PSG vs paiN -285, W, +1u
PSG vs IW -285, W, +1u
IWC vs paiN -110, L, -1.1u
PSG vs MAD +130, L, -1u

IWC MAD O 28.5 -110, L, -1.1u
PNG MAD O 28.5 -110, L, -1.1u
-2.3u

Total -1.06u on 39.91 risked

For tomorrow:
Detonation FocusMe vs DWG, +711, 1u
Infinity vs C9, +386, 1u

Totals:
C9 DWG O 23.5 -109
DFM INF U 24.5 -116
DWG INF O 22.5 -115
DWG DFM O 22.5 -115

Eyeing INF vs DWG as well, might fire a half unit on it if the line moves any more.

it's pissing me off i keep forgetting to watch these

You're not missing a ton, group stages have pretty much happened as everyone predicted (aside from Pentanet advancing).

DFM vs DWG, +711, L -1u
INF vs C9, +386, L -1u

C9 DWG O 23.5 -109, W +1u
DFM INF U 24.5 -116, L -1.16u
DWG INF O 22.5 -115, W +1u
DWG DFM O 22.5 -115, W +1u

None of the longshots hit and we're at -1.22u on 46.46 risked

Next phase of the tournament starts up on the 14th, its basically 1 big group of the 6 remaining teams that play a double round robin, with the top 4 advancing to a bracket phase. So more to come then.

For tomorrow:

RNG vs DWG, -109, 1.09u
Pentanet vs MAD Lions, +475, .5u
PSG Talon vs DWG, +344, 1u
C9 vs MAD, +111, 1u

RNG vs DWG O 26.5 -115
PGG vs MAD O 25.5 -115
C9 vs RNG O 25.5 -115
PSG vs PGG U 27.5 -115
C9 vs MAD O 25.5 -115

Got smoked yesterday for -5.1u

For tomorrow:
C9 vs DWG +255, 1u
PSG vs C9 +139, 1u
PSG vs MAD +159, 1u
RNG PGG O 25.5 -115, 1.15u
C9 DWG O 25.5 -118, 1.18u
RNG MAD O 27.5 -116, 1.16u

Dota had a similar issue in late March/early April. After about a year of no cross-region play (except close regions like NA/SA, CN/SEA, EU/CIS with ping issues) there was finally a LAN event with top teams from all regions. Finally we got to see how the top EU and CIS teams were against the Chinese and NA teams. Betting markets favored EU/CIS pretty strongly because they are historically strong regions.

I use two power ranking based models: one for teams and one for individual players. Then I take a hybrid of the two and apply adjustments. For example, if the exact 5-man lineup hasn't played many games together, I'll weight the player model more. I was curious how it would handle this exact cross-region issue. The only concrete adjustment I made for the tourney was to increase the "k-factor" slightly to adjust faster based on what happened. In the end, it did well but I'm not sure my adjustment made much of a difference.

Long story short, I did try to think of an adjustment to make for this lack of cross-region play but I didn't do much. I thought about trying to have a factor for historically stronger regions but Elo/similar should do that already. It probably won't matter in the future much either as we're back to LAN play.

I appreciate the input, I feel like I have something similar to your player based model and I view a team as the sum of its parts. It's been fine for same league play but international/cross region play is tough for me to crack.

Yesterday:
PGG vs C9 +471, L, -1u
PSG vs RNG +302, W, +1.51u
MAD vs DWG +248, L, -1u
C9 vs DWG +275, L, -.5u
PSG vs MAD +150, W, +1.5u

MAD DWG O 26.5 -105, L, -1.05u
PGG RNG O 26.5 -114, W, +1u
DWG C9 O 24.5 -109, W, +1u

+1.46u

Today:
PSG vs RNG +291, 1u
MAD vs DWG +264, 1u
PSG vs DWG +285, 1u
C9 RNG O 26.5 -105, 1.05u
PGG MAD O 27.5 -111, 1.11u

Adding C9 PGG O 25.5 -112

Last day of round robin as we lost another 2.2ish units yesterday, feelsbadman

RNG vs DWG +100 1u
MAD vs RNG +246, .5u
PSG vs C9 -130 1.3u
RNG DWG O 25.5 -115
RNG MAD O 27.5 -115
PSG C9 O 25.5 -120
MAD C9 O 26.5 -120

Much needed 7-0 day for +7.23u. Still at -1.7u for the tournament but our PSG futures are very live as they're in the semis vs RNG. Doubt I'll be hedging it since I dont think the RNG side is +ev, but I have a day or two until lines are out on sites I can bet at so not sure what I'll be on for either semi yet.

nicely done, sadly have been too busy to watch/tail but with you in spirit

Not hedging the +2200 PSG future, taking these though:
PSG +2.5 Maps, -155, 1.55u
PSG +1.5 Maps, +187, 1u
PSG RNG O 27.5 Map 1, -115
PSG RNG O 27.5 Map 2, -115
PSG RNG O 27.5 Map 3, -115
PSG RNG O 27.5 Map 4, -110
PSG RNG O 27.5 Map 5, -110

+.7u but our PSG futures are dead, maybe should have hedged.

