@tycobb:

I mulled over this exact idea yesterday but realized I don't have the skillset to make it happen. Instead, I learned how to not crash 100% of the time after taking off in DCS World. I'm tempted to say that I took on the more difficult project :P

Hmm...I didn't really think about doing this, but now I think I have to do it. May the best SaltyBot win!

Awesome, will be looking forward to this.

You are crazy.

The biggest problem with this is definitely going to be handling characters with the same name, but the statistics might work themselves out if those different versions show up with roughly the same frequency.

@ravenlight:

Very nice. One of these days I will break down get myself into those games. Perhaps next year after I upgrade my video card this year.

@chaser324: Oh man, challenge accepted! Yea, that's the biggest hurdle and I am pretty sure there's nothing I can do about it except let it ride. My original idea of looking at the avatars has gone out the window. I noticed that the bet screen does not show them. That was going to be my only chance of doing it.

Updated: Now logging fight results.

Alright! This bad boy is going to run all night and tomorrow while I sleep and work. Will see about exporting my results to a Google Doc so I can share my findings.

I thought about doing this after losing so much damn Salty money tonight, but quickly realized I had neither the skill nor the time. Godspeed!

Wow, that is super cool. Are you planning on building a serious predictive model?

(...You know what would also be cool? A virtual betting site for SaltyBet bots.)

I've got all of my automated screen scraping/interaction code working. Now, I just need to setup my code to track results and make educated guesses on how much to bet on who.

I'm just going to leave mine auto-betting $10 on red all day. I'll see about getting it to do something more meaningful tonight.

why do this? IT's only fake money. You're gonna game the system for fake money?.......cool. okay. :)

I think Salty was trying to prevent this yesterday and had the auto-update stop momentarily. Would suck if this will somehow change what we have now.

@gnatsol: It's not really about the money. It's primarily just a programmer curiosity thing.

@forteexe21: There are some ways he could combat this if it becomes a widespread problem (periodic CAPTCHA, etc.), but it remains to be seen if that actually happens.

@audiobusting: I like the way you think. I am going to try to and build to it to make the most money, so... as serious as I can get it. The only problem is I do not have much data to scrape which is why I am also planning on implementing tiers.

@gnatsol: Because I want out of the salt mines! It's going to be my time! Also, this thing predicts stock prices.

Not really.

But maybe

@chaser324: Glad to hear you got something going. Curious what luck you have betting Red. I ended up in the hole.

Updated:Current results are in a Google Doc. Link in post.

I say everyone is better off at learning more about the mugen community. The rich are rich because they came in knowing about most of the characters being used. Their strengths, weaknesses, how they play.

This is awesome, can't wait to see what comes of this. Gotta get them saltybux

Hmmm.... just noticed Cammy fought under the name "Psycho weapon". Hope that name is legit for her seeing as how the name "Cammy" was displayed in the game.

@tycobb: My "Bet $10 on Red" policy has put me at about +$200 today.

so basically you are trying to ruin the fun? Great goal guy great goal. If I could report you I would...

This isn't getting released anywhere and I paid for a subscription to help support the site. Don't be mad because what I am doing affects you no possible away except for a few fake dollars here and there.

Hey, do you want us to give you our win/loss records? That's, like, free data, and it tracks when the match happened so you can avoid dupes.

@itwongo: I would love any data you have prior to 8/5/2013 8:30 PST. Just let me know what time zone the times are in so I can convert to UTC. Much appreciated! I am currently in the process of building in the display for giving me the fighter win/loss records once the bets are open so this will be perfect. So far I have recorded 417 fights since last night.

@tycobb: It doesn't record times on the better's timezone, as far as I can tell. I'm practically addicted to this thing, but I definitely wasn't up til 3:40 am last night. I'm on Pacific time, so that'd be three hours ahead of me. So it's recording data in East Coast time.

And separating same-name characters is gonna be difficult, unfortunately.

Big ol' table:

@itwongo: Awesome. Thanks a lot for this. I'll try to get it imported soon before my blog update. I'll also update my Google Doc of seen characters and results.

Google Doc has been updated and updated this blog wtih Version 0.04. Hopefully I can @itwongo's results imported in during lunch today.

Sounds like you're having some solid success. I haven't had time this week to get my own bot doing anything "smart", but I'll let you know when I do.

@chaser324 What are you writing it in? I meant to ask before, but forgot. I am curious as to what you are using to interact with the page.

@tycobb:

I guess I'm taking a slightly different approach from you. What I'm building is a bit more of a hands off thing that doesn't really allow for any user interaction while it's running.

