please keep me updated on this project, i find it fascinating

Yeah, I’m pretty interested in seeing results. Poker already has pretty fine grained performance measures (money) so I’m not sure I’ll find anything meaningful or interesting. Golf, for example, has a pretty coarse performance measurement (score vs par or number of strokes) and so the shots gained approach really ended up adding a lot of insight.

Mostly I wanted a “biggish” data set to work with and a problem I was interested in solving in order to try out some new data engineering / data analysis tools. But I need something I actually want to know to force me to do the work.

I’m going to start with just relative play by position measured by net funds won in the hand. Then I can do things like compare the bottom 20% of players to the top 20% to see if there are particular areas where players maximize their skill advantage. But this lets me get started without all of the complex hand history parsing/modelling or worrying about not seeing hole cards. Then I can go from there seeing what I find or if anything looks promising.

we should go 50/50 and create a training course, you do all the work and be the before/after marketing lead of zomg this knowledge made rickroll a breakeven rakeback pro w00t!

but seriously, your results would probably be quite valuable in the right hands - something to think about

Hah, I don’t think I’ll have the energy to see this that far. I know nothing about the online poker companies but I assume they have a bunch of models like this. But maybe not?

quite pricey training courses have been sold based on far less

either way, even though i don't play anymore, i would absolutely love to see your results and play around with your outputs - but i also do pga modeling re: sg data so maybe this is just something very specific to my wheelhouse

What kind of stuff do you do? One project I’d always wanted to do was a better leaderboard for golf. Taking into account difficulty of holes left and where players balls are (a long drive ending in the fairway on a par 5 is obviously very different than a shorter one ending in deep rough). Sort of a shots gained / expected value look at the scores. But getting that data seemed either impossible or crazy expensive so not really worth pursuing. Plus I only really watch golf that I’ve previously recorded and so my own live leaderboard isn’t even useful to me.

I also felt like there’s a whole world of visualizations that could be added to broadcasts based on the advanced data they have. But that’s really out of my wheelhouse.

that's something you would need to apply for the private API

there's a world of SG data that's not available to the public - they only grant academic access to it (i think the concern is sportsbooks and sharps using it for degen purposes) and that project if presented in the correct way may get you granted access

i just use it for dfs, i'm terrible at it, can't code for anything (i'm a dreaded non technical product manager) so primarily work in excel (meaning to switch to python long term but i try for shortcuts and have bugs/errors and am too much of a noob to fix and have a deadline so then back to excel)

what i produce is hot garbage, but surprisingly similar to vegas h2h lines quite often - which just goes to show everything is hot garbage with golf

but hey my hot garbage is still better than what the field does, which is "Finau is really feeling it I'm playing him" or "Horschel is from Florida I gotta play him there"

Would generating "legitimate-looking" hand histories be more or less difficult than the analysis you are doing on said HH's?

I'm imagining Super Smash Bros. when we would set up 4 computer players to duke it out. Turn one AI to 9 and all the others on 1 and you see some funny stuff.

iiuc he's not trying to monte carlo ideal playing strategy a la a solver but instead review actual play history to see which players/strategies excel the most in given situations -which in theory shouldn't be an "optimal" line but rather an exploitive one that the rest of the field hasn't yet adapted to

ie if you want to measure what makes tiger woods such a good golfer, you can't just sim golf events, you need to review what he actually did


also, on a side note, i'm considering doing a startup that is highly related to the general 2p2 demographic - plan is to go unfunded until there's a protoype to show to vcs so huge likelihood of spending some free time building something that fails with no compensation or even worse, something that gets funded and then sputters out after a year or two - pm me if interested - very much at the very early stages so at this point would just be a general discussion to gauge interest - could be something you moonlight on given lack of funding for the project

played a **** ton of golf from an incredibly early age

Yeah, basically what rickroll said.

So in the Tiger Woods example it’s not about training/background but about what parts of his game were dominating. Two players could have the same general results (wins, average scores, etc) but arrive at that same destination very differently. One could be a massive driver and mediocre putter. The other could be an average driver but amazing putter. The raw golf scores aren’t going to tell you that. Even coarse metrics like fairways in regulation / average puts don’t help very much. Shots gained is one way to try to have a consistent metric that can be applied to all parts of a golfers game so you can compare the value of being an amazing putter directly with the value of being an amazing driver.

Similarly with poker you might have two winning players that get there very differently. We have some sense of that with things like aggression or vpip but it might be interesting to have a consistent metric across all parts of the game.

I find this very reassuring and relaxing.

Stripped screw in a GTX 980Ti PCB

Any recommendations to get this mother****er out?

one of those wider rubber bands between the screw and your tool
and go with a larger bit or mby a flat head

Try a sharp-edged flathead and hammer it in a little. At least it works with cars, never tried it with a circuit board.

The problem with hammering is that the other side is plastic, which I can only take off after unscrewing (lol)

this happened to me, took it into a specialist and he used a very tiny drill

chinese repair techs earn poverty level income and are known for quick shortcuts and brute force (i've seen one open up my phone with a chisel and mallet once) so don't know if that's the optimal approach or the "i'm only getting paid a dollar and want to be done with this ASAP and it's not mine anyway" approach

Can you sharpen a small flathead screwdriver and lightly try to tap it in? You just want it to gouge into the inner edges of the stripped hole a bit, then you can screw it back out (maybe). You're not trying to hammer through.

Gabe,

I can't tell from the picture if the screw head is raised. If so, could you grab it with some vise grips or pliers, then unscrew it that way?

Use a Dremel, carefully cut a flat head in to it.

@jj - those HH selling sites are censored sitewide due to previous spamming campaigns, and also the major sites ban use of such "datamined" hh databases. Though for research purposes only, it's all good.

The screw is raised a decent amount

I can try it that way, but I'm shitscared that applying force I will screw up and knock off some parts of the PCB

I get that. Probably better than trying to hammer in a new face. I like Dave's idea with the Dremel. Seems like that shouldn't put too much stress on it.

ez
https://www.youtube.com/watch?v=jeUL...ountyFasteners
but use a bigger bit for more contact area and preferably no power tools on pc components, tho you might need one here
if you dremel and leave metal shavings on the pcb ,you're going to have a bad day

I ended up doing this, after using a file to file a bit off opposing sides so that I'd have a bit more grip and the screw came out pfffff

Bought that GPU sh for 250€ and didn't want to screw it up, literally

Thanks for the support, guys

Incidently I'm selling a PC Gaming with Ryzen 2700X, MSI GTX 980TI Gaming, 16Gb RAM 3000 MHz and 512 GB SSD NVMe for 700€