The title pretty much says it. Getting deeper early gives us a compounding advantage across the timeline of the event, but is it worth making slightly -EV decisions to build a stack quickly?

I dont know if there is a consensus on this topic or not.

Strictly from a game theory perspective and playing against a vacuum (opponent(s) won't wildly deviate based on your previous -ev actions), then there is no incentive to ever make -ev decisions.

It really really depends on the structure, especially how deep stacks are to start, how quickly the blinds are going to raise, and how many spots payout.

No not really, but thankfully tourney players are typically either very loose passive or very tight passive in early stages, which enables a good player to open raise hands that would be -ev open raises in a cashgame.

Antes only amplify this effect.

I disagree with the above responses.

ICM is one manifestation that shows that chip EV <> dollar EV. So there are many instances in which it is "optimal" to take a slightly -EV (chip EV) option that actually increases your dollar EV. And many people who have studied it have found that the traditional method of calculating ICM undervalues a chip leaders expected dollar winnings in most situations.

Everybody who has played any decent amount of tournament poker knows that having lots of chips pays dividends in winning more chips, often without the best hand. Smaller stacks don't want to mess with you, you can take down pots with inferior holdings, etc. Tournament poker is a game where "non-linearities" abound.

So I think the answer is Yes. Chips beget chips and getting chips early can definitely pave the way to tournament success.


Edit to add: I realize that I gave the easy answer invoking ICM. I also feel the same is true even ignoring "ICM" considerations.

I guess you are correct, however, it would be calculated in a $ev calc from a game theory perspective I'd think. So what might be a -ev chipev situation is actually +$EV-- i could see this

I'd also imagine that really close -chipev situations may be +$ev, but I'm sure there has got to be some threshold as it doesn't intuitively make sense to me that any reasonably large (subjective statement) -chipev play would also be +$ev.

I don't play tournaments fwiw.

The question was about earlier stages of the tourney. Whosnext is correct if the question was about later stages.

So at beginning of tournaments chipev ~ $ev correct?

I imagine that many of us have seen some of the WSOP Main Event coverage. It is still fresh in my mind and perhaps has colored my previous post.

There are many, many, many players even at the early stages of the tournament (say Day 1) who are "willing" to give up their chips at any sign of any aggression from a player with significantly more chips than they have.

Of course, much of this can rightly be labelled "value of aggression" in poker tournaments. I agree completely. But I would argue that this widespread phenomenon also speaks to the "volatility is good" argument OP brought up.

In tournaments, there are a great number of players (not only in the main event) who choose not to take many risks. And this applies largely early in tournaments. Some have argued that they would fold even if they knew they had a 60/40 edge in the hand.

So, a whole bunch of things are being co-mingled here. Aggression, taking free chips from inexperienced/scared players, etc. But I do think that there is general merit to the "volatility is good" argument, even early in tournaments.

I guess we could do a thought experiment. You have a 50/50 flip on the very first hand of the Main Event. If you win you will have 120,000 chips (and be table captain). If you lose you are out. I have to think that the overall outcomes of the 120K/0 mix is better than the set of outcomes from holding 60K chips.

Yeah I would also imagine taking the 50 50 is better but I don't have my experience. So how do we mathematically prove this?

The other example of the situation where you say people are essentially over folding would be exploitative imo and not really part of the question.

I am not so sure. My hypothesis is that a large fraction of the over-folders are especially over-folders vs players with large stacks. Meaning that the over-folding player phenomenon is a "hidden" benefit of having a large stack.


Edit to add: I guess that there are two aspects to OP's query. (1) "Theoretical" which can encompass ICM and ICM-type non-linearities. (2) "Practical" elements including over-folders and the like.

thanks for feedback guys, whoesnext and brokenstars you guys are saying everything i was trying to figure out. also prob should have framed the question as "slightly -chip ev decisions"

As whoesnext said, this 'compounding advantage' that depth gives us is clearly worth a ton of EV, but its difficult to quantify that advantage. Intuitively it seems like the expectation that we gain from this advantage outweighs the chip-EV we lose making a volatile play early on. The decision may be slightly losing in chip EV, but actually have a positive expectation when this advantage is factored in (as brokenstars said)

If someone can double up quick, and they know how to wield a big stack across the timeline of the event their EV skyrockets (even in a 'nash' setting). So while there is an exploitative aspect to this advantage, it is also a structural advantage

I tried to do something quick and dirty to investigate the ICM aspect of OP's query. I assumed:
- N players
- buyin = 1
- 15% of players get paid
- last payout = 1
- each payout jump is same pct (so I have to solve for this pct based upon N)

Half the time hero "starts" with 2 times as many chips as everyone else (I eliminate one person so total chips is always N) and half the time hero immediately busts.