My basic approach here is to build this as two separate components:

• I have my site scraper for interacting with the SaltyBet website written as a CasperJS based script. If you've never used CasperJS, it's an offshoot of PhantomJS which is essentially just a headless WebKit browser. It's generally used for testing web apps, but I've found it to be extremely useful for web scraping, too.
• Then, I also have a very simple NodeJS server that manages all of the data in a MongoDB database and accepts GET/POST/PUT requests from the CasperJS scraper to access and update data. Part of my thought process for separating this portion out is that I could potentially build a nice looking interface and throw it up on a public webserver so that everyone could have access to this data.

Right now, I've got the scraper and server working as separate components, and I really just need to get them tied together.

Once I get it all in a somewhat complete state, I'll make my GitHub repo public and drop a link in here.

@chaser324: That's a pretty nice approach. I am going to have to look into MongoDB. I've always been curious about document databases, but always forget about them when I start a project. It's my comfort zone and usually I want to get stuff done ASAP.

No development last night and probably won't be any until this weekend. Video card fried yesterday and the machine won't boot unless I have a working one. New one should arrive tomorrow and I will of course want to test games out with it ;)

Just imagining all the stats I am missing out on just pisses me off. lol

I seem to be doing okay betting $50 so far. I still have some losses in there, but somehow the odds have made it so I am actually matching by bet or more on a few of my wins.

I would probably have a lot more money if I just let the bot decide. I keep wanting to go all in on ones I think will win and that usually just ends up resetting me. D'OH!

@tycobb: Run multiple accounts with different browsers and you might have better results (one where you input one where you don't etc).

This is simultaneously ridiculous and rad.

Good job dude, this sounds amazing.

You're famous, dawg!

This seems to be spot on for something like Saltybet. Good job!

@itwongo: Holy Shit! haha

Are you planning on releasing this when it's done? Looks great.

Does your bot have access to the actual Twitch video stream? If so, you could use the character icon next to the health bar to determine who it is when there are name collisions. The icons are always the same and in the same place.

@ttthrasher: I thought about scraping for the icons, but the problem is that the icon isn't in the weigh-in screen. I could have more accurate fight records for different flavors of characters, but when time comes to bet, I still wouldn't know who to bet for and leaves me now having to guess which fight record I want to use.

@granderojo: I am still throwing the idea around. There are still a lot of things that need to be done in order to make that happen. Change UI controls to get off of third-party due to license issue and change the database to something else. Most importantly is somehow getting my win rate to 75%-80% (shooting for 80%). Right now it is at 72% and there's just no way that is good enough to release. SaltyBet is growing and changes seem to be getting made. I need at least another week of fight data in order see where my stats are and where I can improve.

I'm still trying to piece together my brain that exploded when I tried to follow your post. Awesome stuff, duder.

Gonna pop in here and warn you guys. Apparently, using bots is expressly forbidden by Salty Bet. Scroll down to the bottom.

@fadedoasis said:

Damn...well, I can pretty easily just remove the automated betting portion of my project and just use it for the purpose of stat tracking. The Illuminati stats don't go all that deep.

Ah, you may have bots watching bots, but who watches the botwatchers?

They may delete my bet history, but they won't delete my database! xD

I am not releasing this so oh well. This is more of a personal project to just see if I can get great results. Releasing it would just be a bad idea because all the stats odds would be so one sided... wait it already is like that.

I don't really think my automatic bets of $50 and my $3800 in Salty Bucks are hurting anyone's bets lol

A few tips from an ex pro poker player.

Risk of ruin(going broke) was a key factor in what level of poker one would play, as your bankroll is your livelyhood. This includes any money that you could use to resupply your roll(I played online only, so it was easy to mix your life roll with poker roll if you didn't make strict rules). The problem is, in SB's case, there is no risk of ruin, as you constantly have a resupply of Sbucks if you go broke. The only time you risk "ruining" anything is when you've finally built up a roll to where losing it would mean hours or days of work.

In this case I would never make the minimum bet less than your base resupply(depends on what level you have achieved or if you've given money to SB), according to this: http://saltybet.com/leveling . after that, well, how risk adverse do you want to be? Cash game players that play at stakes just high enough to make a living(in Canada, I would never play lower than $.50/1.00 blinds for no limit holdem, if you know what that is), usually never go below 100 buyins. Others would call this extremely conservative, and for the purposes of SB, I would as well (however 100 buyins would put you well below a 1% risk of ruin, even if you were living off of poker site bonuses). To get a real handle on this, though, you really need to know your winrate.