Tomorrow:
MAD +2.5 Maps, -135, 1.35u
MAD +1.5 Maps, +244, 1u
MAD ML, +511, .5u

O 23.5 KILLS Map 1, -115, 1.15u
O 23.5 KILLS Map 2, -115, 1.15u
O 23.5 KILLS Map 3, -115, 1.15u
O 23.5 KILLS Map 4, -115, 1.15u
O 24.5 KILLS Map 5, -115, 1.15u

+3.64u, close to a much bigger day as MAD were up 2-1 but lost both match point games. Also bricked 2 kill totals by 2 kills. Cant really complain though.

Overall at +2.64u on 101.04 risked, not counting the -.25u we have on the futures.

Bookmaker doesnt have lines for tomorrows final posted yet, I'll post what I have (if anything) when they put them up.

I think lines tomorrow are pretty fair, so I don't see myself having any action on the final. I have fair as DWG -108 so if you see better than that for the series go for it, but I'll probably sit it out barring any side gaining major steam.

MSI ends and we won 2.64u on 101.04 risked, along with losing .25u on 7.25u risked from futures so we end:

+2.39u on 108.29u

The next split is probably going to start in ~3 weeks away, not sure if I'll keep posting to the void in here but it was fun to do it for an international tournament.

Im messing around with a side project to try and make an esports model,, which games have the best data available? I was GM in starcraft like 10 years ago so i was gonna start there.. but I cant really think of any approach besides some sort of ELO/power ranking model, there arent many other stats to use and they are so game dependent. Anyone got ideas?

I'm not nearly as knowledgeable on esports as some of the other posters in this thread but for data availability I'd start with csgo and lol. ELO is a great model to start with as it's easy to understand and implement, but unlikely to beat the market long term. There was plenty of discussion about this on the crowdsource syndicate thread a few months ago you might be interested to read.

I think most major esports will have some sort of free data warehouse that someone has put together. I don't know anything about SC2 or CSGO betting but I would second the idea of CSGO as that's a game I think you can break down super granularly. What I mean by that is, you can easily break things down from the smallest question of "How often does Team A win a single round vs Team B on Map Y" and build that up into "How often does Team A win a best of 5 maps against Team B".

Some things that might be worth exploring for single round winrate are things like:
-What map is it?
-Is Team A on T side or CT side?
-Is it a pistol round or one of the rounds following?

Then once you have a solid idea of how to get a single round winrate, you can expand that into the map winrate, and then to the match winrate. I'm far from the sharpest bettor on this forum, and totally willing to be told I'm wrong, but those are the types of questions I would ask myself for any new sport I'm trying to model.

My suggestion is to bet on an esport you actually follow and I think data availability is also pretty important. Another thing to consider is the size of the market and how much you're looking to bet. In esports markets, and this is just my impression, I'd say CSGO > LoL > DotA >> Everything else.

Regardless, start with a simple model and build things into it one by one. If you just run Elo on team wins, that almost definitely isn't enough to beat the books but might be a good starting point. It'll probably be pretty close to widely available lines too and then you just need to find some reason why a certain game should be different.

I also echo the other suggestions of breaking the game up into rounds/parts somehow. In CSGO, this is trivial but in a game like DotA or LoL, that can just be the 5min mark/10min mark/15min mark, etc. This also means you need data at this granularity. Dominating a map on CSGO should be worth more than a win by 1-2 rounds and winning a best of 3 2-0 vs. 2-1 are different. Keep in mind, however, that just as in CSGO where winning the pistol round makes you more likely to win the 2nd round, being ahead at 5 minutes in LoL makes it more likely you'll be ahead at 10 minutes. Also, just as in the NFL, some esports have "garbage time" that should be discounted: being up 20k gold at 20 minutes in DotA is basically 100% to win so you shouldn't give a team any more "credit" for increasing their margin of victory after that.

Team Liquid had this, but it doesn't look like it's been updated since forever. Besides, SC2 just seems to be a case of bet on whichever race is overpowered because blizzard can't balance the game, same as it's been since it came out

have a betting odds question

if A wins B game 1 and u get paid 3.2 and
if A wins B game 2 and u get paid 3.15

can i say for A to win 2-0 u will get 10:1 odds?

are you arguing what should happen vs what is fair?

if first two lines are fair then yes that's a fair line because it's going to happen 10% of the time

but whether they offer that is another thing entirely

computing the odds of winning both is easy so from there you can see if the bet to win both makes sense - but in general parlays have a higher vig

there's a lot of inconsistent lines an high vig in esports

you'll often have situations where two different bets happen with the same outcome

let's say team a won game 1 in a bo5

from there one you can bet team a -2.5 for +180
or bet u3.5 total games +200

it's the same exact bet

i'm trying to understand if team A winning 2-0 will land u a 10:1 odds, cos in game 2 u are risking 352 to win 1108