|
motbob
United States12546 Posts
After I won Liquibet season 1, someone sent me a PM asking for tips on winning. I sent him a long PM back, and I feel like it's only fair to reproduce and expand upon that PM here. This is a highly advanced type of blog: a brag/useful blog hybrid.
How do I win Liquibet?
The most important thing to remember is that points don't matter in Liquibet nearly as much as rank matters in Liquibet.
A correct bet that propels you up to the top of the Liquibet rankings is much more valuable than a correct bet that leaves you in the same position you were before relative to other people. Predicting that IntoTheRainbow will lose is much more valuable than predicting LiberoPrime will lose.
If you want to maximize the EV of your final score, betting on who you think will win every time is a fine strategy. However, if you want to be #1 on the rankings at the end of the season, you're often going to have to bet on people who you think have less than a 50% chance of winning.
Let me explain. In a given Liquibet season, there are going to be upwards of 7000 people competing from beginning to end. When people make a decision about who to vote for, they draw from the same sources: articles, past VODs, Blizzard rank, etc. No one has hidden knowledge on who will win the next match.
What this means is that, somewhere out there on Team Liquid, there is someone with roughly the same predictions as you as to who will win their games. In fact, if you put a lot of time into watching SC2, there are likely a LOT of people with the same predictions as you.
You need to separate yourself from those people in some way. Taking risks is that way.
![[image loading]](http://imgur.com/S97n9.jpg)
FUCK, I saved as a .jpeg
In the above graph, you are a well-educated watcher of SC2. You know a lot about the game.
In strategy A, you bet on who you think will win every time. At the end of the season, you'll likely have a pretty high score, and you'll probably be in the top 1000 or so players. But you have very little chance of being #1. In fact, your chance is almost nonexistent.
If you follow strategy B, you are taking a lot of risks. Your score at the end of the season might be garbage. But it also might be high enough to net you #1.
Here's an example of a smart risk to take: Fruitdealer vs BoxeR[arnc] last season. The TL vote was almost 100% Fruitdealer, a fact that was easily predictable, given that FD was GSL 1 champ and the fact that no one knew who BoxeR was at the time.
Yet BoxeR was near the top of the Korean ladder for a long time, and Fruitdealer hadn't really ever faced a Terran who played really well. Surely the chance of BoxeR winning could be predicted to be a lot closer to 50% than 1%. So the right move would have been to bet on Boxer and hope to move way up in the Liquibet rankings.
To sum up: here's the formula to winning Liquibet.
1. Take calculated chances early and have most of them hit.
2. If you're ahead at the quarterfinals stage, bet on who you think will win.
3. If you're behind at the quarterfinals stage, take crazy chances.
4. Get all your bets at the quarterfinals stage and above correct.
These 4 steps are almost impossible to get right. But how else are you going to beat over 7000 competitors? Liquibet involves a lot of chance. I'm unbelievably lucky to have won last season.
Hope this helps.
  
|
Whoah! I'm gonna win LB now ^_^ thanks.
|
I'm still convinced you made a deal with the mods to be allowed to cheat in Liquibets, in exchange for your soul.
Nice guide, though. Very informative.
EDIT: WTF A STALKER ICON??
|
9070 Posts
|
Yeah I'm confused at your lack of icon, liquibet is hard to win yo!
Edit: Misread, my mind just assumed you meant broodwar liquibet season 1.
|
Motbob, this is really helpful, and makes perfect sense. I recently made a risky bet that paid off 
|
Thanks. I was indeed wondering what your Liquibet-strategy was. This makes a lot of sense.
However, I like to use Strategy A. I think it's much more fun to see how well I can predict the results, even if it didn't lead to high rank.
|
|
|
Nice stalker icon wtf, I want one!
|
lol icon!
I like how we immediately focused on that icon. >.> I think I'm getting used to just looking at it.
|
when i first glanced through this thread, something seems wrong somewhere, i cant detect where is it.. until when ppl mentioned the stalker icon.. Omg! SC2 icons now! if only i can be a baneling =D
|
303 9 motbob
more like how to not win liquibet!
|
apparently we're switching over to sc2 icons 1 at a time
|
It's similar to how i play LB1, except with less risk, since there are so many more statistics given for BW its easier to pick them.
Basically I will look at names, matchup history in terms of race, individual performances in the recent 10-15 games overall as well as in that matchup, and who they have beaten/lost to. Then I check the map they'replaying on and the race matchup history on the map, and then after all that I will factor in the risk I need to take in order to move up. Calculated risk.
but seriously, stalker icon is fucking epic wow
|
Whoever did your stalker didn't make it yellow enough, you should tell them to fix it.
|
Roffles
Pitcairn19291 Posts
Meh, I'm not a fan of that stalker icon for some reason.