Winrate in poker is usually calculated at big blinds per 100 hands(bb/100). Buyins are usually calculated at 100 times the big blind(remember that $.50/1.00 thing in the previous paragraph? the .5 is the small blind and the 1 is the big blind, id going into more poker rules, but that would just confuse things). this number was arbitrarily decided to limit the amount of money at a single given table of online poker by the online poker sites, if you're wondering where it came from.

This doesn't help us in this case, except that it does tell us what kind of expected percentages of a buyin we'd expect to see over a given number of bets. these days, if you are winning at 5 bb/100 you are considered a pretty good poker player. The reason I go into this is because it allows us to use a site like this

http://www.reviewpokerrooms.com/poker-games/general/bankroll-requirements.html

to calculate how big of a bankroll we want. We need to use this site because I don't know of a utility that will do it buyins. WARNING: that site uses BB/100, or big bets, which is double the value of bb. this comes from a time when people played limit, not that you care. so if you had a 2% buyin per 100 bets winrate, it would be 2bb/100 or 1BB/100. We still have the problem of standard deviation, which is explained here:

http://easycalculation.com/statistics/standard-deviation.php

once you calculate those numbers, and put it into the first site I linked, out pops a number of big bets(or 2% of our buyin). I apologize if this was overly confusing, because it didn't have to be if I had the right tools. I have a feeling that because of this, all of this extrapolation will lead you down a garden path as far as actual numbers go. Also, it doesn't help that SD is really advanced(I don't really understand it, I had a database myself for all of my poker hands played, and it popped out a number for me based on them).

Anyway, what it comes down to, is that you need a fairly large database of bets to find an accurate SD(I think, math majors probably should look over this and maybe spruce my post up so it is more understandable). The other problem is, that ODDS CHANGE BASED ON AN EXTREMELY LARGE HUMAN FACTOR(twitch chat in this case), and that you won't know the true odds until betting is over. What we could do, is start to track values based on odds for each individual fighter(which will in turn show their general popularity trending), and extrapolate based on past histories(something you are already doing).

But we can take this a step further. At this point we are basically betting with the community(maybe a little bit better due to new bettors, etc.), which is going to get us to break even over time for the most part. What we need is to start factoring favouritism of heroes, chat trolling, etc and use that to size our bets based on when it is most profitable to bet. This is getting into blackjack territory and card counting, but it comes down to this: if we know the value of a certain hero is higher than another, but the odds will not show those numbers, we should be betting bigger the more disparity there is between those two numbers(which could mean we are betting on the hero more likely to fail, if we are getting the right odds). That means that OP heroes, while guaranteed to produce profit, aren't going to produce the biggest margins(unless they are against other OP heroes). This tends to mean that closer matches are the ones we will net the most profit on, and thus should be bet on with bigger bets. Also, the more powerful the two heroes are combined, the more variance there will be in the match, especially if they are extremely close in value to each other(this is the kind of thing that would lower our betting size). As our database improves, we can also start to predict upsets against conventional wisdom(which is different from trolling or troll wisdom, and usually has something to do with how the big bettors that aren't betting their whole bankroll every time are betting, which means we will probably have to start tracking that as well), which will replace close matches as our big money makers. I also think this means we can start tracking keywords in chat, and extrapolating how that affects odds based on W:L records and newness of heroes.

Anyway, hopefully this can get picked up by someone who actually can do math, as you don't actually need to know math to play poker, you just need to value concepts properly.

If you have any questions, PM me here and hopefully I'll see that I have PMs, and that I can help. I come to this site frequently, but I don't comment/use the forums much, so I'm not logged in often/check to see if I have any PMs cause I don't know anyone here.

Oh yea, don't forget that our own bets will have an effect the overall odds, and that the richer we get, the more of an effect we will have.

It also says "THE UNDERGROUND SYNDICATE" right there at the bottom.

One other thought before I go, A more accurate way of tracking strength is how badly someone won or lost, and factoring that into variance. If a fight is close, either side could've won, thus they may not necessarily be winning every single time, which means we get more meaningful data faster.

So and now you are talking of releasing it and offer charts etc. Yeah sure. Way to ruin the fun. But yeah that is so human like always try to destroy the good things by trying to controlling them...

Have fun with you ridiculous task.....

Yeah, fuck it. I'm calling this whole thing off. Even if I tracked stats and made them publicly available, I feel like I'd be undercutting Salty's business of selling that info (although I think he could and should improve the Illuminati stat tracking).

Who the fuck is Salty?

I go away for a week and the world goes bananas.

I am so confused.