Also you can continue to bet while the draft is going on, but the line doesn't change - probably a decent edge to be made by taking advantage of that

GL dude ur gunna need it!

I didn't get any bets on the match. I went to sleep pretty much after I put up the output for the match. Watching it now.

It's a shame there wasn't a handicap up for this game. I think there would've been a small edge on Newbee, a similar situation to the EG vs DK game where EG winning the BO3 only had a marginal edge but the handicap on EG had a decent edge on it.

Did you actually get a bet on during the draft?

Hero picks play a huge part, teams get out picked very often imo.

I had looked into doing this because you can get an extra edge, I found some code that someone had made to predict the winning team based on hero picks alone. The model could predict the winning team of a game 65% of the time (on average) based on heros alone. This would vary with patches which is not ideal, a big error margin would have to be taken into account. Ideally I could use conditional probability to "update" my model's predictions with the hero picks predictions and get bets down after a few picks.

I have tried to get bets off during the draft a few times. Although the lines stay up on the website during the picks, when I clicked submit it would not go through. I did try after around 8-9 heros picked though.

Results from yesterday:



Here's the output from the model for the first two games tomorrow:



And here's my bets thus far for these games:



Not really sure if I should also bet on CLoud 9 winning the best of 3. I went for the handicap because there's a slightly higher edge, but the best of 3 will pay out for more. I might be being too conservative. Any suggestions?

I'll update the thread before the matches if I got any other bets on, I'll also try to post the output/bets for the matches that follow these two.

Poker Update:

Graph since restarting poker:



I got some rungood and recovered from my mini downswing that I was on when I started this thread. I feel as though I'm playing way better than I was before I burned out.

Going to take a small shot at 25nl. With a bit of luck I'll jump up limits on the 1st try.

Yeah I did get a bet on after the draft to my own demise. EG killed it with their first draft and got close to their dream lineup so I bet even more on them, then they get wiped out 2-0 haha. Oh well.

I went and bet on both cloud 9 +1.5 and them winning the Bo3, but I bet more on them +1.5.

Bad luck, Newbee is on fire! How many heros were picked when you got the bet on? I might look into this in the future.

Pinnacle raised their price on Invictus Gaming for the 2nd game:



Here's my updated bets for the day:

i'm like 80% sure i got my bet in after all the heroes were picked, at least 6-7 were definitely picked when i'd bet

Cool I might try the "update" thing I talked about.

EDIT: Damn, I missed a bet on the 1st iG vs LGD game. The line went offline about 2 seconds after the last pick (all 10 heros picked) on the stream. I will try watching the draft in the game client (should have less lag).

Here's the output for the 3rd series today. Looks like there's quite a bit of value on VG. I've got a bet down, gonna sleep but I should post the output for the last series.

Just curious, but shouldn't it be +EV to just bet on everything that has a positive edge? Or if you're not comfortable with that pick a # for example 5% and everything 5% or better play it?

I tried looking at the picks from the Dota 2 client, there's a good 10-20 seconds after all 10 heroes are picked before the lines close

To be honest I'm not really sure.

I've been having just one bet on the series (the team winning the best of 3) or on the +/- handicap odds, whichever has the highest edge %. Not really sure if I should be betting on both handicap and the series in one game if both have a decent edge. I might be being too conservative. I'd love to hear what people think on this one.

I wouldn't bet on each game from the model alone, just because the hero picks play a big part in who wins a game, this will even out over the best of 3.

I've actually started betting on each game after the hero hero picks and before the lines close on pinnacle as Roy and I have been talking about. I think there can be a pretty big edge from this.

I managed to get a bet on LGD to win game 3 after all the hero picks, they had a huge advantage in the hero picks in this game. The hero pick model (not mine) said they had an 80% chance to win based on heroes alone. Lucky I managed to get that one in because LGD ended up losing the best of 3 and this hedged my losses on the series

Here's the output for the last series of the day:


Looks like the last 2 best of 3's are played over 2 days. So I'll do a big update after the games tomorrow.

Here's the output for the finals:


Doesn't look like there's much value going into the finals, I'll keep checking and hope to get some bets down if the odds improve.

I made a couple of small bets yesterday against ViCi earlier today.

I will do a full update, and the results from betting on the international after the finals tomorrow.

Jks I'll do it now. Here's the results for betting on the Dota2 International since the thread started, starting with a $100 bankroll.



Uni has started back up so will have less time to play poker. I'll post results weekly.

I'm working on the soccer program I gotta get it going before the next EPL. When I get it going I'll also start posting updates about profit/loss, we'll see how it goes

I'm also looking at starting up a site for my dota model, so people can put in their own input (pick each team/players) and get the odds and everything.

great stuff!

Shouldn't you make every bet where you have an edge and just bet smaller portion of your bankroll when edge is smaller like Kelly criterion suggests for maximum growth of bankroll?


Good luck, will follow as I like both DotA and poker

Yeah, not 100% sure if I should stack them though, or how I should adjust the kelly stake if I'm stacking them. Maybe I could do quarter kelly and stack?

I have also thought about doing a middle between the handicap (i.e. -1.5) and the moneyline where I have an edge for it. A middle is where you make 2 bets that "overlap" where you can only lose one and can win two. It's like a hedged bet. So I could bet on the underdog at +1.5 and bet on the favorite's moneyline, then I would win both if the underdog won one game and the favorite won two, otherwise I would make a small loss from the hedging (I'd only lose 1 bet).

The shot at 25nl is going really well, soon I'm going to have an underrolled semi-shot at 50nl. By this I mean I will play 25nl and bumhunt 50nl (play 50 if I can get the jesus seat on a huge fish).

I'll do a poker update next week.

You're removing vig from pinny line to calc kelly/edge right?

This is actually what I thought you had to do too to compare the two (the model and bookie) originally too. This would be immensely complicated and inaccurate since the vig isn't necessarily spread evenly over the selections, I thought you would essentially have to model the odds offered and the true odds of the event happening to do this.

It's way simpler than that. You just compare the two (the model and bookie) using EV, take probability of X team winning from the model and the odds from the bookie for X winning as the return, EV = probability of occurring * return - probability of not occurring * loss.

The edge/kelly stake % is given by:

[(Probability * Odds) – 1]/(Odds-1)

So I'm using Probabilty = model output and Odds = bookie price.

Poker results since starting the thread:



July is over, I'm comfortably playing 25nl now and I'm gonna start bumhunting at 50nl. I will probably have less time to play now that uni has started back up but I should keep making progress.

There's another dota game up, this time with some relatively unknown teams. Gonna use quarter kelly for this one.