Btw, it's usually a good thing to bet against Liquid players, just like Haypro last night.
|
And so, with this, you will never win liquibet again. No one ever escapes the curse 
|
On November 24 2010 00:17 Roffles wrote:
Meh, I'm not a fan of that stalker icon for some reason.
Btw, it's usually a good thing to bet against Liquid players, just like Haypro last night.
I think you're right :<
I already voted for everyone in ro64. But will probably vote against a huge odd in ro32 onwards.
|
Hyrule19057 Posts
On a completely unrelated note....you have a Stalker?
On a related note, now that everyone knows, if everyone tries this, won't it balance out and thus be an invalid strategy?
|
motbob
United States12546 Posts
On November 24 2010 00:35 tofucake wrote:
On a completely unrelated note....you have a Stalker?
On a related note, now that everyone knows, if everyone tries this, won't it balance out and thus be an invalid strategy?
Even if everyone does it, it's still the best strategy.
In fact, as more people do strategy B, strategy A becomes worse and worse.
|
On November 24 2010 00:28 shannn wrote:Show nested quote +On November 24 2010 00:17 Roffles wrote:
Meh, I'm not a fan of that stalker icon for some reason.
Btw, it's usually a good thing to bet against Liquid players, just like Haypro last night. I think you're right :< I already voted for everyone in ro64. But will probably vote against a huge odd in ro32 onwards.
I have no idea why, but I voted for haypro. I don't know what I was thinking, just now I was looking over my choices and it shocked me...
|
Man who cares about Liquibet in this thread, you have a SC2 icon.
j/k
+ Show Spoiler +Gratz on your win and the strategy you employed !
|
You got this icon because you have won the liquidbet ?
|
motbob
United States12546 Posts
No, I don't know why I got the icon.
|
Cool
The colour is kinda off on your icon :/
|
I've known this for a long time. That's basically how every competition works where only the top get a prize. I took some ridiculously speculative stocks in my stock market competition.
|
Nice guide, but I think you should have just called this thread "I have a stalker icon".
|
Well, looking at my picks from last season, I kinda ended up doing half of that B and half of A strat which ended up with me failing almost the whole of ro16 picks ^^ I still ended 22nd so it wasn't all that bad, but maybe I should fully apply that B strat
|
With the dubious credentials of being one of motbob's closest competitors last season, all I can say is that the man is 100% correct. Betting with the herd every time is a surefire way to mediocrity.
That said, I rarely bet against my honest prediction, and only in close cases. I believe the herd is often very wrong :D
|
Do you get epic icons when you win the liquibet?
Gotta win now.
Autovote protoss - antiautovote zerg, nonames and old schoolers over the popular guys. Has to work.
|
|
|
He does all the GSL LR threads during quali's, I think he deserves it
|
Hahahaha, I fully expect there to be a ton of weird bets from now on. As a result, people who bet based on past performances (like me) might not win, but I will get higher in the ranks
|
Wait, is this Motbob's strategy to win the liquidbet: Telling others how not to win the liquidbet?
I am always a person who bet on the more favorable player. Ended up pretty high each season like a top10, top50, and top 200 (I am not entirely sure).
You just need to vote every bets and know the players. Also you might want to take a few risk but only if you are close to the top and only do it at the end of betting season.
To be honest, Motbob's strategy is not bad. But if you follow it, you will probably fail very early in the season and won't have a chance to catch up.
FYI: I missed the first 3 bets this season and decided not to participate in liquidbet this season.
BTW Motbob: What are your current ranking this season?
|
I peaked at 15 in liquibet... thats because I accidentally picked foxer when I meant to pick someone else... that's probably gonna be the high point of my liquibet lol.
+ Show Spoiler +Could I pl0x have a phoenix icon? thnx.
|
I'm number 4 right now and all I did was vote for people I thought would win...maybe I just have poor decision-making and good luck. >o<
I peaked at number 6 the first time I tried after missing like 3 votes and literally not knowing who half the players were haha... It probably is just good luck :/
|
I was like 10th before the first semifinal of last season of liquibet2 and got it wrong. Jumped to 18th place when i could have been second right after motbob before the finals 2 points behind him and 1 point in front of third.
I would have voted BoxeR though as motbob clearly said he voted Nestea in the comments of the liquibet.
=> if you have a particular opponent in the final stages of the season (maybe you're competing against friends) and you're behind, pick the choice your opponent didn't pick, even if his chances of winning are 1%. Seems obvious but A LOT of people don't do it.
|
On November 24 2010 03:56 tyr wrote:
I was like 10th before the first semifinal of last season of liquibet2 and got it wrong. Jumped to 18th place when i could have been second right after motbob before the finals 2 points behind him and 1 point in front of third.