I calculated Hero's ICM (which is fairly easy in this case) and compared it to 1, which is the ICM of each of the N players having equal stacks at the beginning of the tourney.

I was "hoping" to find that the double-or-bust Hero would have a higher ICM than starting Hero (before tourney starts).

I tried the following cases for N: 100, 250, 500, and 1000. In each case the double-or-bust Hero's ICM is about 0.96 (average of around 1.92 and 0) which, of course, is less than 1.00.

Maybe the payouts have to be more top-heavy for the double-or-bust Hero to have the advantage over starting Hero.

I will noodle it some more and maybe try different assumptions if any come to mind. If anybody has any thoughts or suggestions, please chime in.

I thought one of the disadvantages of OCM is that most people believe it doesn't appropriately model the big stack advantage.

Are you simply using ICM to answer the theoretical portion of the question because it's the best theoretical model we have?

I was using the traditional ICM formulation since it is easy to code up and apply to my scenarios. I now believe that there is unlikely to be a $EV advantage to early tournament volatility using the traditional ICM formulation.

Having said that, we know that the traditional ICM formulation significantly undervalues the "true" $-value of large stacks. There have been threads in this forum as well as a few academic papers on that topic. So it seems like it may still be theoretically possible for accepting/seeking more volatility to lead to increased $-value.

Remember bubble factors? A bubble factor reflects the non-linear relationship between chip values and dollar values in a tournament. And, most important for our purposes, a bubble factor can never be less than 1. Meaning that you can never increase your dollar EV at the same time your chip EV decreases. So it appears that that avenue is shut down no matter what $EV model is used.

Beyond the theoretical aspect, as mentioned in previous posts above, there is undoubtedly a practical advantage to having a large stack due to other players over-folding to a large stack at the table in tournaments, even early in tournaments.

Of course, measuring/analyzing/simulating/modeling this practical advantage of having a large stack is quite challenging.

Ok here’s the thing:

When we shove wide (relative to cash game Nash) on the bubble, we’re not making a negative chip ev play. We’re earning chips that would not be ours to earn in a cash game because our opponents should fold positive chip ev hands.

This is how big stacks earn $ev and chip ev near the bubble.

Gotta run or I would elaborate.

Making -EV decisions never makes sense. But one may choose to take the most +EV line by taking a close race or committing chips because they are a dog in the hand but have great pot odds. In prestigious tournaments with good slow structure one may give up the most +EV lines to survive and grind out their advantages over time using lower variance lower EV approach.

A big stack near the bubble means the chip you risk are worth less than the chips your opponent risks.

When you talk about EV in tournaments what you are talking about is $EV. No, you never make -$EV decisions. ChipEV is an irrelevant metric.

If taking a boom-or-bust approach is slightly worse than playing "normally", the high-variance style may still be the way to maximize your hourly rate. This might make it preferable if you have the opportunity to immediately jump into another tournament or play a cash game.

great point, i think time is often undervalued in poker

interesting, i thought it was a relevant metric. my understanding was that chip EV is basically the most practical way of looking at decisions early on in events. then later on we switch to $EV once ICM considerations become more relevant

I addressed the way the question was phrased above, but basically i meant to say 'does a big-stack advantage through the timeline of the event create enough $EV to give slightly -chipEV decisions a positive expectation?" or something like that. Is there enough hidden EV in this advantage?

I guess it would be a direct function of how players deviate vs. larger stacks and therefore is largely subjective.

If it was all bots playing vs bots, then I doubt there is any advantage.

there is an exploitative aspect to this advantage, but it is also a structural advantage, present in a nash setting

Then under the assumption that your assumption is correct (which we're currently having trouble proving). Then under what structural constraints do you think it would be more of an advantage?