build the tree, activate database mode and load all the flops you want, wait for them to show up in the list, then save this empty template in your temporary solving folder. can have this anywhere you want, mine is C:/gtotemp and you have to go to your c: drive and physically create it there. then in gto+ go to the most top right button, the icon looks like an open folder, tick all the options and process files. after it's done solving, the complete file will be moved to the regular folder your solves show up in. if there's an error at some point or power goes out, you will find your incomplete file in the temporary folder but after every flop it finishes, it auto-saves so you don't lose anything. it's the only way to have it save itself automatically, saved me several times.

Can I ask why only 1 IP cbet size OTF? I usually play 33% (rangebet) or 66% OTF when IP as cbettor.

sixers fan, my man

I solve separately, if you give it 2 or 3 cb sizes concomitantly, 1) it mixes between them in a way a human never will be able to and that affects turn play, 2) takes waaay longer to solve

what you can do is have a test run with small, mid, pot+ sizes, then sort the board types according to the cb size the solver largely prefers or the node you have more experience with, copy paste the flops from the editor and re-run separately with said size. not only do you get a better understanding of board textures when you need to put the work in, but you can see how turns and rivers are affected when you take different cb lines (and this is where the true value of studying with solvers lie, you're not trying to memorize anything, that would be impossible)

How does this look? I'm not sure about using 1 or 2 flop sizes. Its clear it is going to add a ton of time as it jumps from 10gb to 14gb. I think having two sizes is better for making this a hand review database (i.e, when I want to review a hand just find the board in this DB if sizes are close enough instead of solving it.) but probably one size is better for study as you said? When you mentioned splitting cbetting sizes would you run one 1755 DB with 33 and one with 66? Do you think that would take longer combined than one with both?

Also for raise sizes is it okay to leave it at default sizing (66%) instead of adding in a bunch more sizes? Also for turn IP should I use the c feature and use different sizes for turn cbetting and turn delayed cbetting? I'm not sure if having a smaller size for delayed cbetting makes sense?

Also what do you use for starting pot and effective? I'm thinking 2.5x open and call with .5 sb should be fine? Close enough even though some people use 2.25x or 3x? Though I am in the stars zoom pools with antes right now, but I feel like I should just ignore that dead money?

Also this is for BBvsBU SRP, using RC 100z solved ranges, but I'm not sure I should use these ranges, especially for BB, as nobody really 3bets like BB is supposed to?