I would have voted BoxeR though as motbob clearly said he voted Nestea in the comments of the liquibet.
=> if you have a particular opponent in the final stages of the season (maybe you're competing against friends) and you're behind, pick the choice your opponent didn't pick, even if his chances of winning are 1%. Seems obvious but A LOT of people don't do it.
OMG your post count! Dun dun dunnn.
Okay, I'm an idiot...but I couldn't resist.
Edit: but how do you tell what your opponent picked? Don't you just have to hope he announces it or something?
|
Nice post, but I have to wonder if this will change the metagame of LB. If people start taking more and more risks, dont you now increase the variance of Strategy A in your diagram AND decrease the points needed for #1, and so increasing the viability of "standard play" (picking the favorites)? I wonder where the equilibrium is.
EDIT: Nevermind, it doesn't actually change the variance, this is nonsense. Great read!
|
Liquibet is about getting the most possible points, the other people's points shouldn't concern you until you want to know your ranking. You're talking about trying to predict the upsets, but seem to say you should always go for upset bets, when the reason its an upset is because it doesn't have 90%+ of the time. Therefore a lot of people picking the "favourites" will get a decent ranking, and only those clued up//lucky enough to guess good upsets will do really well PROVIDED they correctly predicted the other 95% of games that weren't upsets correctly.
Basically you have to know your shit or get lucky or both.
|
Interesting material.
I think the Stalker icon looks pretty cool, but personally I wanna be a Lurker forever.
|
United States4053 Posts
Let's say I have a clone of myself who follows strategy B throughout the season, while I use strategy A. Since he's always betting on the guy with <50% chance of winning, in each bet I am expected to gain a point differential on him.
Say there's a 1 point bet of player X versus player Y and player X has probability p > 0.5 to win. If I bet on X and my clone bets on Y, then my expected point gain is p and my clone's expected point gain is 1-p. So the point differential between our gains is 2p-1 in my favor, which is guaranteed to be positive.
So how does he have a better chance of winning Liquibet? >_____>
|
This thread is now about the stalker icon.
Motbob, I am curious how many "upset" bets did you succeed with?
I mean where you one of the few that voted Foxer beating FD(not clear if that was an example or you actually did) and what was the biggest upset bet that you got away with?
(PS sorry for derailing the thread from the stalker icon  )
|
motbob
United States12546 Posts
On November 24 2010 06:02 infinitestory wrote:
Let's say I have a clone of myself who follows strategy B throughout the season, while I use strategy A. Since he's always betting on the guy with <50% chance of winning, in each bet I am expected to gain a point differential on him.
Say there's a 1 point bet of player X versus player Y and player X has probability p > 0.5 to win. If I bet on X and my clone bets on Y, then my expected point gain is p and my clone's expected point gain is 1-p. So the point differential between our gains is 2p-1 in my favor, which is guaranteed to be positive.
So how does he have a better chance of winning Liquibet? >_____>
?
When did I say strategy B means always betting on the guy with <50% chance of winning? That would be an incredibly stupid strategy.
|
The point is betting on people who may be perceived as having a <50% win chance (but still in the area of >30%) but who a tiny fraction of the betting community actually bet on. This way, if they win, a lot of people win and no one really gets any higher, but if you win, you are one of the few who actually gained points.
It's fairly all or nothing, and most of the people who do it will be in the lower area of the points curve, but a few will be way higher than everyone else.
In short: Liquibet is srs bsns
|
I think this is the betting equivalent of "screwing up the metagame"
|
So basically, lets say player A has 30% chance of winning against player B (quite likely given that the game is still young and no one is untouchable) but 90% of people will bet on player B.
Identify these "arbitrage" opportunities, pick wisely and PROFIT (i.e. win Liquibet)
|
i am now sticking to my guts - still voting for players who have >50% chance of winning because people in TL will now vote to those who have <50% chance of winning after this blog. lol
anyways, thanks for sharing your insights. pretty nice, i hope i can win the LB also, top 20 would make me happy
|
United States4053 Posts
On November 24 2010 08:03 motbob wrote:Show nested quote +On November 24 2010 06:02 infinitestory wrote:
Let's say I have a clone of myself who follows strategy B throughout the season, while I use strategy A. Since he's always betting on the guy with <50% chance of winning, in each bet I am expected to gain a point differential on him.
Say there's a 1 point bet of player X versus player Y and player X has probability p > 0.5 to win. If I bet on X and my clone bets on Y, then my expected point gain is p and my clone's expected point gain is 1-p. So the point differential between our gains is 2p-1 in my favor, which is guaranteed to be positive.
So how does he have a better chance of winning Liquibet? >_____> ? When did I say strategy B means always betting on the guy with <50% chance of winning? That would be an incredibly stupid strategy.
It's an example, but even for one bet where I bet on guy with >50% chance, clone bets on guy with <50% chance, on average I will get points. So how does losing points on average translate to better chance of winning Liquibet?
|
On November 24 2010 11:01 infinitestory wrote:Show nested quote +On November 24 2010 08:03 motbob wrote:On November 24 2010 06:02 infinitestory wrote:
Let's say I have a clone of myself who follows strategy B throughout the season, while I use strategy A. Since he's always betting on the guy with <50% chance of winning, in each bet I am expected to gain a point differential on him.
Say there's a 1 point bet of player X versus player Y and player X has probability p > 0.5 to win. If I bet on X and my clone bets on Y, then my expected point gain is p and my clone's expected point gain is 1-p. So the point differential between our gains is 2p-1 in my favor, which is guaranteed to be positive.
So how does he have a better chance of winning Liquibet? >_____> ? When did I say strategy B means always betting on the guy with <50% chance of winning? That would be an incredibly stupid strategy. It's an example, but even for one bet where I bet on guy with >50% chance, clone bets on guy with <50% chance, on average I will get points. So how does losing points on average translate to better chance of winning Liquibet?
He just means that you'll sometimes have to take a risk if you want to get ahead. He's not telling you always to pick the one that seems less likely to win -.- he's just telling you that you have to get lucky and vote on people on whom others won't (sometimes), or else it will be impossible to win.
|
United States4053 Posts
On November 24 2010 11:07 Z3kk wrote:Show nested quote +On November 24 2010 11:01 infinitestory wrote:On November 24 2010 08:03 motbob wrote:On November 24 2010 06:02 infinitestory wrote:
Let's say I have a clone of myself who follows strategy B throughout the season, while I use strategy A. Since he's always betting on the guy with <50% chance of winning, in each bet I am expected to gain a point differential on him.
Say there's a 1 point bet of player X versus player Y and player X has probability p > 0.5 to win. If I bet on X and my clone bets on Y, then my expected point gain is p and my clone's expected point gain is 1-p. So the point differential between our gains is 2p-1 in my favor, which is guaranteed to be positive.
So how does he have a better chance of winning Liquibet? >_____> ? When did I say strategy B means always betting on the guy with <50% chance of winning? That would be an incredibly stupid strategy. It's an example, but even for one bet where I bet on guy with >50% chance, clone bets on guy with <50% chance, on average I will get points. So how does losing points on average translate to better chance of winning Liquibet? He just means that you'll sometimes have to take a risk if you want to get ahead. He's not telling you always to pick the one that seems less likely to win -.- he's just telling you that you have to get lucky and vote on people on whom others won't (sometimes), or else it will be impossible to win.
I agree that there's luck involved, but I feel like somewhere in this argument there is a severe logical fallacy. >___>
|
Hyrule19057 Posts
Oh yeah, there's an addendum to this method: Never, ever vote for ACE.
|
Seeing as I am doing much better in SC1 liquibet (which I watch almost none of), I can draw the conclusion it is better to be lucky than educated. Obviously the OP knows what he is doing, how else would he get the icon!!! I want to see the SC2 liquibet trophy though, give me something to aim for!
|
The idea, infinitestory, is that EV in terms of points doesn't matter, only EV in terms of getting a Stalker icon, and point EV being positive/negative doesn't guarantee that stalkericon EV is the same sign. To simplify the scenario, in the last pick of the season, unless you are in 1st place, your stalkericon EV is 0 if you are making the same bet as anyone ahead of you, so you have to sacrifice some point EV to give yourself a chance to win if you are in 2nd place. I'm not really sold that motbob's idea is best, though, since I don't think TL netizens are particularly adept at picking winners.
|
I followed a similar strategy to decent success(58th overall, top 5 before ro4) in GSL2. Although I did pick the 'favorite' over likely-upset multiple times.
I think the biggest killer to me is not knowing the players well enough. Given the information, I rarely feel confident one way or another (esp past ro64). Like your FD v BoxerPrime example, I thought it was a HUGE upset- only seeing his ro64 game- but after seeing more games and his KR ladder rank etc it's not really that big of an upset. I need to fix that before I even consider higher liquibet strategies.
Oh also, getting rocked hard this season lol. Sunk 50%, so pretty much no better than guessing. I guess I did watch less than last season, but I still thought I had enough knowledge to make the right pick haha. Granted I got cute with some underdog picks but have mostly been plain wrong.
|
|
|
|
|
|
EPT
